Improving the backtest

Correct backtesting

Unfortunately during my tests I discovered a bug that was increasing incorrectly the performance of HMA strategies (BTC and  big caps). The bug is fixed and the live backtest page is showing the correct values.

You can imagine I was not very happy to discover that. But on the other side the current results are more realistic and shows an interesting behaviour. During 2017 the low cap BMP strategy was outperforming the big caps HMA hybrid strategy. Last year was spectacular for alts and the calculated performance makes sense. In contrary, during the 2018 bear market, the HMA hybrid strategy is more performant. HMA hybrid is quite fast to switch back to USD, protecting the profits. BMP strategy moves slower, and is obliged to stay much longer in alts to catch the potential big alts uptrends.

screenshot_02 copy
Current backtest performance

Transaction fees

In order to have a more realistic backtesting algorithm I decided to integrate transaction fees during phase transitions. Trading fees are quite high in crypto, around 0.2%. Moreover you have to consider that switching from a 100% BTC/USD position to a TOP30 alts index is not a perfect operation, there are delays, market spread, volatility. To take in account I’m using a total 0.5% fees on the phase changes during the backtesting. With these changes the current backtesting should be more realistic and reliable.

Performance tuning

With the improved backtesting unfortunately the HMA strategy performance is lower. But I don’t want to leave you with bad news. I started a systematic test of different HMA and momentum parameters. The testing will take some time, but preliminary results show that there is a very good potential for improving the performance. Finger crossed!