Adrian Reid explains why equity-curve regime filters, which toggle strategies on and off based on recent performance, fail in practice. A 30-year backtest typically triggers only 5 to 15 regime changes, too few to validate the rule reliably. The result is overfitting: apparent success in backtests that won't survive live trading.
Equity curve filters in a trading strategy that turn a system on and off depending on how well it's performing sound like a great idea, but in fact, they're one of the trickiest parts of a trading system to get right. Let's talk about why. This turns out to be the most influential rule in the system if you add it. Now, when the system turns on, you're gonna have to buy in to everything, and when the system turns off, you're gonna have to sell out of everything. So there's a lot more transactional friction in this. There's one more bigger problem. And imagine this rule is the equity curve must be above the 200-day moving average of the equity curve. Okay, just to keep it simple, how many times does that rule come into play in your entire backtest? In your backtest, how many times does a regime filter turn on and off roughly in a 30-year backtest? What do you reckon? Few times, maybe 10, maybe 5, 10, 15 times depending on the filter, right? That's 5, 10, 15 data points to make a decision if you're going to turn your system on and off based on its equity curve. And that's only happened 5, 10, 15 times in the backtest. How do you choose the rule to turn your system on and off? It's a trick question. Very, very carefully, because there's not enough data points to really make that decision. This becomes the most important rule in the system, and you've only got a handful of data points to make the decision on. So it might work, and if we had infinite data history, we could probably make some good decisions here. But there's a very high chance of overfitting because we just don't have enough data. So I would say there's, there's actually, there's probably quite a few trading concepts that if you had infinite data to really validate, then they might actually be helpful. But given we don't have infinite data, the risk of overfitting to the 5 or 10 data points we do have is just too high to risk using the technique. If you want to learn more about systemizing your trading, go to enlightenedstocktrading.com/free to get your free Trader Acceleration Bundle to learn all about systematic trading and how it can help you. There's a link in the description, and don't forget to subscribe so you don't miss the next video.
