AQTIS - Mean Reversion (Long)
We are excited to announce the release of our first performance report for our mean reversion trading strategy. In this report, we will showcase how we utilise mean reversion on 4H timeframes across various assets, as well as the potential returns that AQTIS can generate.
Introduction
AQTIS strives to create a diverse system that comprises various strategies to form an intelligent system overall. Over the past year, we have developed five strategies (with an additional two in the next month) to create a diverse quant technology stack.
Our AI utilises the best strategies for the appropriate market conditions since not all strategies work well in every situation. Some strategies only flourish in specific market moments. A linear approach to web3 markets is insufficient for long-term profitability.
In the coming weeks, we will present performance tests of our individual strategies and we'll close off with an overall performance data . We optimise each strategy based on performance vs. exposure, ensuring that liquidity is used as efficiently as possible. This approach makes it possible for AQTIS to function as a complex system. This update focuses on our first strategy, Mean Reversion V0.1.
Mean Reversion
As promised, we will now present the results of our mean reversion strategy.
Applied in High Time Frames (HTF). We are using this strategy for Swing Trading and as a trigger for a change in the market regime.
Simply put, mean reversion strategies are based on the idea that asset prices tend to return to their average values. This suggests that when a movement has become overextended and we are in areas of overbought or oversold conditions, the market tends to return to the original trend.
The Process
Our team of data scientists and quantitative traders worked together to gather all relevant metrics from cryptocurrency markets. Once we completed our database, we conducted an exploratory analysis of the time series to identify correlations and causalities.
After filtering out the most relevant time series, we created various statistics and fed them into a machine learning system that helped us predict movements and optimise our strategy's parameters.
The Model
Once our machine learning model was trained, the system generates a series of inputs and outputs for our trades. To optimise our entry and exit points, we used a Multiple Time Frames analysis.
This means that to make a trade on a daily time frame, we looked for confirmation on lower time frames (4H, 1H) to obtain higher probability trade.
Verification
As we know, many machine learning models perform well in trading systems because they tend to overfit. The system over-adjusts the strategy's parameters to obtain the best possible results, but the system is not profitable in real-time with new data.
Therefore, we applied a cross-validation with Walk Forward Analysis (WFA) to determine if the model is profitable in real-time.
Key trading metrics
Sharpe Ratio
Measures the risk-adjusted performance of the strategy by taking into account the returns and volatility.
Maximum Drawdown
Measures the peak to trough decline of a portfolio in percentage terms.
Profit Factor
The ratio of the strategy's gross profit to its gross loss.
Total Return [%]
Measures the capital gain or loss generated by the strategy over a specific period of time.
Benchmark Return
Measures the capital gain or loss generated holding an asset over a specific period of time.
Time Exposure [%]
Measures the total amount of time that an investment is held within a portfolio.
Measuring the strategy's performance includes a set of well-defined statistical indicators. It's crucial to analyze these indicators to identify the strengths and weaknesses of a trading strategy.
