Abstract: We present an in depth report into the cryptocurrency exchange ecosystem. The market is broken down by almost all the possible characteristics (Exchange type, exchange region and trading pairs). The robustness and authenticity of exchanges are evaluated  using metrics such as web traffic, average trade sizes, order book depth, security polices and price reliability. The report was produced by CryptoCompare and uses the CryptoCompare’s Aggregate Pricing Index (the CCCAGG), for much of the analysis.

 

(Note: Current CCCAGG Constituent Exchanges, Sized by 24H Volume)

 

Please click here to download a PDF version of CryptoCompare’s report

 

Executive Summary

Major Exchange News in October

• Bitstamp was acquired by Belgium-based Investment Firm NXMH for ~400 million USD according to reports.
• Cryptoassets on Gemini are now fully insured with Aon.
• Coinbase adds 0x to its trading platform as well as USDC after announcing its collaboration with Circle on the CENTRE Consortium.
• Korean exchange Bithumb starts a new DEX, while Huobi and OKEX list stablecoins GUSD, TUSD, PAX and USDC.
• Chainalysis will help Binance comply with anti-money laundering (AML) regulations around the globe, and
• Coinfloor becomes the first exchange to obtain a Gibraltar license.

Exchange Market Segmentation

Spot volumes constitute less than three quarters of total market volumes on average (less than 7 billion USD) compared to futures volumes (3.2 billion USD). BitMEX and BitflyerFX average more than one quarter of total volumes while traditional exchanges such as CME and CBOE constitute just under 1%.

Within total spot volumes, exchanges with taker fees represent approximately 90% of the exchange spot market volumes, while transaction-fee based and no-fee exchanges represent the remaining 10%.

Exchanges that offer fiat to crypto pairs constitute just under a quarter of spot market volumes on average (~2 billion USD) while exchanges that offer only crypto to crypto pairs constitute approximately three quarters (~4.7 billion USD). In terms of exchange count however, approximately half of all exchanges offer fiat to crypto pairs.

Transaction-Fee Mining Volumes

The top trans-fee mining exchange by average 24h volume was EXX (160 million USD), followed by Coinex (114 million USD) and Coinbene (113 million USD). The total average 24h-volume produced by trans-fee mining associated exchanges on CryptoCompare totals just over 550 million USD. This constitutes approximately 10% of total exchange volume over the last 30 days.

Decentralized Exchanges

The total average 24h-volume produced by the top 5 decentralized exchanges on CryptoCompare totals just under 2.4 million USD. This constitutes just 0.4% of total exchange volume. The top 3 on CryptoCompare by 24h volume include Waves Dex, IDEX and Dex.

Volume, Pairs and Coins

Binance remains the top exchange in terms of 24h volume with an average of 977 million USD. This is followed by OKEX (405 million USD) and Bitfinex (368 million USD). Yobit offers the highest number of pairs at 7,032, followed by Cryptopia (4,321) and CCEX (2,140).

Bitcoin to Fiat Volumes

The US Dollar represented half of BTC fiat trading on average over the past 30 days, followed by JPY (21%) and KRW (16%). Bitcoin trading to Korean Won (KRW) increased sharply after the 7th of October. The pair previously represented a tenth of bitcoin trading among the top 5 fiats on average. Between the 7th and 15th of October it represented a third on average, a 230% increase stemming from Korean exchange Bithumb’s spike in trading volumes.

Country Analysis

Maltese-registered exchanges produce the highest total daily volume at just under 1.4 billion USD, followed by those based legally in South-Korea (~840 million USD) and Hong Kong (~560 million USD). Among the top 10 volume-producing countries, the highest number of large exchanges (with significant volume) are based legally in the USA, the UK and Hong Kong. Binance and OKEX represent the vast majority of Malta’s volumes, while Bithumb and Upbit dominate in South Korea.

Trade Data Analysis

CoinEx, a well-known trans-fee mining exchange, has a significantly higher trade frequency and lower trade size than other exchanges in the top 25. This may point to algorithmic trading, given its almost 176 thousand trades a day at an average trade size of 125 USD. In contrast, Bithumb and HuobiPro had an average trade size of just under 3,000 and 1,500 USD respectively and significantly lower trades per day (12-18 thousand).

Web User Analysis

IDAX and CoinBene appear to have lower average daily visitors compared to similarly sized exchanges by daily volume. Binance has the highest average daily visitor count, in line with its high trading volumes. Meanwhile, exchanges such as Coinbase, Cex.io and Bittrex have significantly greater numbers of daily visitors than other exchanges with similar daily volumes. ZB and EXX attract significantly lower daily visitors than similarly-sized exchanges.

Order Book Analysis

ItBit, Kraken and Bitstamp have relatively more stable markets compared to exchanges such as CoinEx, ZB and Coinbene. These exchanges appear significantly less stable given their relatively low average order book depth values over the specified period of analysis.

Exchange Security

Out of the top 100 exchanges by 24h volume, only 86% have both a public privacy policy and a terms & conditions page. A third of top exchanges store the vast majority of users’ funds in cold wallets. Exchanges itBit, Coinfloor, Bitfinex and Coinbase are among those that store the highest proportion of users’ funds offline. As a proportion of the top 100 exchanges, 11% have been hacked in the past.

KYC

Just under half of top exchanges impose strict KYC requirements, while more than a quarter do not require KYC.

Total Exchange Volumes and Market Segmentation

This section aims to provide a macro view of the cryptocurrency exchange market as a whole. An area of interest is the proportion of spot trading vs futures trading historically. We will also assess the relative proportion of exchange volumes that represent exchanges that charge fees, as well as those that implement models with no-fees or trans-fee mining. Finally, we will take a look at exchange volumes that represent crypto-crypto exchanges versus those that represent fiat-crypto exchanges.

Historical Spot vs Futures Volumes

Spot volumes constitute three quarters of total market volumes on average.

Total spot volume averaged less than 7 billion USD, while futures volume averaged over 3.2 billion USD over the period of analysis.

Futures exchanges such as BitMEX (XBT to USD perpetual futures) and BitflyerFX (BTC to JPY futures) average just under a quarter of total cryptocurrency market volumes. Traditional exchanges such as CME and CBOE trading bitcoin futures, only constitute a very small proportion of the total market at just under 1% on average.

Historical BTC to USD Futures Volumes

BitMEX’s Perpetual Bitcoin to USD Futures volumes continue to dominate the Bitcoin to USD futures market

When compared to CME’s and CBOE’s futures volumes, BitMEX has represented an average of just over 90% of the market over the last month.

Historical Spot Volumes Segmented by Predominant Fee Type

Exchanges with taker fees represent approximately 90% of the exchange spot market volumes.

On the other hand, exchanges that implement transaction-fee mining represent just over 9% of the total spot market, while those that offer no-fee spot trading represent just under 1% of the market.

Historical Crypto to Crypto versus Fiat to Crypto Exchange Spot Volumes

Exchanges that offer fiat to crypto pairs constitute just under a quarter of spot market volumes on average.

Adjusted Historical Spot Volumes

The cryptocurrency exchange market trades an average of 5.26 billion USD in adjusted volumes over the period of analysis.

Adjusted spot volumes exclude all exchanges that operate trans-fee mining or no-fee trading models.

Historical BTC to Fiat Spot Volumes – Top 5 Fiat Currencies

Bitcoin trading to Korean Wan (KRW) increased sharply from the 7th of October.

BTC to KRW previously represented a tenth of bitcoin trading among the top 5 fiats on average. Between the 7th and 15th of October it represented a third on average, a 230% increase. This increase stems from Korean exchange Bithumb’s spike in volumes.

Proportion BTC Trading to Various Fiat Currencies

The US Dollar represented half of BTC fiat trading on average over the past 30 days, followed by JPY (21%) and KRW (16%).

Summary of Volumes, Coins and Pairs

Top Exchanges by Average 24H Volume in USD

Exchange	24H volume (USD million)	Coins	Pairs
Binance	977.5	160	408
OKEX	405.0	171	511
Bitfinex	368.5	96	281
Bithumb	323.2	13	13
HuobiPro	310.2	128	293
HitBTC	295.2	427	889
ZB	247.6	58	167
Upbit	211.0	132	261
Bibox	208.9	87	210

Top Exchanges by Number of Pairs

Exchange	24H volume (USD million)	Coins	Pairs
Yobit	27.7	1,180	7,032
Cryptopia	3.5	785	4,321
CCEX	0.1	628	2,140
EtherDelta	0.2	2,058	2,059
HitBTC	295.2	427	889
TradeSatoshi	0.1	200	840
Bittrex	49.1	514	637
Livecoin	12.5	249	595
WavesDEX	0.9	163	592
IDEX	0.7	563	563
OKEX	405.0	171	511
Kucoin	10.1	189	450
Binance	977.5	160	408
Gateio	48.8	172	358
Zecoex	1.4	119	342

Historical 24h Volume – Top 8 Exchanges

The top exchange by 24h spot trading volume was Binance with an average of just under 980 million USD.

By average 24h volumes, Binance was followed by OKEX and Bitfinex with volumes of 405 million and 368 million respectively.

Bithumb saw a 356% spike in trading volumes from an average of 140 million USD to an average of 640 million USD after the 7th of October. This follows after Singapore-based BK Global Consortium bought a controlling share in the exchange.

Bitfinex saw a spike in volumes towards the 15th of October as the Bitcoin premium on Bitfinex vs Coinbase reached an all-time high of 11.28% according to CrypoGlobe.

Month on Month Average 24H Trading Volume – Top Exchanges

Average Bithumb volumes increased 187%, while those for Binance and OKEX dropped by 8% and 35% respectively

Korean exchange Bithumb saw a significant increase in average trading volumes from 96 million USD between August/September to 276 million between September/October. Meanwhile, Binance’s volumes over the same time period dropped from 974 million USD to 893 million USD. Finally, the 2nd largest exchange by 24h volumes, OKEX, saw trading volumes drop 655 million USD to 423 million USD.

Country Analysis

Exchanges maintain operations in a variety of countries, in order to serve the wider global community of cryptocurrency traders. They often change legal jurisdiction to avoid regulation in countries that might restrict their abilities to conduct business as they wish. The following country analysis aims to highlight the top 10 legal jurisdictions by the total 24h volume produced by the top exchanges legally based in each jurisdiction.

Top 10 Exchange Legal Jurisdictions – 24h Volume vs Exchange Count

Maltese-based exchanges produced the highest total daily volumes, while the highest quantity of top exchanges are based in the USA and the UK.

Maltese exchanges produce the highest total daily volume at just under 1.4 billion USD, followed by those based legally in South-Korea (~840 million USD) and Hong Kong (~560 million USD). Among the top 10 volume-producing countries, the highest number of exchanges (with significant volume) are based legally in the USA, the UK and Hong Kong.

Top 10 Exchange Legal Jurisdictions – Constituent Exchanges by Impact on Volume

Binance and OKEX represent the vast majority of Malta’s volumes, while Bithumb and Upbit dominate in South Korea.

Top 10 Exchange Legal Jurisdictions – Constituent Exchanges and Count

 

Well-known USA-based exchanges include Coinbase, Poloniex, and itBit, while those in South Korea include Upbit, Bithumb and Coinone.

Hong Kong exchanges include HitBTC, CoinEx and Bit-Z, while those in more remote jurisdictions include HuobiPro in the Seychelles, ZB in Samoa and Coinbene in Vanuatu.

Pair Offering Analysis

The following analysis aims to highlight both the total volumes produced by crypto-crypto vs fiat-crypto exchanges as well as the total number of exchanges that fall within each category.

Crypto to Crypto vs Fiat to Crypto – Average 24H Volume and Exchange Count

On average, exchanges that offer only crypto-crypto pairs constitute approximately three quarters of the total spot trading market (~4.7 billion USD)

Those that that offer fiat-crypto pairs constitute only a quarter of the total exchange market (~2 billion USD) on average. In terms of exchange count, approximately half of all exchanges offer crypto-crypto.

Trade Data Analysis

This analysis aims to shed light on the trading characteristics of given exchange. It helps to answer whether an exchange’s volumes might be the product of consistently large trades, or the product of many small trades which may suggest the use of algorithmic trading or bots.

Average 24H Trade Frequency vs Average Trade Size – Top 25 Exchanges

CoinEx, a well-known trans-fee mining exchange, has a significantly higher trade frequency and lower trade size than other exchanges in the top 25.

This may point to algorithmic trading, given its almost 176 thousand daily trades at an average trade size of 125 USD. In contrast, Bithumb and HuobiPro had an average trade size of just under 3,000 and 1,500 USD respectively.

Average 24H Trade Frequency vs Average Trade Size – Top Exchanges

	Exchange	AVG 24H Volume (Millions)	Average Trade Size (USD)	Trades in 24H (Thousands)
1	Binance	977.5	950	95.7
2	OKEX	405	701	48.5
3	Bitfinex	368.5	1,438	38
4	Bithumb	323.2	2,788	12.4
5	HuobiPro	310.2	1,483	18.7
6	HitBTC	295.2	2,873	12.1
7	ZB	247.6	702	29
8	UPbit	211	732	22.5
9	Bibox	208.9	1,253	16.4
10	EXX	159.9	1,134	24.1
11	BitZ	143.9	2,333	8
12	IDAX	131.5	520	37.4
13	CoinEx	113.6	125	175.6
14	CoinBene	113.2	298	35.2

Web Traffic Analysis

This analysis examines the web traffic stats of the top exchanges within CryptoCompare’s total pool of exchanges. It is based on similar studies that have attempted to make a connection between the number of unique web users per domain and the subsequent 24h trading volume for that specific domain. This analysis assumes that the more unique visitors an exchange attracts, the higher its trading volume.

