Signal Decay Awareness
Every trading edge has a half-life — assume yours is shrinking.
Trading signals don't last forever. As soon as your idea is discovered by others, capital floods in and the edge erodes. A few signals last 20 years; most don't. Treating your edge as permanent is the surest way to blow up.
The framework forces a continuous research budget. You should always be hunting for the next signal because the current one is decaying. The market is reflexive: by trading on a finding, you and others change the very behaviour that produced it.
Volume capacity matters too — short-term signals can only absorb so much money before their own success kills them.
- Markets change as you interact with them.
- Once a signal is widely known, the price action that produced it disappears.
- Short-term signals have a volume ceiling.
- A 20-year-old signal that still works is the exception, not the rule.
- The cost of an edge is the ongoing research to find its replacement.
- Catalogue every active signalWrite down each rule, signal, or pattern you trade on. Include when it was discovered and what conditions it relies on.Pro tipIf you can't articulate a signal cleanly, you don't actually have one — you have a vibe.
- Estimate its capacityAsk how much capital can flow into the signal before its own footprint moves prices. Short-term, mean-reverting signals usually have low capacity.WarningIf your AUM grows past the signal's capacity, the edge will silently disappear.
- Track decay metricsMonitor the live performance of each signal versus its backtest. Sustained underperformance is evidence of decay, not bad luck.Pro tipSet a tripwire — e.g. live Sharpe falls below half of backtest for N months — that triggers a review.
- Maintain a research pipelineAlways have new candidates being tested. The point is to have a replacement ready before the current signal dies.
- Retire signals without sentimentWhen evidence shows a signal has decayed, kill it. Defending an old signal because it used to work is the same bias as results-driven research.
Boyle says he has a few signals that have worked for 20 years, but calls this rare. Most of what he's built has had a finite life — useful for a stretch, then competed away as others independently found it.
He uses the example: if you find a way of finding cheap stocks and start buying them, and others do too, they're no longer cheap. The signal kills itself through use.
Boyle credits the insight to a Plain Bagel conversation about edge erosion. After 20 years in markets, he's seen most signals fade as competitors discover them. A handful of his own signals are 20 years old and still work — but he calls that rare. The lesson is to hold any individual edge with a light grip and keep the research engine running.