Trading bitcoin using a fast Shepherd's Momentum signal
Posted by
Joao TeixeiraRelated reading
What are Byzantine fault tolerance (BFTs)?
A brief introduction to Byzantine fault tolerance
Introduction
This blog is the first of a series of four where I explore using our proprietary Shepherd's momentum (SM) signal to trade cryptocurrencies.
The complete series list is:
- this blog, "Trading bitcoin, using a fast Shepherd's momentum signal",
- "Trading ethereum using a fast Shepherd's Momentum binary signal" (to come soon),
- "Trading ripple (XRP) using a fast Shepherd's Momentum signal" (to come soon),
- and, finally, "Trading solana using a fast Shepherd's Momentum binary signal" (to come soon).
In this series of blogs I backtest two different version of the SM, all with no short sell allowed, to trade bitcoin, ethereum, XRP and solana.
In this blog I will use a no shorts version of the Shepherd's momentum to trade bitcoin. I aim to make this blog both accessible to beginners and engaging for experienced investors.
This blog is not an invitation to trade. Also does not intend to provide trading advice. Trading involves risks, and readers are solely responsible for their trading decisions.
The next graph shows bitcoin's price and volume, on a semi-logarithmic scale, since the first of January, 2014. I start at this date since at this point in time bitcoin's volume is already quite reasonable.
As it is usual in finance I will compare Shepherd's momentum performance with a Buy and Hold (B&H) strategy. The buy and hold strategy serves as a simple benchmark to evaluate whether the Shepherd's momentum adds value. If the Shepherd's momentum strategy underperforms the buy and hold strategy over time, it may not be worth the extra effort and costs.
Shepherd's momentum on bitcoin
As discussed in my Shepherd's momentum blog, the SM is a twist on the traditional moving average crossover indicator. SM is defined between -1 and 1. A value of 1 would mean full long and a value of -1 fully short. Values in between indicate that part of the capital is invested and that the remaining is held in cash.
The following graph shows the bitcoin price together with the Shepherd's momentum signal.
I assume that most investors are not able to easily short sell bitcoin , consequently if the signal is less than zero, the position will be zero. That is, I will not have short positions.
The number of units of bitcoin held (the position) is given by: position = signal x capital / price, where signal is a value greater or equal to 0 and less or equal to 1.
Shepherd's momentum and buy and hold performance
The next graph shows, on a semi-logarithmic scale, the Shepherd's momentum and buy and hold portfolios value over time, starting with an initial capital of 1000 USD.
The next graph shows the Shepherd's momentum and buy and hold strategy percent invested. While B&H is always 100% invested, SM is mostly only partially invested.
The next table shows the main statistics for the both SM and B&H strategies.
| Name | Shepherd's momentum | Buy and hold |
|---|---|---|
| Initial Capital (USD) | 1,000 | 1,000 |
| End Capital (USD) | 233,990 | 124,169 |
| Mean daily log returns | 0.13 | 0.11 |
| Highest return | 25.04 | 25.09 |
| Lowest return | -15.32 | -51.81 |
| Yearly standard deviation | 37.19 | 70.51 |
| Information ratio | 1.31 | 0.61 |
| Skewness | 0.84 | -0.89 |
| Kurtosis | 18.08 | 14.02 |
| Historical VaR 95% | 2.40 | 5.68 |
| Conditional VaR | 4.71 | 9.14 |
| Maximum % drawdown | 45.84 | 83.34 |
As it can be seen in the table above, the Shepherd's momentum strategy is superior to buy and hold in most of the presented indicators, in particular:
- SM's Information Ratio is more than twice that of B&H.
- In spite of being frequently out of the market, by the end of the backtest the SM's capital is almost double than B&H.
- Both Value at Risk (VaR) and Conditional VaR (CVaR) are better for the SM strategy. The CVaR (also called expected shortfall - ES) is particularly relevant because it gives the expected loss in the worst 5% (this threshold was arbitrarily set by me) of cases.
- SM's maximum drawdown is almost half of B&H.
The kurtosis is better (lower) for the B&H strategy. That is due to the fact that SM has many zero returns, which makes the returns distribution thin at or close to the mean and, by consequence, the returns' distribution is fat tailed. The logarithmic returns distribution for both strategies can be seen on the two following charts.
Note also that while SM yearly standard deviation is about half of B&H is still high at about 37%. This means that a yearly loss of 37% is just one standard deviation away, making it a likely occurrence over time. Not everyone can deal having the value of their investment cut by 40% in a year.
Final remarks
There are a few points to have in mind:
- The results presented here are a simulation (a backtest).
- Trading involves risks, and readers are solely responsible for their trading decisions.
- Losses, when investing in leveraged products like futures and options, may exceed the original investment's value.
- Past performance, even when actually realized, is not indicative of future results.
- Buy and hold is easy and low-cost. Active trading is complex and incurs costs (slippage and commissions).
- Trading costs were not considered. Trading costs can significantly impact performance, especially in a relatively fast-moving strategy like this version of the SM.