Average Daily Visitors versus 24H Volume – Alexa Rankings Above 100,000

IDAX and CoinBene appear to have lower average daily visitors compared to similarly sized exchanges by daily volume.

The figure above represents the top exchanges by volume that have an Alexa ranking above 100,000. The reason for this is that according to Alexa, any ranking below this may not be statistically significant.

What we can see that exchanges such as IDAX and CoinBene have lower Average Daily Unique Visitor numbers than other exchanges with similar volumes such as Kraken, Bitstamp, and CoinEx.

Binance has the highest average daily visitor count, in line with its high trading volumes. Meanwhile, exchanges such as Coinbase, Cex.io and Bittrex have significantly greater numbers of daily visitors than other exchanges with similar daily volumes. In Coinbase’s case, this can be attributed to the exchange’s reputation and age.

Average Daily Visitors versus 24H Volume – All Alexa Rankings

ZB and EXX attract significantly lower daily visitors than similarly-sized exchanges.

The above figure represents the top 20 exchanges by 24h volume regardless of whether their Alexa rankings are below 100,000. Noticeably, unique visitor counts for exchanges ZB and EXX are significantly lower than other exchanges within a similar 24h volume band.

These exchanges maintain average daily trading volumes of 248 million and 160 million USD
respectively. Despite this, their daily unique visitor counts amount to no more than 700 visitors per day.

Although there is a chance that these web statistics may present errors given Alexa rankings below 100,000, in the interests of mitigating any potential risks, these exchanges will be flagged until clarification is provided.

Order Book Analysis

The following order book analysis investigates the relative stability of various cryptocurrency exchanges based on snapshots of the average order book depth for the top markets on each exchange in 10-minute intervals over a period of 10 days. In the context of this analysis, average depth down is defined as the cumulative volume required (in USD) to reduce the price of a given market by 10%. This is compared to the average daily volume for the top 5 pairs. The result of this analysis is that we are able estimate the relative stability of a given exchange based on the ratio of depth down to average daily pair volume.

Average Order Book Depth Down vs Average Daily Exchange Pair Volume

In relative terms, CoinBene, ZB and CoinEx have the thinnest markets.

Despite relatively large average volumes per top pair (~12 million USD), CoinBene’s average order book cumulative depth down (order book buy side) totals only 33 thousand USD. In other words, to move the price 10% downwards, a trader would need to sell 33 thousand USD worth of currency.

In contrast, Kraken which has similar average daily pair volumes (~13.5 million USD), has an average order book cumulative depth of 4.2 million USD. This is almost 130 times larger than that of CoinBene’s and therefore suggests a much more stable exchange.

Average Depth Down to Average 24H Pair Volume Ratio

ItBit, Kraken and Bitstamp have relatively more stable markets compared to exchanges such as CoinEx, ZB and Coinbene.

In the case of ZB for instance, its depth to volume ratio was just 0.4%. I.e. in order to move the price down 10%, a trader would only need to sell 0.4% of average daily pair volume. These ratios are similarly low in the case of CoinEx (0.7%) and CoinBene (0.3%).

Meanwhile other exchanges such as Bitstamp and ItBit, had ratios of 30% and 40% respectively. This is a factor of 100 times greater than those of CoinBene’s for instance.

Transaction-Fee Mining Exchanges

Average 24H Trans-Fee Mining Volumes

The total average 24h-volume produced by trans-fee mining associated exchanges on CryptoCompare totals more than 550 million USD. This constitutes approximately 10% of total exchange volume over the last 30 days.

Decentralized Exchanges

Average 24H DEX Volumes

The total average 24h-volume produced by the top 5 decentralized exchanges on CryptoCompare totals just less than 2.4 million USD. This constitutes just 0.4% of total exchange volume.

Security Analysis – Top 100 Exchanges by 24H Volume

This security analysis aims to evaluate a pool of the top 100 exchanges by 24h volume considering the proportion of exchanges with both a public privacy and a terms & conditions page. In addition, we analyse the proportion of exchanges that have been hacked in the past as well as the publicly stated proportion of cold wallet vs hot wallet storage for users’ funds. In theory, the higher the amount of funds stored in “cold storage” (i.e. offline), the less exposed the funds held by a centralized exchange will be to hackers.

Proportion of Exchanges with both a Public T&C and Privacy Policy Page

Out of the top 100 exchanges by 24h volume, only 86% have both a public privacy policy and terms & conditions page.

Proportion of Users’ Funds Held by Exchanges in Cold Storage

A third of top exchanges store the vast majority of users’ funds in cold wallets.

Proportion of Users’ Funds in Cold Storage by Exchange

Exchanges itBit, Coinfloor, Bitfinex and Coinbase are among those that store the highest proportion of users’ funds offline.

Proportion of Exchanges Hacked in the Past

11% of top exchanges have been hacked in the past.

KYC Requirements Among the Top 100 Exchanges

Just under half of top exchanges impose strict KYC requirements, while more than a quarter do not require KYC.

Those that impose partial requirements (25%) require KYC verification in order to conduct certain activities such as to withdraw fiat, to trade fiat pairs, or to increase maximum trading amounts.

Trade Data Assessment of New Exchanges

A visual inspection of the trades on the new exchanges is now carried out. Snapshot data cannot capture volatility, so these trade graphs allow the characteristic trading to be assessed in light of its effect on the CCCAGG. Graphs were produced of all trades vs the CCCAGG for the top 5 trading pairs for each new exchange over the last month.

BCEX

BCEX displays high volatility on both of the pairs that it trades. Buying of large amounts of the order book is visible, suggesting a very thin market. The price on this exchange will accordingly not reflect the price of the cryptocurrency well, so it will not be included.

CoinTiger

Top trading pairs on CoinTiger display agreement with the CCCAGG, but due to anomalous volumes further monitoring will be carried out before considering inclusion into the CCCAGG.

iCoinBay

Pairs on ICoinBay show agreement with the CCCAGG. This exchange is a possible inclusion to the CCCAGG.

Iqfinex

A flash crash on the largest trading pair elicits a longer period of assessment before consideration for inclusion into the CCCAGG.

Liqnet

Pairs on Liqnet show agreement with the CCCAGG. However, large amounts of API downtime can be observed. The quality of the exchange API will be monitored and the exchange will be considered for inclusion in the event of an improvement in API provision.

P2PB2B

Poor agreement with the CCCAGG gives grounds to exclude P2PB2B.

StocksExchange

StocksExchange displays some unusual trading activity and a flash crash. The exchange will not be included due to trading behaviour.

Example Assessment of BTC to USD and Future Exchange Methodology Additions

This section provides a quantitative analysis of trade data received from exchanges. The purpose is to provide an understanding of what the exchange trading ecosystem looks like, and to allow for selection of exchanges that best represent the price of a cryptocurrency.

In order to make comparisons across exchanges, an estimate of the trading price of the cryptocurrency needs to be ascertained. For the BTC-USD pair, all trades over a 30-day period were collated and plotted. In this time period, there were around 6.5 million unique trades. The trades are plotted such that colour indicates the density of points in the area.

All BTC to USD trades over 30 days

This graph represents the entire ecosystem of the price of BTC-USD trading over a 30-day period. This is now used to generate a representative price for BTC. The median was selected to calculate a trading price for the cryptocurrency. The motivating factor behind this measure being used was the large number of outliers in the trade data set. To keep the computation tractable, trades were grouped into 1-hour long time bins, and the median for each of these bins was computed.

For the purposes of this investigation, volume weighting was not used. This was due to high volume buying up of order books being observed when looking at individual exchange trade data. It was hypothesised that the arithmetic median would better reflect the mid-price of the order books of the exchanges, as the majority of trades take place at the mid-price. The median should therefore reflect the price that the average trade was carried out at.

The 1-hour median line was then plotted on the trade data, and a visual inspection of a section of the above graph shows that the line follows the highest trade density, which is indicative that it is a good estimate of the trading price of the cryptocurrency.

BTC to USD trades over 30 days with hourly median price line

CryptoCompare’s CCCAGG is an aggregation of trade prices, and aims to reflect the current trading price of an asset. It is possible to validate the CCCAGG price by comparing it to the median trade price. It can be seen that there is agreement between the two measures, suggesting that the CCCAGG is accurately capturing the trading price.

CCCAGG Price vs Median Trade Price for BTC to USD

 

 

Abstract: We are proud to announce the launch of ForkMonitor.info, a new website sponsored by BitMEX Research. The website is connected to several different nodes, both Bitcoin and Bitcoin Cash implementations. It displays various pieces of information regarding the chains followed. This website can be useful for monitoring the situation during network upgrades (softforks or hardforks), as well as being potentially useful in helping to detect unintentional consensus bugs. Thanks to Sjors Provoost for helping develop the site.

 

New Website: ForkMonitor.info

 

(Website screenshot as at 3rd Nov 2018)

Overview

The website is currently connected to the following 13 nodes:

Bitcoin Nodes	Bitcoin Cash nodes
Bitcoin Core 0.17.0.1	Bitcoin ABC 0.18.2
Bitcoin Core 0.17.0	Bitcoin ABC 0.18.0
Bitcoin Core 0.16.3	Bitcoin ABC 0.17.2
Bitcoin Core 0.16.0	Bitcoin ABC 0.16.2
Bitcoin Core 0.10.3	Bitcoin ABC 0.14.6
	Bitcoin SV 1.0.1
	BUCash 1.5.0
	BUCash 1.3.0.1

(Nodes run by forkmonitor.info as at 4 November 2018)

The website is primarily geared towards Bitcoin Cash, running 8 Bitcoin Cash nodes compared to 5 Bitcoin nodes. The reason for this is the upcoming Bitcoin Cash hardfork, where several different nodes appear designed to follow different chains.

The forkmonitor.info website may also be useful in monitoring the situation during this upgrade. After the Bitcoin Cash hardfork is complete, the website’s intention is to move some of the focus over to Bitcoin. The plan is to run more versions of Bitcoin Core (especially older versions), as well as independent implementations such as Bcoin, BTCD and Libbitcoin. This may be helpful in spotting any consensus bugs, such as the inflation bug CVE-2018-17144, which was discovered in September 2018. The website’s code will be made open source, which may hopefully encourage other organisations to spin up multiple nodes and monitor the chains in a similar way.

The Bitcoin Cash hardfork

At around 16:40 UTC on 15th November 2018, Bitcoin Cash is expected to hardfork. There is potential for three competing chains:

• a hardfork implemented by Bitcoin ABC
• a second hardfork implemented by Bitcoin SV
• potentially the original rules chain

A list of some of the main clients and their respective positions on the hardforks is provided below:

Client name	Comments
Bitcoin ABC	Versions of Bitcoin ABC after and including 0.18.0 are expected to activate a hardfork at around 16:40 UTC on 15 November 2018, according to the median past time. Versions of Bitcoin ABC prior to this, are not expected to follow this new chain. In our view, Bitcoin ABC is the most popular implementation and the economic majority of Bitcoin Cash users are likely to support the hardfork and follow the new chain. It is unclear to us what will happen to older versions of Bitcoin ABC; however, the likely outcome is that no additional blocks are produced on the original chain.
Bitcoin SV	Bitcoin SV (or Bitcoin Satoshi’s Vision) is a client promoted by Craig Steven Wright, who is popularly known as the “Fake Satoshi”. In 2016, Mr Wright produced what he claimed was proof that he was Satoshi, however it quickly emerged that this was a digital signature copied from Bitcoin’s Blockchain, presented in a manner designed to be confusing. Bitcoin SV is also expected to activate a hardfork at the same time as Bitcoin ABC, however, this hardfork is supposedly incompatible with Bitcoin ABC. In our view, Bitcoin SV is likely to have limited support from users, investors and traders. However, some of the large Bitcoin Cash mining pools, apparently support Bitcoin SV or are otherwise affiliated with Mr Wright: Coingeek: 25% share (A pool owned by Calvin Ayre, allegedly a financial backer and  supporter of Wright) BMG Pool: 12.5% (Another pool believed to be linked to Wright, with BMG being a division of Wright’s company nChain Group) SV Pool: 7% market share (A pool set up to support Bitcoin SV) (Source: cash.coin.dance) In addition to the above, the listed Canadian mining company Squire Mining (SQR CN), with a CAD$65 million market capitalisation, is likely to be supporting Bitcoin SV. According to Squire’s investor presentation, Stefan Matthews is a director while Bloomberg data shows that he owns 9.3% of the company’s shares. Furthermore, a June 2016 book on Wright entitled “The Satoshi Affair” implies that Matthews is a long standing and close friend of Wright. Matthews was the CEO of “nChain”, another company deeply involved in the shenanigans of the “Fake Satoshi”. Squire’s investor presentation states that Mr Matthews: is currently the chairman, of the nChain Group, known for global leadership in blockchain and Bitcoin research. BMG, a division of the nChain Group Therefore we believe it is likely that despite the lack of community support for Bitcoin SV’s hardfork, the chain could have considerable hashrate, even if it’s only for a limited period. Although, despite all the noise and promotion generated by nChain, Coingeek and Wright, we do not know for sure if the mining pools related to these entities are actually running Bitcoin SV. Even if Bitcoin SV does have significant or even majority hashrate, if the Bitcoin Cash economy ignores it, the chain should have little financial impact. We view this as the most likely outcome.
Bitcoin Unlimited	There is a third client group called Bitcoin Unlimited. This group’s BUCash 1.5.0 client is designed to follow the hardfork of Bitcoin ABC. Earlier versions may behave differently.

