UPDATE: November 2021 brought difficult cryptocurrency markets and the strategy hit a drawdown which meant I paused trading.
In early August 2021, I launched a crypto trading strategy called the Crypto Alpha One fund. The starting capital of the fund was £1000, which I funded with my own money.
Before launching the strategy, I tested it on historical cryptocurrency data. This testing is an effective way of objectively evaluating the idea the trader has. The idea could be something the trader has seen in the market.
Alternatively, the trader may have seen something about how the price behaves at certain key price levels. This observation becomes the hypothesis that trader believes might give them an edge, an advantage over traders.
Around the observation, the trader creates some rules. These could be straightforward rules such as, if this happens, do this, or if that happens, do that.
The trader then applies the rules to the historical price data of a portfolio of instruments. A key point to understand is that testing a set of rules on a single instrument such as Apple stock, Gold futures, or the EURUSD will lead to skewed and unreliable results.
Backtesting rules on a single instrument is flawed
When backtesting, the trader should aim towards a robust set of results. The results will influence whether the trader decides to trade the rules with real money.
Without confidence in the results, the trader places themselves in a vulnerable spot. If the results from live trading don’t match the results from the backtests, the trader can start to question their rules and the confidence they originally had from the backtest.
In my opinion, backtests should have two critical elements.
Firstly, historical price data going back ten, twenty, or more years. Additionally, the backtest should be on a portfolio basis. In other words, the trader should select several instruments such as 20 or 30 futures instruments or stocks and apply the rules across all instruments.
The result will be far more robust and give confidence that even in several different market environments such as a pandemic, credit crisis, or dot com boom and bust, the trader can consistently apply the rules in live trading.
Cryptocurrencies are still a very new addition to the investing world, and most coins have only been available for around five years. With this in mind, the backtest performed on the Crypto Alpha One strategy would be limited.
That said, the results from the backtest were encouraging enough to live trade, albeit with a small amount of capital that I could afford to lose. The strategy was backtested in the QuantConnect platform, using the coding language of Python.
The goal with the live trading is to try and replicate the backtest results. However, I need to be careful about making conclusions too early; I need five or six months of living trading results.
This is the same for you or any other swing trader. If you’re going to do your own backtests and then try and replicate those results in live trading you need a large enough sample size.
In the short term, randomness and luck (both good and bad) will dominate your trading results.
Starting live trading of a long-only strategy at the beginning of a strong bullish trend will give you different results to starting live trading at the end of, for example, 2018, when the crypto market started to nose-dive. The same strategy, but the results from the first will give vastly different results to the second, and you’re at risk of drawing the wrong conclusion from both.
To avoid this, you’ll need to start with the assumption that no conclusion about the profitability of the strategy for the first few months.
The Crypto Alpha One is a long-only strategy and trades a portfolio of cryptocurrencies. There is no leverage involved, so it is cash only. The code determines the size of each investment based on the parameters I’ve coded into the strategy, so each cryptocurrency receives a portion of the Fiat cash (GBP) balance.
The algorithm runs once a week, early on Monday morning. The strategy decides what to do with any existing positions in the portfolio. So the code will either increase the investment, decrease the investment or sell 100% of the investment.
An API connects the algorithm to the Coinbase Pro platform, so the process I’ve written runs without my interference. I’m interested to see how the strategy performs, and I do regular reports to my YouTube channel.