I audited every losing trade looking for a pattern
I have spent some time analyzing 1,166 trades from my core trend-following system to see if I could identify common traits in losing positions.
I have spent some time analyzing 1,166 trades from my core trend-following system to see if I could identify common traits in losing positions. By tagging each entry with variables like ATR (average true range, which is a measure of volatility), ADX, RSI, time, and day of the week, I wanted to see if I could filter out the losers to improve my overall results. The analysis revealed two distinct patterns in the losing trades. The first is that trades with a very short holding period, specifically those under 12 bars, almost always result in a loss. While this is a clear sign of a failed trend, it is only visible in hindsight. Because there is no way to identify these trades at the moment of entry, they cannot be filtered out. The second pattern is more interesting because it is visible before the trade is placed. Trades entered when the RSI (a momentum indicator) was low, indicating a weak breakout, consistently showed a negative expected value.
| Metric | Weak Momentum (RSI 30-60) | Strong Momentum (RSI 72+) |
|---|---|---|
| Win Rate | 24.4% | 45.5% |
| Expected Value | -0.068% | +0.142% |
| Seeing these numbers, I tested an RSI filter to exclude these weak breakouts. The results were underwhelming. While it slightly improved the profit factor (PF), which is the ratio of gross profit to gross loss, in the in-sample weak period from 0.98 to 1.02, it had no impact on the out-of-sample (OOS) performance, where the PF remained flat at 1.47 to 1.48. In other words, although I successfully identified the losing trades, removing them did not improve the system because it also meant cutting out the occasional winners that were mixed in with those weak breakouts. | ||
| This highlights a fundamental reality of trend following. The profits in this strategy rely on a few “fat tail” events, which are large outlier gains. By trying to avoid small losses, I end up sacrificing the very trades that contribute to those major wins. The “weak momentum” filter does not make the system more profitable; it only serves as a conservative option to slightly increase survival during the worst market regimes. | ||
| Ultimately, this data-driven deep dive confirms that my core system is already quite robust. Even when I can pinpoint why a trade loses, the structure of trend following is such that there is simply no room left to “trim the fat” without hurting the system’s overall performance. I have decided to keep the RSI filter as an optional setting for extreme market conditions, but the version 1.4.1 logic remains unchanged. |