While it appears that the economic majority will support Bitcoin ABC’s hardfork, there is significant uncertainty over how each client will behave and which chains they will follow. Therefore, BitMEX Research has sponsored this new website which has launched before the hardfork is due to occur. This will hopefully provide useful information to some stakeholders, as the events get underway next week.

 

Abstract: We examine the power and dynamics of the “Bitcoin Core” software project and we draw distinctions between the various different ways one can compete with the project. We address the misconception that the Bitcoin Core software repository has the unique capability to change or prevent changes to Bitcoin’s consensus rules. We also discuss some common misconceptions and explain that if the Bitcoin Core repository becomes hijacked by nefarious actors or deleted, Bitcoin should be largely unaffected.

Venn diagram illustrating the various ways to “compete” with Bitcoin Core

(Sources: Bitcoin ABC, Bitcoin UASF, BTCGPU, Bitcoin XT, BTC1, Bitcoin Classic, Bitcoin Cash Cobra, Bitcoin SV, Bitcoin Unlimited, BitcoinX, Bitprim, Bcoin, Parity Bitcoin, BTCD, Libbitcoin, Caesure, Bits of Proof, Bitcoinj, Ufasoft Coin, Bitcrust, Picocoin, Bitcoin Addrinex, Bitcoin Knots, Bitcoin-RBF, Bitcoin BitMEX Research)

The three kinds of competition

One can categorise competing software projects with Bitcoin Core into three different groups:

Type of competition	Explanation
Competition between chains	This is when the competing software project deliberately has a different set of consensus rules to the implementations the users currently run. This includes both hardforks and softforks. Running such software can be considered risky in certain circumstances, as it can split the coin into two chains. Therefore this kind of competition is between different coins/chains, rather than  merely competing with a different implementation of Bitcoin. Indeed if one does a software fork of Bitcoin Core and changes the consensus rules, most of the code is still likely to be written by the same development team, so it is not really competing against the team, but potentially launching a new coin whose code was written by that same team.
Competition between independent implementations	This form of competition occurs when Bitcoin is re-implemented without using the code from Bitcoin Core. Typically a new coding language is used; to try to capture some advantages other languages may have. Like the above form of competition, many consider this form of competition risky, as it may increase the chance of unplanned chain splits, caused accidentally by different consensus rules. The alternative client needs to match the consensus behaviour of the software users currently run, even matching bugs or unintended behaviour in the majority client.
Other competing software projects (which neither change the consensus rules nor re-implements the codebase)	One can compete with Bitcoin Core by neither trying to change the consensus rules nor by writing a new independent codebase. One can do this by creating a software fork of the project and then making only non consensus changes. This type of competition does not share many of the risks mentioned above.

The debate over competing consensus rules

This topic has been widely discussed in the Bitcoin community, largely in the context of the “blocksize war”, which ran from the summer of 2015 to November 2017. We are not going to repeat all those arguments in this report, where the primary purpose is to articulate the different types of competition.

In favour of competition

Opposed to competition

Competition over the rules should be encouraged, since this ensures the coin is flexible and able to adapt and compete. The model of the status quo ruleset always prevailing mean that the rules may never change, even when the case is highly compelling, as in this contentious environment a minority will always oppose any change. Competition over the rules is far less likely to cause significant disruption than many people think. In reality large businesses and the community will quickly rally behind one coin and change the client they run to follow the economic majority or hashrate majority.

It is best to try to avoid competition over the consensus rules, as doing so is risky and damages the stability of the coin. In the event of a dispute, the existing consensus rules should prevail, this keeps the existing rules of the coin, such as the 21 million cap robust, a key and unique property of Bitcoin. The disruption which can be caused by changing the consensus rules without widespread agreement, is therefore a highly desirable characteristic of Bitcoin. Changing the consensus rules should therefore occur in one of the following two ways: With widespread agreement across the community of coin users and technical experts.  Sufficient time must also be given for users to upgrade their clients If developers are unsure if a sufficient number of users will upgrade to the new rules, this could result in the launch of a new coin. In this case various safety measures such as strong two way replay protection and chain wipeout protection (for both fully verifying clients and light clients) may be necessary to reduce the risk of users losing funds (If the change in the rules is a softfork (as opposed to a hardfork), it may be possible to prevent a chainsplit if the majority of miners upgrade)

The debate on competing independent implementations

As above, this is also a very controversial and divisive topic, however we still think it’s a fundamentally different issue to competition over deliberate changes to the consensus rules.

In favour of competition

Opposed to competition

Although one dominant implementation may protect the network from unexpected consensus bugs, it may leave the coin exposed to certain types of critical bugs, such as bugs which caused clients to crash or allow unexpected coin inflation to occur. A recent example of this is CVE-2018-17144, a critical inflation bug only discovered in September 2018. If, for example, there were ten independent implementations, each with a 10% market share, if a bug occurred on one of the implementations which caused it to crash or caused inflation, 90% of the network could continue as normal. The network would therefore become more resilient. Diversity of the clients users run is therefore a key strength.

The strongest opponent of this form of competition was probably Satoshi, he/she famously said: I don’t believe a second, compatible implementation of Bitcoin will ever be a good idea.  So much of the design depends on all nodes getting exactly identical results in lockstep that a second implementation would be a menace to the network.  The MIT license is compatible with all other licenses and commercial uses, so there is no need to rewrite it from a licensing standpoint. A second version would be a massive development and maintenance hassle for me.  It’s hard enough maintaining backward compatibility while upgrading the network without a second version locking things in.  If the second version screwed up, the user experience would reflect badly on both, although it would at least reinforce to users the importance of staying with the official version.  If someone was getting ready to fork a second version, I would have to air a lot of disclaimers about the risks of using a minority version. This is a design where the majority version wins if there’s any disagreement, and that can be pretty ugly for the minority version and I’d rather not go into it, and I don’t have to as long as there’s only one version. (Source: Bitcointalk) Although ten popular implementations might be good, the issue is the transition from one dominant implementation to a diversity of popular clients, without entering dangerous territory such as two popular independent implementations, each with a 50% market share, leaving the network vulnerable to consensus bugs. Therefore a better plan may be to have one dominant implementation which is highly scrutinized, to keep consensus bugs to a minimum. This way the network may be reliable for all users, even 10% of a minority chain may be a problem for that 10%.

Other competing clients

Even if one really likes a robust ruleset, opposes competition over the consensus rules and one religiously follows Satoshi’s negative view about competing implementations, this does not mean one cannot have competing software projects. The competition can simply be in the white area, outside of the circles in the above venn diagram. This form of competition, which neither initiates a deliberate change to the consensus rules nor re-implements the code, is not controversial at all, as far as we can tell.

Therefore in theory Bitcoin never needs to suffer from the apparent problems of who controls a particular software repository in Github or arguments over who has commit access to the repository. In our view, many of these apparent problems are based on a misunderstanding, by people who appreciate some of the risks of competing software projects, but fail to distinguish appropriately between the different types of competition. Therefore many seem to overestimate the power of the Bitcoin Core software repository, thinking that any competition is risky or somehow unacceptable.

Bitcoin Core’s genesis

Prior to 2013, there was no software project named Bitcoin Core. The Satoshi client was sometimes just called the reference implementation or Bitcoin-QT/Bitcoind. Then in February 2013, Gavin Andresen, a prominent Bitcoin developer, posted to the Bitcoin Foundation forum asking:

There was some discussion about renaming Bitcoin-Qt and the reference implementation in general in IRC today; I thought some of you smart people might have good name ideas.

Mike Hearn, another developer, then responded:

Oh good, about time. This has irritated me for a while. How about Bitcoin Core?
(Source: Bitcoin Foundation Forum)

Many then started referring to the software project as “Bitcoin Core”, but nothing actually changed. Bitcoin Core then began to develop a strong brand, associated with prudence and stability, or as Gavin said at the time, “[its] like a rock”.

The impact of the “blocksize war”

During the blocksize war, many characterised the debate as being Bitcoin Core vs miners or large businesses, with the Bitcoin Core side opposing hardforks and blocksize limit increases. In our view the characterisation was mostly incorrect. However, many who made this characterisation then subsequently concluded that Bitcoin Core won, since there was no hardfork. This same group therefore currently overestimate the power of Bitcoin Core, in our view.

Bitcoin Core is not as powerful as many people think

It is not the Bitcoin Core software repository that defines Bitcoin’s consensus rules. The rules are defined by the clients economically significant users currently run. These are typically previously released versions of Bitcoin Core. The Bitcoin Core software project cannot change what software users are running and the users are a lot more independent minded than many people think, in our view. Even if Bitcoin Core had released a hardfork client, which increased the blocksize limit, it is not clear if the community would have upgraded. Therefore concerns about the Bitcoin Core software repository becoming deleted, hacked or hijacked should be far less of an issue than many people think. If this happens it will not affect clients users are already running and if further upgrades or improvements are needed, one can simply switch to a different repository or many different repositories, without worrying about any coordination problem or other risks.

Actually, in the summer of 2017, in some ways, a client competing with Bitcoin Core, Bitcoin UASF, overthrew Bitcoin Core and deliberately changed the networks consensus rules. Therefore, concluding that Bitcoin Core is all powerful, is the wrong lesson to learn from the blocksize war.

BitMEX Research is launching a new client to compete with Bitcoin Core (For illustrative purposes only)

Today BitMEX Research is announcing a new client to compete with Bitcoin Core, Bitcoin BitMEX Research. Since it is a software fork of Bitcoin Core, it carries none of the risks of not being bug for bug compatible, like Satoshi was concerned about. The BitMEX Research client also doesn’t change Bitcoin’s consensus rules, so the concerns about contentious chainsplits do not apply. Therefore, if the Bitcoin Core repository gets hijacked or deleted, the codebase can still improve using the Bitcoin BitMEX Research client or any other set of clients.

Conclusion

Following the resolution of the blocksize war, there is too much emphasis on the power of the Bitcoin Core software repository. Common questions now are “Who controls the repository?”, “What if they delete the Bitcoin Core GitHub?”. In our view, these questions may illustrate one is missing the point of Bitcoin.

People tend to look for somebody who is in control of Bitcoin’s protocol rules. Prior to and during the blocksize war, many thought it was miners, large businesses or Gavin Andresen. One of the unexpected negative consequences of that war is that many seem to have switched their opinions to believing Bitcoin Core is incharge, an equally flawed view. The truth is, as hard as it is to appreciate, end users are ultimately in charge of Bitcoin.

Of course this could be unrealistic, in reality, ASIC manufacturers, large mining farms, developers, large custodians, large exchanges and even an individual software repository are highly influential. We may be idealistic in saying that users are ultimately in control. However, isn’t that what “user controlled money” means? If one doesn’t think users control Bitcoin, what exactly is Bitcoin for anyway?

 

 

Abstract: In March 2018, we wrote a piece on the SegWit capacity increase and compared it to Bitcoin Cash transaction volume. Another topic we have focused on is coins moved for the first time since the split, on both sides of the chain (our September 2017 report). In this piece we briefly provide an update on the metrics we were tracking. The data shows that SegWit is enjoying strong and consistent growth, while Bitcoin Cash volume is also slowly increasing from its lows, to around 9% of Bitcoin transaction volume. As at October 2018, very few pre-split coins are moving for the first time since the fork.

 

 

SegWit transaction volume – Percentage of Bitcoin transaction volume (Daily data)

(Source: BitMEX Research, Bitcoin blockchain)

On the Bitcoin network, SegWit adoption has grown substantially since our first article on the topic in September 2017. Adoption now approaches 50% and the growth has been reasonably consistent and gradual throughout the period.

 

Daily transaction volume

(Source: BitMEX Research, Bitcoin blockchain, Bitcoin Cash blockchain)

As the above chart indicates, Bitcoin Cash transaction volume declined from the c10% of Bitcoin level in March 2018, when we last commented on the topic, to around 6%. Then in the late summer of 2018 Bitcoin Cash volume picked up again, to around the 10% level. The Bitcoin Cash numbers are somewhat skewed by the “stress tests” which occurred in August 2018 and then September 2018. However, the median daily Bitcoin Cash percentage transaction volume compared to Bitcoin in the last six months is 9.0%, a recovery compared to earlier lows of around 5% or 6%.

 

Cumulative transaction volume since the launch of Bitcoin Cash

(Source: BitMEX Research, Bitcoin blockchain, Bitcoin Cash blockchain)

Since the launch of Bitcoin Cash, 22.1 million SegWit transactions have taken place, only 17.0% more than the cumulative number of Bitcoin Cash transactions, which stands at 18.9 million. Although, as the chart above illustrates, this appears to be skewed somewhat by the stress tests.

Prior to the start of the stress tests, in July 2017, there had been 15.5 million SegWit transaction, 95.1% more than the number of Bitcoin Cash transactions.

