10/10/22 - Deep Dive on Cleargate Backtesting

Launch App

10/10/22 - Deep Dive on Cleargate Backtesting

BTC-ETH Momentum and Trend Strategies Performance

We wanted to provide some insight into the backtesting performance of the momentum and trend following strategies in BTC and ETH. For those that don’t know, backtesting is a way to see how well the trading algorithms would have performed in the past, and can serve as a useful reference for potential future performance. ClearGate has done extensive backtesting of both strategies as they developed and refined the machine learning algorithms that will implement these strategies, so users can feel confident in the historical basis for these Cellars.

Disclaimer: These metrics aren’t set in stone and do not promise any definitive results in the future.

The benchmark (black line) used for this backtesting is a 50/50 portfolio of BTC and ETH, spot holding.

BTC-ETH Trend following strategy

Performance graph and info: (black line for benchmark)

Backtesting results in brief:*

* past performance is not indicative of future results.

After a first glance, the most notable metrics displayed here for most people would probably be the win rate and APY. In traditional finance, anything near or over 100% is almost unheard of, it's very difficult to achieve these metrics. Another important interpretation is that the best performing month resulted in a 52.51% return, while the worst month resulted in a negative return nearly 3 times as small, -14.21%. After compounding returns over 3.5 years, a staggering profit of 1576.59% is achieved. To the financially inclined, the sharpe ratio for this trend following strategy is eyebrow raising, to say the least. For those that aren’t aware, sharpe ratio is a measure of risk adjusted return, the higher the value the more excess return relative to risk. A sharpe ratio above 2.0 is extremely impressive and uncommon.

BTC-ETH Momentum Strategy

Performance graph and info: (black line for benchmark)

Backtesting results in brief:*

* past performance is not indicative of future results.

Generating similar results as the trend following strategy, the momentum strategy yields 84% APY and a 62% win rate! As the strategy that’s more sensitive to price appreciation of the two, you can see that its best month (51.97%) and worst month (-27.50%) are both of greater magnitude than with the trend following strategy. Similarly, its worst drawdown is significantly larger than in the trend following strategy, which can completely exit BTC and ETH into USDC, whereas the momentum strategy always maintains some exposure to those assets. Additionally, the Sharpe ratio is very impressive here, boasting a 1.402 ratio, indicating the excess returns warrant the risk.

The win rate depends more on how the strategy is set up. The actual return is affected by how big the wins and losses are. For example, the win rate can be very high but if all wins are very small and losses are very big, then the strategy will still lose money. Here the win rate and loss rate are more dependent on what have been the market conditions. If the market is trending, the win rate will be high. However, if the market is moving in zig-zags (which might happen in the future), then the win rate might be clearly lower. It shows the results of individual trades, not the overall picture as trade size will not be constant in time (trades can be smaller or bigger depending on quant signals received).

In reality all of the positive metrics in backtests can be achieved by over optimizing the strategy which may have little value for the future. For an experienced eye, looking at various metrics together can give more indication about the characteristics of the strategy and thus it is useful not to concentrate too much on one particular metric (although Sharpe ratio incorporates both return and risk and thus is quite useful). We tried not to over optimize the strategies and followed all the best practices of backtesting but nevertheless all the metrics here become more relevant when the strategy is already running. It always takes some time for the benefits of the strategies to emerge and those metrics are not constant in time but averaged over the whole 3,5 year backtest period.

We hope that this backtesting provides more depth into our upcoming strategies that we are so excited to share with you!