When to retire a trading strategy

Drawdown vs death look the same

Every strategy has drawdowns. Some are normal turbulence within historical expectations; some are signals the underlying edge has decayed. The mistake retail traders make is treating any drawdown over X% as a kill signal, when X% might be perfectly normal for the strategy. The opposite mistake is treating clearly-anomalous drawdowns as “just turbulence” until the account is gone. Both are emotional decisions dressed up as analysis.

Statistical underperformance thresholds

A strategy's expected drawdown distribution comes from Monte Carlo or walk-forward analysis. If your live drawdown exceeds the historical P95 worst-case, that's a statistical signal — not proof, but a meaningful event. If it exceeds P99 or significantly crosses the historical max DD, the edge is probably gone.

Run a Monte Carlo simulation up front, before going live, so you have the P75/P95/P99 thresholds in hand — not improvised during the drawdown. Use the Monte Carlo simulator to generate these. See also risk of ruin and statistical significance.

Regime change vs strategy death

Some strategies don't die — they just hit their bad regime. A trend-following strategy in a long range looks dead. A mean-reversion strategy in a strong trend looks dead. The question: when this regime ends, will the strategy work again?

Walk-forward analysis across historical regimes answers this. If the strategy has historically recovered from similar regimes, this is likely temporary. If the current drawdown is unprecedented even across diverse historical regimes, the edge is probably gone. See strategy types and regime dependence.

Walk-forward predicts retirement

Walk-forward analysis isn't just for development. Running it continuously on a live strategy gives the earliest warning available. If the most recent walk-forward windows show consistent OOS degradation while older windows showed strong OOS performance, the strategy is decaying — not just in a drawdown but in its underlying edge. This pattern is much more informative than recent live P&L alone. See walk-forward vs Monte Carlo.

Honest retirement decision process

1. Define exit criteria before going live. P95 DD threshold, time underwater, consecutive losses. Written down. Not negotiable in the heat of the moment.
2. Track live performance against the historical Monte Carlo distribution. Not against your hopes.
3. When threshold hit, pause — don't delete. Pause trading and run diagnostic.
4. Diagnostic includes: walk-forward refresh on latest data, regime analysis, parameter stability check.
5. If diagnostic confirms degradation, retire. If it's just regime, wait or sit out.

Replacement, not extension

The honest move when a strategy retires is to replace it, not to retune it. Retuning a dying strategy is overfitting in disguise — you're optimizing to the noise of recent underperformance, the most fragile data you have. New strategies built from scratch and validated on full historical data tend to outperform “rescued” versions of dying ones. See choosing parameters without overfitting for why retuned strategies usually fail.

The Puravida Edge methodology applies this honestly: portfolio strategies that show consistent walk-forward degradation get pulled from the production roster, not retuned. Replacement, not extension.

FAQ

When should I stop trading a strategy?

When live drawdown exceeds the historical P95–P99 of expected drawdowns from Monte Carlo simulation, that's a strong statistical signal of edge degradation. The exact threshold depends on your risk tolerance, but objective criteria beat emotional ones. Set the exit threshold before going live, not during a drawdown.

How long should I let a strategy underperform before quitting?

Time alone is a poor metric — it's about whether the underperformance is statistically anomalous. A strategy that's been flat for 6 months might be in its bad regime (fine, wait) or might have lost its edge (retire). Walk-forward analysis on the most recent live data is what distinguishes the two cases.

Can a dead strategy come back?

Rarely. Most strategies that show consistent walk-forward degradation across multiple recent windows don't recover, because the market dynamic that produced the edge has changed. The exception: regime-specific strategies that look dead in the wrong regime but recover when conditions return. Walk-forward analysis distinguishes these two cases.