 

Coins moved for the first time since the fork

(Source: Forks.network, Original chart idea from BitMEX Research)

As for our investor flow analysis system, 9.1 million Bitcoin which existed prior to the spit has moved at least once since the fork, compared to 8.6 million Bitcoin Cash. As the chart above indicates, the gradient of the spend for the first time since the fork lines are flattening out on both sides of the split, potentially indicating further significant changes in the investor flow dynamics are unlikely.

 

Abstract: Bloomberg is reporting that Noble Bank, which back in February 2018 we speculated could be Tether’s primary reserve bank, may be facing financial difficulties. Tether is said to be diversifying away to other banks and this hypothesis is supported by Q2 financial data from Puerto Rico.

New Puerto Rico Financial Data for Q2 2018

Bank deposits in the International Financial Entities (IFE) category, which includes Noble Bank, were $2.9 billion, down 18.4% in the quarter. This is despite continuing growth of  Tether, which is illustrated in the below chart. In our view, this data supports the assertion that Tether is moving its reserves out of Noble and into other banks outside of Puerto Rico. BitMEX Research has also been informed by Tether insiders that the Tether funds have been diversified into other banks.

Puerto Rico’s IFE aggregate deposits versus the Tether balance in millions of USD. (Source: IFE Accounts, BitMEX Research, Coinmarketcap)

Bloomberg are also commenting on the financial data from Puerto Rico, stating:

Puerto Rico has seen a surge of cash related to cryptocurrencies. By the end of 2017, cash and equivalents held by so-called international financial entities, such as Noble, soared to $3.3 billion from $191 million a year earlier, according to data from Puerto Rico’s bank regulator. As of June 30 this year, the total had dropped to $2.6 billion. The majority of that money on the island was held by Noble, people familiar with the matter said earlier this year.

(Source: Bloomberg)

Bloomberg is also reporting that Bank of New York Mellon is no longer Noble’s custody bank.  Tether’s hunt for more reserve banks continues.

Abstract: Following on from our first piece on ICOs in September 2017, which focused on the team members and advisors, in this report we work with TokenAnalyst to track the Ethereum balances of the ICO projects over time. We look at the amount of Ethereum raised and the US$ value of the gains and losses caused by changes in the Ethereum price, for each project. We conclude that rather than suffering because of the recent fall in the value of Ethereum, at the macro level, the projects appear to have already sold almost as much Ethereum as they raised (in US$ terms). Of the Ethereum still held by the projects, even at the current c$230 price, projects are still sitting on unrealised gains, rather than losses.

 

Please click here to download the pdf version of this report

 

Ethereum raised by 222 ICOs – Macro analysis

	ETH	US$m
ETH raised by EOS	7,211,776	3,824
ETH raised by other projects	7,972,003	1,639
Total ETH raised	15,183,779	5,463
ETH sold by EOS	(7,211,776)	(3,892)
ETH transferred out/sold by other projects	(4,113,345)	(1,560)
Total ETH transferred out/sold	(11,325,121)	(5,452)
ETH Balance remaining (Sept 2018)	3,858,659	830

(Source: Ethereum Blockchain, BitMEX Research, TokenAnalyst, Token Data, Price data from Etherscan)

Overall profits & losses caused by changes in the price of Ethereum – US$ million

Realised gains
EOS project gains	68
Gross realised ETH gains by other projects	692
Gross realised ETH losses by other projects	(34)
Net realised gains	727
Unrealised gains
EOS unrealised gains	n/a
Gross unrealised ETH gains	403
Gross unrealised ETH losses	(311)
Net unrealised gains	93
Total net gains	819

(Source: Ethereum Blockchain, BitMEX Research, TokenAnalyst, Token Data, Price data from Etherscan)

Notes:

1. This analysis only considers the Ethereum balances of the ICO projects, which we have tracked on the Ethereum blockchain. Funds raised in currencies other than Ethereum are not considered nor is the balance of the new token created by the project. Our reported totals are therefore lower than some other sources. Therefore while our figures may be an underestimate, one at least has a degree of assurance that the balance is calculated independently of the project. At the same time we are missing several projects such as Tron, as we have not identified a treasury address or an address cluster.
2. The estimate of the value of Ethereum raised is calculated by taking the highest value of Ethereum inside the address cluster of each project at any point in time (with the exception of EOS). This will result in some inaccuracies.
3. The estimate of the value of US$ raised is calculated by using the average ETH price during the ICO period. This should therefore be considered as a rough and unreliable estimate.
4. The estimate for the realised gains was calculated by taking the month end Ethereum balance for the address cluster of each project every month and then looking at the reduction in the Ethereum holdings. The average Ethereum price for each month was then used to estimate the US$ value of Ethereum that was sold. This is likely to be inaccurate and it is possible the project retains ownership of the Ethereum or that the Ethereum was not sold for US$.
5. While we believe our estimates at the macro level may be reliable, at the individual project level our figures are likely to be unreliable. We apologise for any errors or inappropriate assumptions.

Commentary on the overall Ethereum holdings and sales

The Ethereum price has fallen almost 85% from the US$1,400 peak price in around December 2017. As we mentioned back then, the value of Ethereum and the associated crypto-currencies was high and there was significant downside risk. The large fall in the value of Ethereum led some to question if there could be a “downward price spiral” due to the concentrated Ethereum holdings of the ICO projects. The theory being that many ICO projects were sitting on a large treasure trove of Ethereum and that as the price of Ethereum fell, these projects were going to “panic sell’, fearful of being the last project holding their Ethereum bags. Read more “Ethereum holdings in the ICO treasury accounts”

Abstract: Following on from our August 2018 piece on Bitmain’s IPO, in this note we look at new information made available in Bitmain’s IPO prospectus, which was published in the last few days. The new filing confirms our suspicion that Bitmain has been making large losses recently, with a net loss of US$395m in Q2 2018. The magnitude of wasted production costs is also revealed, with almost US$0.5 billion spent on failed chips in the last 18 months. However, the document also confirms that Bitmain successfully raised US$442m from investors in August 2018, significantly strengthening their balance sheet. At the same time, this brings the IPO closer, which is good news for Bitmain and something its rivals should be concerned about.

The Income Statement to June 2018

The prospectus discloses financials up to June 2018, one extra quarter compared to what had previously been available. The new income statement confirms our suspicion (driven primarily by lower sales prices) that Bitmain has been making losses recently. As the below table shows, the company lost US$395m in Q2 2018. The IPO prospectus document shows the company making a net profit of US$742m in the first half of 2018, however since we know from the “leaked” pre-IPO presentations that Bitmain made a $1,137m net profit in Q1, we can tell that Q2 was a loss making period.

	2015	2016	2017		2018 Q1	2018 Q2
Sales	137.3	277.8	2,529.3		1.896.4	949.1
Gross Profit	71.5	158.1	1,447.1		1,137.3	(107.3)
Net Profit	48.6	118.9	1,249.4		1,137.7	(395.0)

(Source: Bitmain IPO prospectus, BitMEX Research)

However, the losses only relate to a period of one quarter and business conditions may change. One quarter of losses should not be a significant concern to long term investors, especially in a volatile business like crypto-currency mining. Although mining machine prices remain low and Q3 is also likely to be a loss making period, therefore moving back into the black may be challenging. Bitmain may need to raise prices to return to profitability, in our view.

In the document, Bitmain do acknowledge some potential strategic mistakes which may have contributed to the losses, and how they plan to address these issues going forwards:

Cash injections

The balance sheet position improved significantly in Q2 2018, mainly due to new cash injections from new investors. The net cash balance improved from US$104.9m to US$343m in Q2. Investors essentially rescued the company as it neared a cash crisis. As the table below shows, Bitmain raised even more money in Q3, which is likely to improve the cash position even further going forward.

Bitmain issuances of shares

Date	Amount raised
August 2017	US$50.0m
June 2018	US$292.7m
August 2018	US$442.0m
Total	US$784.7m

(Source: Bitmain IPO prospectus, BitMEX Research)

Inventory

The inventory balance fell to $887.2m in Q2, compared to the $1,243.8m in Q1 2018. This reduction is likely to be primarily driven by impairments. In H1 2018 Bitmain suffered an inventory write-down of US$391.3m. Therefore a significant proportion of the pain related to the overproduction could have already occurred.

Pre-payment to TSMC & the current mining industry outlook

Worryingly the TSMC pre-payment situation has not materially improved as a drain on working capital, with the balance as at Q2 being US$652.9m, only down slightly from US$666.0m in Q1. This could relate to Bitmain’s new 7nm mining product, which was recently announced. The fact that this was officially announced by the company is a positive, since the failed chips were not announced and therefore this product could finally be successful. This could rescue Bitmain from a difficult business enviroment. However, skeptics would point to the following:

• This new 7nm project could also be a failure, the company is only announcing it as they are under more pressure (in our view this is unlikely)
• Moving to 7nm is very challenging and it could take around 12 to 18 months until these devices are as reliable as the 14nm and 10nm products on the market
• Producing at the 7nm level is too expensive and Bitmain’s rivals, Innosilicon, Ebang, Bitfury have out-smarted Bitmain by selecting the cheaper and larger wafer size in their new products, which have also all been announced in the last few weeks.

Impairments related to failed chips

As we mentioned in our previous piece, “Bitmain has tried to release at least three new more efficient Bitcoin mining chips, one at 16nm, one at 12nm and more recently 10nm in March 2018. Each of these releases failed, costing Bitmain hundreds of millions of dollars”. The disclosure in these documents may reveal that our assessment may have been accurate and the scale of the cost of these failures can now be determined.

Provisions for impairments related to TSMC prepayments & inventory write-downs

Period	Value
2017	US$240.4m
2018 H1	US$252.7m
Total	US$493.1m

(Source: Bitmain IPO prospectus, BitMEX Research)

The above illustrates just how risky and potentially financially costly it is to produce new chips. Bitmain have paid a high price for some of the failures.

Bitcoin Cash & the crypto-currency holdings

The prospectus does not reveal any significant new information compared to our previous report, individual holdings by coin were not disclosed. The value of crypto-currency on the balance sheet fell to US$886.9m in Q2, compared to US$1,172.4m in Q1. This is likely to be driven by a fall in value in Bitcoin Cash and the other coins. Bitmain disclosed an impairment of its crypto-currency holdings of US$102.7m in Q2, which is likely to have accelerated further into Q3.

Conclusion

The prospectus enables us to quantify the financial impact of mistakes we already suspected that Bitmain had made.

• Bitmain lost US$0.5 billion on production costs associated with mining chips that failed (or other inventory write-downs)
• The company was dependent on US$784.7m on new investment to retain a positive cash position
• Bitmain incurred a net loss of almost US$400m in the most recent quarter, due to having too much inventory and needing to lower sales prices

Obviously many of these mistakes could have been avoided, but all they really show is that Bitmain take risks. If Bitmain didn’t take such risks the company would not have built $1,617m of shareholder equity in the last few years and Bitmain would not have been the largest and most profitable mining company in 2017.

We now know the IPO is close and could occur within a few months. This could provide Bitmain a substantial cash war chest. Although Bitmain’s rivals have very recently successfully began releasing a wave of new more efficient mining products, Bitmain’s new large cash reserves is something they should worry about. Even though Bitmain obtained this money from investors, rather than generating it from free cash flow.

(The timeline of the IPO or number of shares which will be sold has not been disclosed in the filling)

 

Abstract: In this piece we review and analyse Bitmain’s financial data, which was made available (or leaked) as part of the pre-IPO process. The figures indicate Bitmain was highly profitable and cash generative in 2017, but may currently be loss making. Bitmain also spent the majority of its operating cash flow acquiring Bitcoin Cash and may have suffered mark to market losses of US$328 million as a result. We conclude that the IPO itself may go well, however going forwards the allocation of investor capital will be key and management may need to improve in this area.

 

Please click here to download the pdf version of this report

 

The IPO process

Bitmain expects to submit IPO documents to the Hong Kong Stock Exchange at the end of August 2018, with a public listing expected towards the end of 2018. The company has just conducted a pre-IPO round, raising several hundred million dollars at a valuation of around $14 billion. Therefore we believe the company is likely to attempt to raise several billion dollars at the IPO stage, with a valuation north of $20 billion.

Documents outlining this process, which also contain Bitmain financial data, have been leaked on Twitter. We have reason to believe the authenticity of these documents, which forms the basis for this report.

Bitmain’s position in the mining industry

Area	Companies	Bitmain’s position
Chip manufacturing/foundry	TSMC, Samsung, Global Foundries, SMIC	Bitmain has no presence or prospects in this area
ASIC design, mining machine assembly and distribution	Bitmain, Canaan Creative, Ebang, Innosilicon, Bitfury	Bitmain is the dominant player in this area and this is the company’s core business. In 2017 Bitmain claims to have had a cryptocurrency market share of 85% and a Bitcoin share of 77%
Mining pool operator	BTC.com, AntPool, BTC.TOP, Slush, ViaBTC & F2Pool	Bitmain has a dominant position in this space. BTC.com & Antpool are the two largest pools, which are both owned by Bitmain. Bitmain is also an investor in ViaBTC. In the last six months these three pools had a combined global market share of around 48%
Mining farm operator	Hive Blockchain, Genesis mining, Bitmain	In 2016 Bitmain is likely to have been a dominant player, however the pre-IPO documents show Bitmain has significantly scaled back in this area.  Revenue from own mining operations has fallen from 18.4% of total revenue in 2016 to just 3.3% in Q1 2018.

