Monitoring for Strategy Decay

advanced7 min read

Edges erode as others find them. The signs your strategy is dying, and when to retire it.

This lesson confronts the truth from Module 1 head-on: *every edge eventually decays.* Markets change, edges get crowded, and inefficiencies get arbitraged away. Monitoring for decay — and knowing when to retire a strategy — is the difference between exiting gracefully and riding a dead strategy into the ground.

The hardest skill in systematic trading is distinguishing temporary underperformance (which you should ride through) from genuine, permanent decay (which means retire now) — and getting this wrong in either direction is costly. Quit too early on normal variance and you abandon a still-good edge at its worst moment; hold too long on a truly dead strategy and you bleed capital chasing an edge that’s gone. The signals of real decay (not just bad luck): live results persistently diverging below the forward-test expectation (beyond the Monte Carlo range), the character of returns changing (win rate and payoff structurally shifting, not just a slump), evidence the edge is now crowded (everyone’s running it), and — most importantly — the mechanism you identified (Module 1) no longer holding (the behavioural bias or structural quirk has changed). The defences: pre-define your retirement criteria before going live (so the decision is rule-based, not emotional), diversify across multiple edges (so any one’s death isn’t fatal), and accept that retiring a strategy is success, not failure — it means you caught the decay instead of being destroyed by it. The goal was never a strategy that lasts forever; it’s a process that finds, exploits, and gracefully retires edges.

Decay is inevitable — markets change, edges crowd and get arbitraged (Module 1); plan for it.
The hard call — distinguish temporary underperformance (ride it) from permanent decay (retire); both errors cost.
Real-decay signals — persistent divergence below expectation, structural change in win rate/payoff, crowding, and the mechanism no longer holding.
The defences — pre-defined retirement rules (not emotion), diversification across edges, and treating retirement as success.

ExampleA strategy’s edge rested on a behavioural quirk. Over a year, live returns drift persistently below the Monte Carlo range, the win rate structurally falls, and the trade is now in every newsletter (crowded). These aren’t a slump — they’re decay signals plus a broken mechanism. Your pre-set rule (“retire if 12-month live Sharpe falls below X”) triggers, and you exit cleanly — having caught the decay rather than donating capital to it.

Key takeawayEvery edge decays (markets change, edges crowd). The skill is telling temporary underperformance (ride it) from permanent decay (retire) — watch for persistent below-expectation results, structural win-rate/payoff change, crowding, and the mechanism breaking. Pre-define retirement rules, diversify across edges, and treat retiring as success.

FAQs

When exactly should I retire a strategy?

Ideally per *criteria you set before going live* — e.g. live performance staying outside the expected (Monte Carlo) range for a sustained period, a structural change in return character, or the original edge mechanism clearly no longer applying. Pre-committing removes emotion from a decision that’s otherwise dominated by hope (holding too long) or fear (quitting too early). Diversifying across edges makes any single retirement survivable.