After over a year in existence, it is time to analyse the predictive properties of the XBTUSD funding rate. The XBTUSD 100x leveraged contract is a Bitcoin / USD total return swap that has no expiry date. To anchor the price of the swap back to the spot market, an interest payment (we call this funding) is exchanged between longs and shorts. The interest rate by and large is determined by the previously observed 8-hour time weighted average premium of the swap vs. the spot price.
The funding rate is published with an 8-hour grace period before it is charged. That allows traders who do not wish to pay or receive funding to exit their positions before the funding timestamp. The question is, can you predict the future price of Bitcoin by the published funding rate?
I have analysed data from March 2017 until now. My data series consists of the funding rate every 8 hours, and the log return of the XBTUSD swap over the next 8 hours.
T0: Now
T1: 8 hours in the future
X-axis: Funding Rate published T0 to be charged at T1
Y-axis: Log(XBTUSD P1 / XBTUSD P0)
Simple Regression
The above chart is a XY scatter plot of the data. The chart clearly illustrates the funding rate contains no significant predictive power.
Digging Deeper
When the funding is extremely positive or negative, this could signal a reversal in the market’s direction i.e. mean reversion. Using an extreme funding rate as the signal, we can take the counter trend position.
Example:
If the published funding is at the maximum +0.375%, does that predict with greater accuracy whether the return of XBTUSD in the next 8 hours will be negative?
To further analyse this hypothesis, I calculated the sample mean and standard deviation of the funding rate in basis points (bps).
1bps = 0.01%
Mean: 1.66bps
Standard Deviation: 17.13bps
I constructed one and two sigma bands. I then conducted mean reversion tests.
1 Sigma = 1 Standard Deviation
Hypothesis:
A large negative funding rate predicts a positive return for the next 8 hour period. A large positive funding rate predicts a negative return for the next 8 hour period.
The magnitude of the funding rate tested depends on the number of sigmas away from the mean.
The following table lists the results.
| Sigmas | Funding Rate | Sample Size | % Success | Cumulative Funding | Cumulative XBTUSD Return | Cumulative Return | % of Total Observations |
| -2 | -32.61 | 21 | 47.62% | -7.71% | 3.36% | 11.07% | 4.01% |
| -1 | -15.47 | 62 | 53.23% | -16.76% | -18.15% | -1.38% | 11.83% |
| 1 | 18.80 | 81 | 45.68% | 25.10% | -14.77% | 10.33% | 15.46% |
| 2 | 35.93 | 36 | 44.44% | 13.49% | 6.90% | 20.39% | 6.87% |
Sample Size – Out of 524 funding periods, this is the number of times that the funding rate was less than or equal to the sigma adjusted test (assuming a negative funding rate).
% Success – Out of the sample size, this is the number of times where the funding rate was negative and the next period return was positive or vice versa.
Cumulative Return – This is the net return, including funding, of both success and failure situations. If the funding rate is negative, you go long, and you receive funding because the rate is negative. If the funding rate is positive, you go short, and you receive funding because the rate is positive.
% of Total Observations – Sample Size / 524 (Total Number of Funding Periods)
Conclusion
The data clearly illustrates that traders may use an extreme funding rate as a signal to take the counter trend position. The added benefit of receiving funding for bucking the trend is what provides a significant majority of this strategy’s returns.
A simple trading algo can be constructed to capture this alpha. At each funding timestamp, if the funding rate is above or below your limit, place the counter trend trade. Immediately after the next funding timestamp, close your XBTUSD position.
The one caveat is the sample size is still relatively small. I will revisit this study early next year to observe if the results change.