(Source: Bitmain pre-IPO documents, BitMEX Research)

Will there be demand for the IPO?

Bitmain has a strong or dominant position in most areas in cryptocurrency mining, as the table above illustrates. Bitmain is likely to be the largest and most profitable company in the blockchain space, which is likely to make the company attractive to many investors.

In the last few years and decades the key lesson many technology investors have learnt, rightly or wrongly, is to always invest in the number one company. The number one company typically benefits the most from network effects and as a result smaller rivals tend to fail. The below list of usual suspects illustrate this basic point:

Sector	Company
Messaging app	Whatsapp, Line (Japan), WeChat (China)
Search	Google
Ride sharing	Uber
eCommerce	Amazon (US), Alibaba (China)
Music streaming	Spotify
Online Video streaming	Youtube

Justified or not, the blockchain space is now regarded by many as one of the next big internet based technologies and Bitmain is the number one player in this space. Whether this network effect type logic can apply to ASIC design and distribution is not clear to us and the benefits of being big may be limited to the more traditional economies of scale. We think it is important to not only look at cryptocurrency mining through the technology lens, but also to look at it from the angle of an energy intensive industry, like mining for natural resources. In these sectors the benefits of scale are more limited than for internet based networks. Therefore we don’t completely agree that one should blindly invest in the largest cryptocurrency mining entity, we just think than some investors, perhaps naively, may think this way.  

Bitmain’s own mining operation declines

As we alluded to above, one of the most interesting discoveries in the Bitmain pre-IPO documents was the sharp decline in Bitmain’s own mining farm business. Although the share of revenue has dramatically fallen, in absolute terms own mining revenue still grew by 250% in 2017, its just that the 948% growth in equipment sales overshadowed this.

Bitmain – share of revenue from own mining operations

(Source: Bitmain pre-IPO documents, BitMEX Research)

We believe this decline represents a smart strategic decision by Bitmain to divest (relatively speaking), from an increasingly competitive and lower margin area. In our view, as one moves down the mining supply chain, the competition is likely to increase faster and Bitmain made a sensible move by trying to focus their efforts up the chain, where an increasingly large proportion of the value from mining may accrue. In some ways this is good news for Bitcoin decentralisation, as a dominant mining player has stepped back. We believe ASIC design and distribution is less critical to network security than mining farm operation, which in theory choose the pools who construct blocks and select which blocks to build on top of. Of course Bitmain’s power and dominance in the ASIC space still remains as a significant problem for Bitcoin.

Currently Bitmain are likely to be making losses 

In the documents, Bitmain disclosed the revenue, sales and crucially gross profit margin for each of the main mining products. We have displayed the relevant data in the table below. The data shows that Bitmain sold over a million S9’s in 2017 and then over 0.7 million in Q1 2018 alone.

Financial metrics by mining product

	2016	2017	Q1 2018	BitMEX  projection (Current prices)
Revenue by product – $m
S9 (Bitcoin)	98.1	1,347.4	1,225.9
S7 (Litecoin)	106.3
L3 (Litecoin)	0.9	421.6	344.0
D3 (Dash)		411.6
T9 (Bitcoin)		34.9	84.2
A3 (Siacoin)			76.8
V9 (Bitcoin)
Other	3.3	9.3	9.1
APM Power	13.6	104.8	78.7
Total	222.2	2,329.0	1,829.7
Price – $
S9	1,429.0	1,257.0	1,719.0	499.0
S7	593.0	212.0
L3	1,315.0	1,685.0	1,404.0	209.0
D3		1,581.0	179.0	176.0
T9		1,031.0	991.0
A3			1,431.0
V9			145.0
APM Power	108.0	110.0	95.0
Gross profit margin
S9	58.8%	55.7%	69.9%	(11.6%)
S7	58.6%
L3	21.2%	71.0%	44.6%	(133.8%)
D3		76.2%	(108.8%)	(113.8%)
T9		19.9%	49.7%
A3			78.3%
V9			(34.5%)
Total	57.9%	58.9%	61.9%	Loss
Volume (units)
S9	68,672	1,071,901	713,158
S7	179,315
L3	676	250,181	245,007
D3		260,313
T9		33,885	84,932
A3			53,703
APM Power	125,513	952,785	828,194
Total	374,176	2,569,065	1,924,993

(Source: Bitmain pre-IPO documents, BitMEX Research, Bitmain website)

Using the disclosed gross profit margin from 2017, we calculated the implied cost price of each machine. Assuming these costs remain unchanged (which may be unrealistic), we were able to calculate gross profit margins based on the latest prices on the Bitmain store. This analysis implies Bitmain are currently loss-making, with a negative profit margin of 11.6% for the main S9 product and a margin of over negative 100% on the L3 product. In reality costs are likely to have declined so the situation may not be as bad, however we think it is likely Bitmain are currently making significant losses.

These low prices are likely to be a deliberate strategy by Bitmain, to squeeze out their competition by causing them to experience lower sales and therefore financial difficulties. In our view, herein lies the key to one of the main driving forces behind the decision to IPO. A successful IPO may increase the firepower available to continue this strategy and eliminate an advantage rivals could have by doing their IPOs first.  

Another reason for these low prices and apparent losses may be that Bitmain has too much inventory on the balance sheet. As at March 2018 Bitmain had $1.2 billion of inventory on the books, equal to 52% of 2017 sales. Bitmain may therefore have had to suffer inventory write downs, which could have generated further losses in addition to the loss making sales.

Use of operating cash flow and balance sheet

The documents contain summary balance sheet data. On the positive side is that Bitmain has no debt and the company was highly cash generative in 2017. The negatives include:

1. Large prepayments to TSMC, totalling almost $866m in 2017, which weaken Bitmain’s working capital situation;
2. A large inventory balance, of around $1.2bn (over 50% of peak annual sales) illustrating overproduction;
3. A large portfolio of altcoins, with a cost base of $1.2 billion which represents the primary use of Bitmain’s cash flow.

Balance sheet – US$ million	2016	2017	Q1 2018	BitMEX projection (Current value)
Cryptocurrency
Bitcoin Cash	–	673.5	887.5	558.7
Bitcoin	69.1	216.1	148.2	153.1
Litecoin	2.2	49.0	51.2	56.1
Dash	–	103.0	103.4	55.0
Ethereum	–	0.6	0.8	0.3
Other adjustment	(40.0)	(336.7)	(18.7)	–
Total coin assets	31.3	705.5	1,172.4	823.2
Fixed assets	54.0	355.7	175.7
Other	2.4	2.7	3.0
TSMC prepayment	42.7	866.0	666.0
Receivables	7.7	66.4	167.4
Inventory	61.9	1,034.1	1,243.8
Cash	18.2	60.6	104.9
Total assets	218.2	3,091.0	3,533.2
Liabilities	81.3	1,638.3	896.1
Net assets	136.9	1,452.7	2,637.1

(Source: Bitmain pre-IPO documents, BitMEX Research, Prices from Bitfinex)

(Notes: Projections based on prices as at 28 August 2018, coin holdings as at 31 March 2018)

One of the key assets of the company is its portfolio of cryptocurrencies, valued (on a cost basis) at almost $1.2bn as at March 2018. As at March 2018 this consisted of over 1 million Bitcoin Cash. The market value of the altcoin portfolio has fallen in value since Bitmain invested, with almost all the losses attributable to Bitcoin Cash, as the chart below shows.

Bitmains investment in cryptocurrency – change in value vs cost price – $ million

(Source: Bitmain pre-IPO documents, BitMEX Research, Prices from Bitfinex)

(Notes: Prices as at 28 August 2018, coin holdings as at 31 March 2018. Chart assumes coin holding do not change)

As the following chart below illustrates, the Bitcoin Cash investment itself is very significant, to the extent that the company spent around 69% of its 2017 operating cash flow on purchasing Bitcoin Cash. Although this could be an exaggeration, some of the Bitcoin Cash would have been inherited from pre-fork Bitcoin. The figures imply that around 71,560 of the 1,021,316 Bitcoin Cash coins could have been inherited in this way. 

Bitmain use of 2017 cashflow – $m

(Source: Bitmain pre-IPO documents, BitMEX Research)

The situation is even worse than the above indicates. Not only did Bitmain spend a majority of the 2017 cash flow into Bitcoin Cash, they also spent a majority of cash flow from their entire history of operations, into Bitcoin Cash. The documents show that Bitmain generated no cash flow in in 2016 and then only $25m in Q1 2018 (perhaps due to large TSMC prepayments).

Bitmain operating cashflow – $m

(Source: Bitmain pre-IPO documents, BitMEX Research)

In a sense of course none of this matters. Bitmain spent their own funds on risky assets and they knew the risks. For a public company the situation could be a different, with investors expecting the company to invest in core operations or return money to investors. Although perhaps our expectations for governance here are too high for Hong Kong.

Why are Bitmain doing the IPO?

In our view the primary motivating factor for the IPO is simply that Bitmain’s competitors are also planning on doing them and the industry is fiercely competitive, as Bitmain’s loss making prices indicate. Rival Canaan Creative are planning on an IPO and Bitmain are unlikely to let them obtain such a funding advantage. Bitmain’s IPO should deduct money from the pool of capital that could otherwise be invested in Canaan as well as the other miners and it is therefore a good complement to the strategy of lowering prices.

The other reason for the IPO may be to strengthen the balance sheet after investing the majority of the operating cash flow into Bitcoin Cash. Bitmain only had around $105 million of cash on the balance sheet as at March 2018, when this figure could have been nearer a billion dollars if the company hadn’t acquired so much Bitcoin Cash. At the same time the business does require a lot of cash, for example the large advance cash payments TSMC require, which reached a peak of $866m in 2017.

The debate over Bitmain’s technological advantage

Nobody can challenge the performance and scale of Bitmain’s operations

Bitmain is losing its lead and has not increased the performance of its miners in over two years

It is true that Bitmain competitors have recently released more efficient mining machines than Bitmain, however this is only part of the picture. The figures show that in the past 27 months, Bitmain has delivered 1.9 million S9’s and 3.0 million mining machines in total. No competitor has the capability to deliver on anything like that scale. Rivals can at best deliver a few hundred thousand machines per annum At the same time, although machines from rivals are more efficient, Bitmain’s S9 product is more reliable and has less variance with respect to the hashrate For example, although the Dragonmint T1 product is more efficient, according to official figures. the hashrate variance is higher than expected, this is not acceptable for low margin mining farm operators, who need a predictable product for budgeting purposes. The S9 is the only product that has the necessary reliability Bitmain is the largest player and is lowering prices, such that the other ASIC design companies are now under severe financial stress. Bitmain has already attracted investments from some of the top VC funds in Silicon Valley and the upcoming IPO will ensure Bitmain’s dominance for years to come

The Bitmain S9, a 16nm product, was released in December 2015, with an efficiency of around 110 W/TH. The company has not successfully innovated or improved its Bitcoin miner performance since then. This is over 2 and a half years ago Since early 2017 Bitmain has tried to release at least three new more efficient Bitcoin mining chips, one at 16nm, one at 12nm and more recently 10nm in March 2018. Each of these releases failed, costing Bitmain hundreds of millions of dollars. Even TSMC themselves have mentioned that they think the Bitmain investment strategy is too optimistic, which may be part of the reason they insist on such large prepayments These failed tapeouts have finally resulted in competitors producing better machines, today the Innosilicon T2 and ShenMA M10 are more efficient than the S9, with a 80W/TH and 65W/TH performance respectively Bitmain has lost its technological edge as key staff, such as former director of design Dr Yang Zuoxing, have left. Dr Yang is said to have founded a rival mining firm which was sued by Bitmain for a patent violation  Without the ability to innovate and produce better equipment, the only way Bitmain can generate sales is by lowering prices, until eventually the company loses its dominant market position This desperate situation is why some claim that Bitmain tried to mislead investors into thinking it had received an investment from the government investment funds in Singapore. Although we have seen no compelling evidence of Bitmain misleading in the way suggested

The narrative surrounding Bitmain’s technical capabilities can be spun in either direction and as ever the truth may lie somewhere in the middle. However, one thing is clear, if these mining companies do go public, the picture should be far less murky going forwards and we think that could be a significant positive for the cryptocurrency community.

Conclusion

In a way some of Bitmain’s biggest mistakes, such as producing too many units and buying too many speculative altcoins in a bull market, are somewhat analogous to the typical behaviour of mining company management teams.  For instance gold mining firms often invest in high cost assets in bull markets and then fail to invest in quality low cost assets in bear markets. Perhaps it is not totally fair to blame these companies, the hedge funds and institutional investors who own the shares are often just as, if not more, at fault. Greed, fear and the emotions of market moves can affect us all. Therefore although Bitmain has made mistakes, in many ways they are not unusual or unexpected.

We are sure you have heard it before, but “cryptocurrency is here to stay”. In that environment we think Bitmain has the ingredients to be one of the great companies in the space. Bitmain can be a legendary crypto company, generating strong shareholder returns for decades to come, but in order to achieve this (and it’s a lot harder than it sounds) the Bitmain management team may need to improve their management of company resources. Once the company goes public, capital allocation decisions in this volatile and unpredictable market will be difficult enough, letting emotions impact too many investment decisions may not be tolerated.