Backtesting results - Ethereum
Asset: Ethereum [ETH/USDT]
Strategy: Mean Reversion Direction: Long Only
Time frame: 4H and 1H as entry confirmation
Start 2022-01-01 00:00:00+00:00
End 2023-04-20 12:00:00+00:00
Period 474 days 16:00:00
Start Value 100.0
Min Value 97.8641
Max Value 260.163032
End Value 260.163032
Total Return [%] 160.163032
Benchmark Return [%] -46.774773
Total Time Exposure [%] 11.025281
Max Gross Exposure [%] 100.0
Max Drawdown [%] 9.418241
Max Drawdown Duration 75 days 00:00:00
Total Orders 98
Total Fees Paid 3.152463
Total Trades 49
Win Rate [%] 36.734694
Best Trade [%] 7.742659
Worst Trade [%] -1.237563
Avg Winning Trade [%] 7.742659
Avg Losing Trade [%] -1.237563
Avg Winning Trade Duration 2 days 06:00:00
Avg Losing Trade Duration 0 days 09:09:40.645161290
Profit Factor 3.619194
Expectancy 3.268633
Sharpe Ratio 2.823432
Calmar Ratio 11.53047
Omega Ratio 1.699448
Sortino Ratio 6.110543Visual statistics and ETH Chart:
Backtesting Results - Bitcoin
Asset: Bitcoin [BTC/USDT]
Strategy: Mean Reversion Direction: Long Only
Time frame: 4H and 1H as entry confirmation
Start 2022-01-01 00:00:00+00:00
End 2023-04-20 12:00:00+00:00
Period 474 days 16:00:00
Start Value 100.0
Min Value 91.543874
Max Value 209.889718
End Value 209.889718
Total Return [%] 109.889718
Benchmark Return [%] -37.551071
Total Time Exposure [%] 19.066011
Max Gross Exposure [%] 100.0
Max Drawdown [%] 12.260654
Max Drawdown Duration 42 days 16:00:00
Total Orders 84
Total Fees Paid 2.206435
Total Trades 42
Win Rate [%] 28.571429
Best Trade [%] 9.738264
Worst Trade [%] -1.237563
Avg Winning Trade [%] 9.738264
Avg Losing Trade [%] -1.237563
Avg Winning Trade Duration 5 days 01:00:00
Avg Losing Trade Duration 1 days 00:00:00
Profit Factor 3.31045
Expectancy 2.616422
Sharpe Ratio 2.133246
Calmar Ratio 6.267801
Omega Ratio 1.375034
Sortino Ratio 3.887623Visual statistics and BTC Chart:
Backtesting Results - Binance Coin
Asset: Binance Coin [BNB/USDT]
Strategy: Mean Reversion Direction: Long Only
Time frame: 4H and 1H as entry confirmation
Start 2022-01-01 00:00:00+00:00
End 2023-04-20 12:00:00+00:00
Period 474 days 16:00:00
Start Value 100.0
Min Value 98.644292
Max Value 203.603235
End Value 193.228694
Total Return [%] 93.228694
Benchmark Return [%] -36.267496
Total Time Exposure [%] 5.372191
Max Gross Exposure [%] 100.0
Max Drawdown [%] 7.772283
Max Drawdown Duration 57 days 00:00:00
Total Orders 70
Total Fees Paid 2.058748
Total Trades 35
Win Rate [%] 34.285714
Best Trade [%] 8.740461
Worst Trade [%] -1.237563
Avg Winning Trade [%] 8.210344
Avg Losing Trade [%] -1.237563
Avg Winning Trade Duration 0 days 21:40:00
Avg Losing Trade Duration 0 days 15:18:15.652173913
Profit Factor 3.17129
Expectancy 2.663677
Sharpe Ratio 2.51665
Calmar Ratio 8.485308
Omega Ratio 1.966319
Sortino Ratio 5.606496Visual statistics and BNB Chart:
Backtesting Results - Solana
Asset: Solana [SOL/USDT]
Strategy: Mean Reversion Direction: Long Only
Time frame: 4H and 1H as entry confirmation
Start 2022-01-01 00:00:00+00:00
End 2023-04-20 12:00:00+00:00
Period 474 days 16:00:00
Start Value 100.0
Min Value 95.465535
Max Value 271.282019
End Value 271.282019
Total Return [%] 171.282019
Benchmark Return [%] -86.63431
Total Time Exposure [%] 7.58427
Max Gross Exposure [%] 100.0
Max Drawdown [%] 10.793209
Max Drawdown Duration 44 days 08:00:00
Total Orders 66
Total Fees Paid 2.157945
Total Trades 33
Win Rate [%] 42.424242
Best Trade [%] 10.736066
Worst Trade [%] -2.235365
Avg Winning Trade [%] 10.736066
Avg Losing Trade [%] -2.235365
Avg Winning Trade Duration 1 days 23:25:42.857142857
Avg Losing Trade Duration 0 days 10:31:34.736842105
Profit Factor 3.493504
Expectancy 5.190364
Sharpe Ratio 2.447259
Calmar Ratio 10.693655
Omega Ratio 1.748385
Sortino Ratio 4.818083Visual statistics and SOL Chart:
Conclusions
What's next?
With our data-driven approach and commitment to innovation, we are confident in our ability to stay ahead of the curve and deliver exceptional returns.