 

Abstract: We examine the extent to which one miner dominated Bitcoin in 2009. We review Sergio Demian Lerner’s 2013 analysis, where he discovered that the increase in the ExtraNonce value in the block can potentially be used to link different blocks to the same miner. We build on his analysis and conclude that although the evidence is far less robust than many assume, there is reasonable evidence that a single dominant miner in 2009 could have generated around 700,000 bitcoin. Although our analysis itself is weak and there is no perfect way of approaching this problem.

Bitcoin blocks mined in 2009 – ExtraNonce vs block height

(Source: BitMEX Research, Bitcoin blockchain)

The history of the debate over the dominant miner in 2009

Towards the start of April 2013, blockchain researcher Sergio Demian Lerner attempted to argue that Satoshi had mined a million bitcoin in 2009. The logic behind this assertion was that the hashrate throughout 2009 was at a low level, around 7 million hashes per second and this was said to be consistent with one miner being highly dominant. At the same time this hashrate was around the same to that which was present in Bitcoin’s first 14 days, which Sergio assumes may be a period in which Satoshi was the only miner. Many in the community were sceptical of Sergio’s claims. The scepticism seems to be based on the following:

1. The 7 MH/second estimate is based on block timestamps and unreliable, nor is the sample size large enough
2. There is no reason to believe Satoshi mined alone in the first 14 days
3. Many individuals recalled mining themselves in the period
4. Sergio’s assumptions about hardware were incorrect

It is likely that some people could have been biased in dismissing Sergio’s claims, since if Satoshi had mined a significant proportion of the early coins, in the eyes of some that could damage the monetary integrity of the system. However, in our view, the evidence Sergio initially presented was somewhat weak.

The ExtraNonce bombshell

Just a few days later, Sergio then posted a far more persuasive argument on his blog, with much stronger evidence that a single miner was dominant. This then eventually convinced many in the community and to this day many people believe Satoshi is likely to have mined around one million bitcoin.

Sergio’s evidence centred around something called the ExtraNonce. The ExtraNonce is not part of the Bitcoin protocol, in that it is not part of the consensus rules nor is there a formal specification about how to interpret the field. The ExtraNonce is an area in the Coinbase transaction which can vary after several hashing attempts to provide extra entropy for miners, once the standard nonce in the block header has been used up. As the image below illustrates, as the ExtraNonce changes, the impact works its way up the merkle tree into the block header (although in 2009 most blocks contained only the Coinbase transaction, as the network was not used for transactions.)

(Illustrative diagram of the structure of a Bitcoin block)

Sergio published the below picture, with the ExtraNonce on the y-axis and the blockheight on the x-axis (He incorrectly labeled the x-axis as time). The image shows that the ExtraNonce increase over time, in a series of slopes. Some of the slopes (in black) are said to be of a similar gradient, not to overlap and to go back down to zero once they reach a certain height. This is said to demonstrate that all the black lines belonged to one miner (potentially Satoshi) and that this miner now controlled almost a million Bitcoin. Although the technical points about the gradient of the slopes, the height and the lack of overlap may be quite difficult to appreciate and evaluate, the image itself is clearly very powerful and persuasive, in our view.

(Source: Bitslog)

The new BitMEX Research allocation of blocks to the dominant miner

We decided to repeat Sergio’s analysis, except our objective was to count the blocks mined by the apparent single entity and allocate all the blocks. The excercise was challenging, since the slopes interact with many other points. It is therefore impossible to do an accurate allocation. As a result, our analysis is far from perfect and we used a variety of methods, including statistical analysis, random number generators and even manual review to allocate some blocks. We will have made many errors and we do not claim our methodology is robust or scientific. However, as far as we are aware, this is the first attempt to allocate every block in 2009 as belonging to the apparent single entity or not. The below high resolution image below represents our allocation for every block in 2009.

Bitcoin blocks mined in 2009 – Allocation to the dominant miner – ExtraNonce value (y-axis) vs block height (axis)

(Source: BitMEX Research, Bitcoin blockchain)

Analysis of the allocation

Up to August 2009 we agree with Sergio’s conclusion.  There are 22 slopes prior to August, which all have both similar height and gradient (Around an increase in the value of the ExtraNonce by 4 per block found). At the same time these slopes almost never overlap. (The tiny amount of apparent overlap in some instances in less than 5 blocks and therefore may just be coincidence).

After August 2009 the pattern breaks down to some extent. The gradient of the slopes varies considerably (from 1.1 nonces per block to 10 nonces per block). At the same time the height of the slopes is inconsistent and there are many large gaps between them. Therefore although the image still looks compelling, the evidence that the miner is one entity is somewhat weak, in our view. We have presented results below, which include figures both pre and post August 2009.

Category	Number of blocks	Number of bitcoin
Allocated to the dominant miner	14,815	740,750
Weak allocation (After August 2009)	4,553	227,650
Total	19,368	968,400
Not allocated	16,920	846,000
Grand total	36,288	1,814,400

(Source: BitMEX Research, Bitcoin blockchain)

Debate over the strength of the analysis

Supporting the analysis

Weaknesses of the analysis

The image is highly powerful, just looking at it illustrates there is a dominant miner. Although explaining it statistically may be challenging, the conclusion is clear. Many of the slopes have the following characteristics: A similar gradient to each other Similar heights Slopes often start shortly after another slope has finished The slopes rarely overlap Many of the slopes are hundreds of blocks long This is all too much to be a coincidence. Therefore the evidence that many of the blocks were mined by a single entity is highly compelling

Although the ExtraNonce analysis is interesting and revealing, when it comes to estimating the number of coins mined by the dominant miner, it is essentially useless. The methodology used when conducting the allocation involves maximising the number of blocks in each slope, this is required due to the lack of any other available mechanism for allocation. Therefore the number of blocks allocated to the dominant miner is grossly overestimated. The analysis is built on a logical fallacy. In any period there is going to be at least one miner who has the largest share or the steepest rate of increase in the ExtraNonce. There are also going to be at least some types slopes which do not overlap. Grouping these slopes from potentially different miners together is misleading and potentially based on flawed reasoning. Even if the slopes are similar, this could be because different entities had a similar setup. Each miner is not independent, in the sense that they are likely to be running the same software or could be using the same popular hardware, which could produce the same pattern.

Conclusion

In conclusion, although there is strong evidence of a dominant miner in 2009, we think the evidence is far less robust than many have assumed. Although a picture is worth a thousand words, sometimes pictures can be a little misleading. Even if one is convinced, the evidence only supports the claim that the dominant miner may have generated significantly less than a million bitcoin in our view. Perhaps 600,000 to 700,000 bitcoin is a better estimate.

None of the above says much about whether the dominant miner was Satoshi, although we know Satoshi mined block 9, which we have allocated to the dominant miner in our analysis. However this is in a slope of just 11 blocks, so it’s certainly not conclusive. Whoever the dominant miner was, it is of course possible the keys have been lost or discarded by now.

We will end with one famous quote from Satoshi to consider:

Why delete a wallet instead of moving it aside and keeping the old copy just in case?  You should never delete a wallet.

(Source: Bitcointalk)

Although, maybe we are giving that quote out of context…

 

Abstract: We look at the events surrounding the bitcoin price rally in June 2011 to $32 and the following temporary flash crash down to $0.01, on the MtGox exchange. We look at the incompetence of MtGox and examine the causes of the crash. We then look at the political battle and uncertainty which occurred in the aftermath of the crash.

Bitcoin price from May 2011 to 18 June 2011

(Source: YouTube, MtGox, BitMEX Research)

 

Click here to download the pdf version of this report

 

Overview

If one likes price volatility and scandals, the summer of 2011 was an exciting time for Bitcoin. Over the course of a few days, bitcoin plummeted in value from a peak of $32 to just $0.01 on the MtGox exchange, a trading platform based in Tokyo, which was dominant at the time. This was after a recent rally, with bitcoin trading at around $2 a couple of months earlier. The crash down to $0.01 is now a famous part of Bitcoin history.

In this piece, we look at the cause of the crash and its aftermath. Although the major exchange of the time, MtGox, was shown to the community to be largely negligent, which may not have been the best advertisement for Bitcoin. In our view the engaging nature of the events which occurred that summer, ironically made a significant contribution to the level of interest in the space.

MtGox security issues & the context of the event

There was significant uncertainty surrounding the hack at MtGox which caused the June 2011 price crash and the issues surrounding it were never fully explained. The Bitcoin community was riddled with rumours about whether MtGox was solvent and how much bitcoin was stolen.

Thanks to a report published in 2017 by Kim Nilsson, we now have a relatively strong understanding of what occurred in 2011 and the damage that this caused to MtGox.  Ironically, despite the huge impact on the market and the company’s reputation, in terms of MtGox’s solvency, this event was largely insignificant compared to other security incidents. For many however, a new window into MtGox was opened up, which illustrated a severe lack of monitoring systems, governance, controls and security measures.

The  table below lists some of the major security incidents at the MtGox exchange, with the June 2011 hack highlighted in green. This incident may have only directly cost the exchange 2,000 bitcoin, an inconsequential amount, compared to roughly 837,000 bitcoin of total losses.

List of known MtGox losses

Date	Incident name	Description	USD lost	BTC lost
20 Jan 2011	Liberty Reserve withdrawal exploit		50,000
30 Jan 2011	Liberty Reserve withdrawal exploit 2	A user supposedly withdrew US$2billion from their account which never existed. Although it seems no wire transfer actually occurred and therefore there may have been no losses
1 March 2011	Wallet theft 1	Hackers obtained the MtGox wallet.dat file from the server. This is believed to be the withdrawal transaction. As of 19 July 2018, the stolen 80,000 bitcoin has never been moved.		80,000
22 May 2011	Wallet theft 2	Somebody is believed to have accessed a wallet containing 300,000 bitcoin, which was kept unencrypted on a public drive. The thieves decided to return 297,000 bitcoin, keeping a 1% fee. The return transactions are believed to be for 280,000 and 17,000.		3,000
19 June 2011	Price crash to $0.01	Hacker gained access to Jed McCaleb’s administrative account, and sold bitcoins to crash the price, to withdraw as many bitcoins as possible within the US$1,000 per day limit. Other users who purchased bitcoin at low prices may have also withdrawn funds.		2,000
11 Aug 2011	Bitomat	Took over the debts of Bitomat after the company deleted its private keys		17,000
Sept 2011	Database hacked	A hacker gained write access to the database and inflated balances to withdraw funds		77,500
Sept 2011	Wallet theft 3	A hacker obtained the main wallet.dat file again and began withdrawing funds in October 2011. MtGox appears not to have noticed this.		603,000
October 2011	Incorrect deposits	The change from the above hacker’s withdrawal transactions were incorrectly booked as new MtGox deposits, totalling 44,300 bitcoin. This resulted in customers seeing new deposit balances in their accounts. In some ways the hackers therefore caused more damage to MtGox than the value of coins which were stolen. Some of these errors were corrected, so the net impact may be around 30,000 bitcoin.		30,000
28 Oct 2011	Destroyed bitcoin	A software bug caused funds to be sent in such a way that they could not be redeemed. Example of such transactions can be found here & here		2,609
May & Aug 2013	US law enforcement seizures	Federal agencies in the US seized funds from MtGox’s Dwolla account due to allegations the exchange was not complaint with US regulations.	5,000,000
May 2013	Coinlab dispute	Coinlab sued MtGox in a dispute over a licensing agreement.	5,000,000
2011 to 2013	Willy Bot	MtGox trading program designed to make up some of the above losses, but actually ended up making things worse.	51,600,000	22,800
Total			61,650,000	837,909

(Source: Cracking MtGox, BitMEX Research)

Overview of the events in June 2011

In the weeks leading up to 19 June, many users of MtGox were reporting that their accounts had been hacked. At around the same time a database of MtGox users, including an MD5 hash of their passwords (with an unclear/inconsistent salt policy) was leaked and made available. Many passwords were identified. Some traders used the same credentials at the rival exchange, Tradehill, who also experienced security issues. Despite this, MtGox did not suspend trading, a decision which many traders questioned.

On 19th June 2011 (3am on 20th June Tokyo time), there were large sell orders on the exchange and the price crashed from around $17.50 to $0.01 and trading continued at this level for several minutes before recovering. This lead to a high degree of uncertainty, with some assuming there may be a problem with the Bitcoin network.

It now seems likely that what actually happened was that a hacker may have obtained access to the account of Jed McCaleb, the founder of MtGox who sold the exchange to Mark Karpeles around three months earlier. This account appears to have retained administrative rights to the database and therefore the hacker was able to manipulate account balances and grant themselves a large number of bitcoins on the MtGox system. The hacker is the likely to have begun selling some of these coins.

Due to the poor management of MtGox, in our view it is unlikely that the company were aware of this, even in the aftermath of the hack, and therefore the explanations provided at the time of the events were incomplete or inaccurate.

The withdrawal limits

At the time, MtGox had a daily withdrawal limit of US$1,000, this applied to both bitcoin and USD (via Dwolla). This meant that the hacker (or any others who benefited from the hack by buying bitcoin at low prices), would be unable to benefit by withdrawing the funds, except within the US$1,000 limit. However, the US$1,000 bitcoin limit was based on the market price of bitcoin on the platform and since the price fell to $0.01, in theory the maximum each user could withdraw was 100,000 bitcoin, certainly not a small amount.

Fortunately, however, MtGox appeared to also have a bitcoin based withdrawal limit, that many users were unaware of. As the Mark Karpeles said at the time:

2011-06-20 00:16:43 MagicalTux the btc withdrawal limit saved us

(Source: IRC, Note: MagicalTux is the CEO & owner of MtGox, Mark Karpeles)

Mark then mentioned that only 2,000 bitcoin were withdrawn in the aftermath of the event, which was a relatively positive result for MtGox.

Got about 2,000 BTC out

(Source: IRC)

There was widespread scepticism about this number at the time, with many believing much more was stolen. Ironically, this 2,000 bitcoin figure now seems about right, although MtGox had lost far more in other incidents. However, due to the price crash and suspension of trading, this incident was very public at the time and resulted in the incompetence of the MtGox platform being exposed to the community.

The rollback debate

Many trades took place at the artificially low price of around $0.01 during the crash. Some traders & investors were unhappy at missing out on the price rally from around $1 to $32, and therefore had buy orders waiting in the system, all the way down the order book to $0.01. To them, this crash is exactly what they were waiting for. To the dismay of many of these traders, in the aftermath of the incident MtGox said they would reverse the trades which occurred during the crash:

The bitcoin will be back to around 17.5$/BTC after we rollback all trades that have happened after the huge Bitcoin sale that happened on June 20th near 3:00am (JST). One account with a lot of coins was compromised and whoever stole it (using a HK based IP to login) first sold all the coins in there, to buy those again just after, and then tried to withdraw the coins. The $1000/day withdraw limit was active for this account and the hacker could only get out with $1000 worth of coins. Apart from this no account was compromised, and nothing was lost. Due to the large impact this had on the Bitcoin market, we will rollback every trade which happened since the big sale, and ensure this account is secure before opening access again.

(Source: MtGox)

After this announcement there was significant debate in the community as to whether the rollback should occur. Obviously many participants in the debate had a financial interest in the outcome and this was no doubt effecting their views. In many ways, there were some parallels between this rollback and the 2016 DAO “rollback” on the Ethereum network, with some similar arguments being made.

Supporting the rollback

Opposing the rollback

Most traditional exchanges tend to roll back trades in exceptional circumstances, particularly if trades occur at extremely unusual prices. The prices in this instance were certainly extreme. The bitcoin were stolen and therefore users should not benefit from stolen goods. The bitcoin may never have existed and may only have been entries in MtGox’s database and therefore it may not be possible to deliver the coins.

MtGox should take responsibility and compensate all parties involved. In particular MtGox did not act appropriately in the weeks prior to this event when many users reported that their accounts were hacked and they allowed trading to continue. MtGox had no policy with respect to the matter and should therefore honour the trades. If MtGox reversed the trades in this case, then users may not trust them again. Reversal is an arbitrary process, would MtGox reverse trades if a much smaller amount of money was stolen? This is one rule for the rich and another for the poor. Although there are some examples of major traditional exchanges reversing trades in exceptional circumstances, there are examples where they have not done so. Honouring the trades is more consistent with the no bailout, dog eat dog, 24×7 uptime, immutability type culture in the community, which was in some ways more prevalent at the time than it is today.

The community appeared to be split on this issue, with some even favouring a vote to decide.

The trader who bought 260,000 bitcoins for US$2,622

The day after the incident, a trader called “Kevin”, claims to have purchased around 260,000 bitcoins during the crash and was arguing that he should be able to keep the coins. As he explained:

I had around $3,000 USD in my MtGox account, from earlier sales I’d made. I looked at the market stats, and realized that there were tons of orders to buy BTC at $0.01 that would likely eat up any remaining bitcoins this seller had on the order. I figured if I put a buy order in for $0.0101, my order would execute first and I could buy a huge amount of bitcoins

(Source: Bitcointalk)

Kevin posted what he claimed to be the trade confirmation:

06/19/11 17:51  Bought BTC 259684.77 for 0.0101

Kevin then went on to explain the likely reason behind the price crash, which was that the seller was trying to manipulate the price down so that they could withdraw more coins within the US$1,000 limit. In our view this part of Kevin’s story is likely to be an accurate explanation for the price crash. This logic contradicts the claim from MtGox that the person who conducted the hack was also the buyer of the bitcoin.

I could place a reasonably sized sell order for $0.001, crash the market again, and withdraw probably all of the bitcoins, since they’d be valued at $0.001 each and would fit under the $1,000 USD limit. I also decided against this, when I realized that whoever placed the gigantic sell order was probably doing so for the exact same reason

However, some have doubted the accuracy of Kevin’s story, claiming the volume of trades he claims is not consistent with the MtGox feed. The feed appeared to show trading volume of only 55,000 bitcoins during the crash past $0.0101 and only 238,000 bitcoins traded in the period. Only 3,000 bitcoin seem to have been traded at the $0.0101 price. These figures are lower than those implied by Kevin, although Kevin’s trades could have been excluded from this data for a variety of reasons. The feed was also notoriously unreliable and it was not clear if there was a precise definition of some for the information in the feed. In our view, there is no reason to believe the whole truth of any of the parties involved in this incident, but Kevin’s explanation for the crash itself seems plausible to us.

MtGox price feed during the crash

(Source: BitMEX Research, MtGox. Note: Volume in bitcoin)

The proof of reserves

The MtGox exchange was down for several weeks and many users were becoming anxious about the solvency of the platform. There was uncertainty over the amount of bitcoin which were lost and users were concerned about a run on MtGox, eventually leading to the exchange going into liquidation and users losing funds. In an attempt to reduce some of these concerns, as the chat log and bitcoin transaction show below, MtGox attempted to prove it had access to a significant quantity of bitcoin, by conducting an onchain transaction on 18th July 2011.

IRC Chat log – 18 July 2011

(Source: IRC Logs)

(Source: blockchain.info)

At the time, the above action seemed to settle the nerves of many of the traders.

Conclusion

A few weeks after these events, after many false starts, trading at MtGox eventually resumed and the bulk of the trades were reversed. However, to this day, as far as we are aware, MtGox has not been able to provide a coherent explanation for what occurred. The lack of a consistent narrative from MtGox lead many to believe that MtGox had poor monitoring and controls of its systems and that the company was run negligently. Many concluded “never to trust MtGox again”.

Unfortunately, however, MtGox somehow continued to dominate the exchange ecosystem for another three years. However one views the conduct and transparency of some of the platforms and players in the ecosystem today, we can at least conclude that things have  significantly improved since 2011.

 

Abstract: Following our earlier research pieces on Tether, financial information from Q1 2018 has been released by the financial regulators in Puerto Rico, providing more evidence of the impact of Tether. In addition to this, a source close to Tether has confirmed to us that the speculation in our initial report is correct.

After our earlier speculation that Noble Bank in Puetro Rico was Tether’s primary reserve bank, a few months later in May 2018 Bloomberg released an article further substantiating our claims. As Bloomberg put it:

According to three people with knowledge of the matter, Noble Bank International, based in San Juan, Puerto Rico, took over banking duties for Bitfinex last year.

In addition to the above, BitMEX Research has also now spoken to people close to Tether, who have also confirmed the reliability of most of the claims in our February 2018 report. Our initial discovery was based on the disclosure of data from the financial regulator in Puerto Rico, who have recently provided the latest update, for the quarter ended March 2018. In our view, the data continues to support our initial speculation.

New Financial Data for Q1 2018

Bank deposits in the International Financial Entities (IFE) category, which includes Noble Bank, were $3.5 billion, up 6.9% in the quarter. Total assets in the category were $4.1 billion, up 7% in the quarter. This moderate growth coincides with a the moderate increase in the volume of crypto-coin trading, which has likely resulted from the continued growth of the Tether balance and crypto-coin ecosystem, moderated by crashing crypto-coin prices in the quarter. In the quarter, the value of Tether in issue increased by 62.7% to $2.3 billion.

We have updated the chart below from the version in our earlier piece, which compares the Tether balance with the deposits in the banking category in Puerto Rico which contains Noble Bank.

Puerto Rico’s IFE aggregate deposits versus the Tether balance in millions of USD. (Source: IFE Accounts, BitMEX Research, Coinmarketcap)

 

Cash as a percentage of total assets (an indication of full-reserve banking) also increased in the quarter, from 85.8% to 91.0%. This also indicates crypto-coin or Tether-related activity, as we explained in the previous piece.

Puerto Rico’s IFE aggregate cash balance as a percentage of total assets. (Source: IFE Accounts, BitMEX Research)

 

In the quarter the regulator appears to have changed the name of the Tether balance, to “Deposits, money market investments and other interest-bearing balances” from “Cash in banks“. We do not view this as suspicious.

 

Abstract: In this piece we look over the history of distributed stablecoins, focusing on two case studies, BitShares (BitUSD) and MakerDAO (Dai). We examine the efficacy of various design choices, such as the inclusion of price oracles and pooled collateral. We conclude that while a successful stablecoin is likely to represent the holy grail of financial technology, none of the systems we have examined so far appear robust enough to scale in a meaningful way. The coins we have looked at seem to rely on “why would it trade at any other price?” type logic, to enforce price stability to some extent, although dependence on this reasoning is decreasing as technology improves.

 

Please click here to download a pdf version of this report

 

Overview

Distributed stablecoins aim to achieve both the characteristics of crypto-coins like Bitcoin (censorship resistant digital transactions) and the price stability of traditional financial assets, such as the US Dollar or gold. These systems are distinct from tokens such as Tether, where one entity controls a pool of US Dollar collateral, ultimately making the system centralised and thus susceptible to being shut down by the authorities.

Along with the somewhat related idea of distributed exchanges, distributed stablecoins have been referred to as the “holy grail” of financial technology, due to their very strong potential benefits. In our view the transformative nature of such a technology on society would be immense, perhaps far more significant than Bitcoin or Ethereum tokens with their floating exchange rates. Distributed stablecoins could have the advantages of Bitcoin (censorship resistance combined with the ability to transact electronically), without the difficulties of a volatile exchange rate and the challenge of encouraging users and merchants to adopt a new unknown token. Such a system is likely to be very successful and therefore it is no surprise that so many people have attempted to launch such projects:

List of stablecoin projects

Name	Type	Launch Date	White paper link
BitShares (BitUSD)	Crypto-collateralized	21 July 2014	White paper
Nu (NuBits)	Crypto-collateralized	24 Sept 2014	White paper
Steem (SteemUSD)	Crypto-collateralized	19 April 2016	White paper
Corion	Non-collateralized	14 Oct 2017	White paper
MakerDAO (Dai)	Crypto-collateralized	27 Dec 2017	White paper
Alchemint	Crypto-collateralized	Sept 2018	White paper
BitBay	Non-collateralized	Sept 2018	White paper
Carbon	Non-collateralized	n/a	White paper
Basis	Non-collateralized	n/a	White paper
Havven	Crypto-collateralized	n/a	White paper
Seignoriage Shares	Non-collateralized	n/a	White paper

The technical challenges involved in creating such systems are often underestimated. Indeed constructing a distributed stablecoin system, which is robust enough to withstand cycles or the turbulence and volatility linked to financial markets may be almost impossible. For instance perhaps most forms of fiat money, even the US Dollar itself, have not even achieved that, with credit cycles putting US Dollar bank deposits at risk. A stablecoin system which builds on top of the US Dollar is therefore never going to be more reliable than traditional banking, in our view.

In economics there is a concept of money supply, with risk and the potential inflationary impact increasing as the number of layers increase. One could add this stablecoin systems on top, as a new high risk layer:

• M0 – Notes & coins plus deposits at the central banks
• M1 – Money on deposit in a bank current account (including M0)
• M2 – Money on deposit in a bank savings account (including M1)
• M3 – Money in a money market account (including M2)
• MZM – Money in all financial assets redeemable on demand (including M3)
• MSC (Synthetic Crypto Money) – Money inside synthetic crypto stablecoin systems  (including MZM)

However advanced or sophisticated the distributed stablecoin technology is, we believe the token is likely to be less robust than the layers above it in the money supply tree.

In this piece we review some of the most prominent and interesting attempts at building these synthetic US Dollar type systems. BitUSD in 2014 and then a more recent project, MakerDAO (Dai).

 

Case study 1: BitShares (BitUSD) – 2014

Factbox
Coin Name	BitUSD
Launch Date	21 July 2014
Crypto collateral	Yes
Price oracle	No

The first stable coin we will discuss is BitUSD, a stablecoin on the BitShares platform. BitShares was a delegated proof of stake (DPOS) platform launched in 2014 by:

• Daniel Larimer (The primary architect behind EOS and Steem),
• Charles Hoskinson (the former Ethereum Foundation CEO & Cardano architect), and
• Stan Larimer (Daniel’s father).

BitShares is just one in a long line of decentralised autonomous corporation (DAC) type platforms released by Daniel Larimer, as the below image shows:

(Note: Daniel Larmier’s company Invictus Innovations launched a number of token/DAC platforms including Protoshares, Angelshares and BitShares. The black arrows represent Protoshares coin holders being granted tokens in the new chains, which Invictus Innovations promised to deliver on all new DAC platforms. Source: BitSharestalk)

 

BitUSD Marketing material

(Source: Introduction to BitShares Youtube video)

BitUSD System dynamics

Pools of Funds	Description
Bitshares	The native currency of the BitShares platform
Bitshares held as collateral	Separate pools of Bitshares  held as collateral, used as backing for the stablecoin.
BitUSD	The stable token, designed to track the value of the US Dollar

 

Groups of Participants	Description
BitUSD holders	Investors and users of the BitUSD stable coin. Holders of BitUSD are able to redeem the tokens for the Bitshares held in collateral.
BitUSD creators	Those that create new BitUSD, by selling it into the market (creating new loans), by posting BitShares as collateral. This loan may be for a small period of time, after which it needs to be rolled over or have its collateral topped up to the initial margin level.
Traders	Those exchanging BitUSD for Bitshares, and vica versa, on the platform’s own distributed exchange. There is therefore a Bitshares vs BitUSD market price.
Block producers	Bitshares block producers/miners have a role of spending the BitShares backing BitUSD, something they are only entitled to do if the value of the BitShares is less than 150% of the value of the BitUSD it is backing (based on the BitUSD vs BitShares exchange rate on the system’s own distributed exchange). The miner can then uses the Bitshares to redeem/destroy the BitUSD. (After the launch the 150% margin level was increased to 200%)

 

Price Stability Mechanisms	Price Direction	Description
Investor psychology (Unclear/”Why not trade at $1?”)	Both directions	There does not appear to be a specific price stability mechanism in the BitUSD system. One can redeem and create BitUSD, however the price this transfer occurs at is determined by the BitUSD vs BitShares price in distributed exchange, which is not linked to “real USD”. In a way the price references itself. There is therefore no direct mechanism keeping the price of BitUSD at $1, but the argument put forward is “why would it trade at any other price?” In our view this logic is weak.
BitUSD redemption (indirect)	Positive	Should the value of the collateral currency (BitShares) fall, any BitUSD holder can redeem the BitUSD and obtain $1 worth of BitShares, assuming the market price of BitUSD is still worth $1 and there is sufficient BitShares held in collateral. This stability mechanism protects the integrity of the system only in the event that the value of BitShares falls and the BitUSD market price remains at $1. It does not directly stabilize the price of BitUSD around $1, in our view. If the price of BitUSD deviates from $1, this mechanism may not help correct the price. In our view, it is important to draw the distinction between a mechanism designed to protect the value of collateral and that of a mechanism which directly causes the price of the stablecoin to converge.

Weaknesses

Exposure to a fall in the value of collateral – BitShares was a new, untested and low value asset, and therefore its value was volatile. If the value of the token falls by 50% sharply, in a period spanned by one of the loans used to create BitUSD, there may be insufficient collateral and the peg could fail.

Lack of a price oracle – In our view one of the most controversial aspects of this design is the absence of any price oracle mechanism, providing the system with real world exchange rates. However any price oracle system is challenging to implement and may introduce several weaknesses and avenues for manipulation. We will talk more about this in part 2. In our view, the only real way around this may be that any stablecoin system may require a price feed from a distributed exchange, which can in theory publish a distributed price feed from real world US Dollar transactions. The distributed exchange in BitShares did not allow “real USD”. A distributed exchange system like Bisq, without a central clearing could in theory allow “real USD” prices and provide a distributed price feed.   Therefore stablecoins may eventually be considered as a layer two technology on top of liquid and robust distributed exchange platforms, should these systems ever emerge.

Manipulation – Trading volume in the Bitshares vs BitUSD market on the distributed exchange platform was low, it was therefore possible for block producers to manipulate the market by causing the value of Bitshares to fall relative to BitUSD, enabling them to obtain Bitshares at a discount.

Lack of any price stability mechanism – The main weakness of the system is the lack of any mechanism to move the price towards $1, other than the “where else would it trade?” logic.

Daniel Larimer’s defence of the system

In Daniel’s view, the mechanism of BitUSD creation is analogous to how USD are created in the economy, in that financial institutions lend them into existence.

It’s the same way dollars are created in the regular banking system. Dollars are learnt into existence backed by collateral, in the case of the current banking system the collateral is your house. In the case of our system its shares in the DAC itself.

(Source: Lets talk Bitcoin episode 129)

 

In a way Daniel is correct here, however as we explained in the introduction to this piece, these synthetic dollars are far less reliable than those created by more traditional banks, and can be considered as a whole new layer of risk, as they are even further away from base money. In addition to this, when obtaining a bank loan, the bank typically has a legal obligation to provide the customer physical cash should they demand it. While such an outcome for BitUSD holder is possible, its not a legal obligation for the creators of BitUSD. Although obviously banks typically do not have the cash in reserve to pay back their deposits, we think the fact they have a legal obligation to do so is an important distinction to draw when comparing BitUSD to US Dollar banking deposits.

In response to the supposed weakness of a lack of a price peg, Larimer argues in favor of his “hypothesis that the price feed is unnecessary” as follows:

It implements automatic margin calls, such that if the price moves against someone who is effectively short, it forces them to cover and buy it back in the market and that creates a peg. The market peg works on the premise that all market participants buy and sell based on what they think market participants will be buying and selling in the future. The only rational choice is to assume that it’s going to trade based on the peg in the future. If you don’t believe that they you have to decide on which way it’s going to go, up or down. And if you don’t have a way of saying you abstain from the market. If you don’t think it works you sell the shares and get out, as the systems going to fail in the first place. So its a self reinforcing market peg, that causes the asset to always have the purchasing power of the dollar.

(Source: Lets talk Bitcoin episode 129)

 

In our view this idea that a price of $1 is the “only rational choice” is a weak argument. It is basically saying that if the price is not $1, then what will it be? This logic may hold true for some periods, but it is not sustainable and will not scale, in our view.

Conclusion

The volume of BitUSD in existence was a lot lower than many had hoped, in some periods there was only around $40,000 in issuance. At the same time liquidity was very low and the price stability was weak, as the below chart illustrates. The main architect of BitUSD went on to propose a new stablecoin SteemUSD in 2017, this time including a price feed system. Therefore we consider BitUSD as an interesting early experiment, it did not achieve what was hoped nor did it build a robust stablecoin.

(Source: Coinmarketcap)

 

Case Study 2: MakerDAO (Dai) – 2017

Factbox
Coin Name	Dai
Launch Date	27 Dec 2017
Crypto Collateralized	Yes
Price Oracles	Yes (indirect)

The next stablecoin we look at is Dai, which exists on the Ethereum platform. This system is highly complex, with four relevant pools of funds and six possible stability mechanisms. There are currently around $50 million worth of Dai in issuance and the peg seems to be holding up reasonably well.

System dynamics

Pools of Funds	Description
Ethereum	Ethereum is the native token of the Blockchain platform used for Maker & Dai
Pooled Ethereum	Ethereum is placed in pools used as collateral for issuance of the Dai token. These are often referred to a collateralized debt positions (CDPs)
Dai	Dai is an ERC-20 token that is generated by collateralizing pooled Ether. Dai is the stablecoin token, designed to be valued at $1.
Maker	The Maker token is MakerDAO’s governance token. It is used to vote on various initiatives that pertain to the stability of the ecosystem. It is also mandatory to possess during the collateral unlocking process. During such a process, a stability fee is garnered from the user, where payment is accepted exclusively in Maker. Maker is also an ERC-20 token.

 

Groups of Participants	Description
Dai Creators	An individual who sends Ethereum to a smart contract, locking up Ethereum in exchange for Dai. These people are also known as CDP owners.
Dai Holder/User	A Dai holder may or may not be a Dai creator. They may invest in or use the Dai stablecoin token.
Maker Token Holders	Maker token holders vote on several functions and parameters of the MakerDAO system. They manage aspects such as stability fees and liquidation ratios, as well as having responsibility to nominate other groups.
Keepers	These traders monitor the Dai collateral and if it falls to an insufficient level, purchase the collateral in an open auction, by spending Dai.
Oracles	Price feed producers submit price information that is aggregated and used to select a given price for both Maker and Ethereum (but not Dai itself). These agents are nominated by MakerDAO token holders. In order to prevent manipulation, there is a one hour lag between the price publication and when it impacts the system. In addition to this a median type mechanism is used to select the price, which involves ignoring the highest and lowest prices. In our view this may not prove to be robust enough if the oracles have a conflict of interest and try to engage in manipulation.
Global settlers	This is another group nominated by the MakerDAO token holders. This group can unwind the entire Dai system, by giving Dai holders the right to redeem their collateral at a fixed price.

 

Price Adjustment Mechanics	Price Direction	Description
Dai Redemption	Positive	The primary stability mechanism is the ability, in theory, to redeem Dai for $1 worth of Ethereum. Redemption can only be conducted by CDP owners (unless there is insufficient collateral). If the price of Dai falls, CDP owners need to either use Dai they currently hold or buy it in the market, and then they can redeem/delete Dai for $1 worth of Ethereum based on the price feed provided by the price oracles.
Dai Creation	Negative	To complement the Dai redemption process, the mechanism to prevent the price of Dai climbing too high, is the ability of Ethereum holders to create new Dai, by placing Ethereum inside of CDPs.
Target rate (Not active)	Both directions	There is a “Target Rate Feedback Mechanism” (TRFM), which appears to be another price stability mechanism in the system. However, it is not yet active nor have several specifications of the mechanism been worked out yet. The the idea is that a target rate is set by the MakerDAO token holders. The target rate is essentially a spread which applies to the creation or redemption of Dai, designed to correct the price.
CDP liquidation (indirect)	Positive	There is a mechanism by which traders/keepers can redeem the Ethereum collateral held by another CDP. This can only occur if the value of this collateral falls to an insufficient level to backup the Dai, in this case 150% of the value of Dai. This should incentivise CDP owners to keep topping up their CDPs to ensure there is a large buffer of Ethereum. This is a necessary mechanism to ensure the integrity of the system and ensure the value of the collateral is always sufficient. However it is not clear if this directly keeps the value of Dai at $1. This mechanism can be thought of as a building block on the stability mechanism, which merely ensures the level of collateral is sufficient. Other redemption systems are needed to make this meaningful, in our view.
Global Settlement	Positive	This mechanism can be triggered at any time. The triggering essentially gives all Dai holders an option to convert back to a fixed value of Ethereum, worth $1 according to the oracle price feed, at the time of the triggering (or whatever price is possible given the total level of collateral in the system). The difference between this and normal redemption, is that the price is fixed and its open to all Dai token holders and not paired to a particular CDP. The idea is that this mechanism can be used as a threat against CDP holders, to ensure they keep redeeming Dai in the event the price falls, rather than holding out for an even lower price. Global settlement can also be used in the event of bugs or other emergencies.
MakerDAO token issuance (indirect)	Positive	MakerDAO token holders act as the buyer of last resort. If the collateral (pooled Ethereum) in the system were to drop below 100% collateralization, MakerDAO is automatically created and auctioned on the open market to raise additional funds to collateralize the system. Hence, if the system becomes undercollateralized, Maker holders absorb the damage. Again this mechanism protects the value of collateral, but does not directly help the price of Dai converge to $1, in our view.

Analysis of the core stability mechanism – Dai redemption

The primary stability mechanisms appear to be the ability of CDP owners to redeem if the price of Dai is too low and for people to create new Dai if the price is too high. For example if the price of Dai falls to 80 cent, CDP owners could purchase Dai in the market and redeem it, unlocking $1 worth of Ethereum and making a nice profit. This is how the system should work under normal circumstances.

The above appears to be a robust stability mechanism which should keep the price of Dai at or near $1. However, the theory may only work if CDP owners expect the price of Dai to correct back to $1. If the price of Dai has fallen to 80 cent, CDP owners may be reluctant to redeem if they expect the Dai price to fall further to 60 cent, as such a price would enable them to make even more profit. There is no guarantee that once the price reaches 80 cent, it won’t continue to fall.

Therefore the stability mechanism could depend somewhat on the power dynamics between two groups, Dai owners and CDP owners. These two groups are essentially trading against each other in the market, Dai owners are selling of Dai and CDP owners are the potential buyers. If the power balance shifts towards CDP owners, such that they are well capitalised, patient, collaborative and determined, this group could outmaneuver the Dai token holders, drive the price down, and then buy it back and make a large profit. This may seem unlikely, but in our view the stability mechanism may not work in all market scenarios. Although we consider Dai as superior to BitUSD, in some limited ways, the Dai peg relies on market psychology and investor expectations, in the same way as BitUSD. Therefore the Dai peg is also weak and unlikely to scale.

The global settlement system can mitigate the above risk. If CDP owners are successfully manipulating the price of Dai down too far, this could trigger global settlement. Dai holders would then get around $1 of Ethereum back. Therefore the threat of global settlement may keep the price of Dai up. However again the effectiveness of this threat depends on the determination of the various groups, the CDP owners, MakerDAO token holders and global settlement activators.

Conclusion

We consider Dai to be one of the most sophisticated and advanced stablecoins systems which has been produced so far. In our view, when digging into Dai’s stability mechanisms, there is no one powerful mechanism which ensures stability. Instead we have a complex network of systems, which to some extent reference each other and use circular logic.  One could claim this complexity was created to obfuscate the lack of a strong and clear stability mechanism, but it is more likely to be an indication of an experimental trial and error type approach to the design of the system.

Therefore the system is still reliant on investor expectations and psychology, although to a lesser extent than the BitUSD. While the stability systems in place could work, at least for a while, we think they are not robust enough to withstand market turmoil or some types of power imbalances between Dai holders and CDP owners. Therefore, the search for the holy grail continues.