Strategy Decay: Why Edges Fade

[Fairman]

Strategy Decay: Why Edges Fade

Strategies that work for a while often stop working. This is not always a sign of error. Markets adapt.

Several forces drive decay. Competition is one: as more traders learn about a pattern, the opportunity shrinks, since their trades move price toward fair value. Changing conditions are another: volatility levels, interest rate regimes, and market structure shift over time. And some edges come from temporary imbalances, which resolve.

Decay can be gradual or abrupt. A trend-following system built for the strong trends of one decade may struggle in a choppy one. A news-based approach may lose value as execution speed improves among institutions.

How do you detect it? Monitor live performance against expectations. Track rolling expectancy over the last 50 or 100 trades. Compare the current drawdown with the historical maximum. If results fall outside the range seen in testing, investigate.

The difficulty is that normal variance and genuine decay look similar at first. Predefine a threshold, for example, a drawdown 1.5 times the maximum backtest, or a rolling expectancy below zero for a set number of trades, that triggers a formal review.

The response may be to reduce size, adapt the rules, or retire the strategy.

Having more than one approach, and the willingness to research new ones, offers resilience.

Practical step: calculate the rolling expectancy of your last 50 trades, and write the level at which you will pause and review.

[Fairman]

Tracking rolling expectancy is easy in a spreadsheet. One column of R results, then an average of the last 50 rows dragged down. Plot it and you'll see drift much earlier than you'd notice from the account balance.

[Fairman]

Something that looks like decay is sometimes just you. Before blaming the market, check if your rule adherence slipped. Bad patches line up with broken rules more often than people expect.

[LondonNewsTrader]

Changing conditions are another: volatility levels, interest rate regimes, and market structure shift over time.

The rate regime point is a big one for anyone using older test data. Through most of the 2010s, policy rates in Europe and Japan were pinned near zero, and many data releases barely moved the euro or the yen because nobody expected the central bank to react. Once inflation came back in 2021 and 2022, the same releases started producing large moves. A strategy tested mostly on the quiet years can look decayed today when really it's facing a different market.

So before calling it decay, I'd check whether the test period and the live period share the same rate backdrop.

[Shadow Trader]

The point about variance and decay looking alike can be put in numbers. Say a system wins 45 percent of the time at 1.6R and loses 1R otherwise, which gives an edge of 0.17R per trade.

For a rolling 50 to dip below zero, that system needs 19 wins or fewer out of 50. With nothing wrong at all, that happens roughly 1 time in 5, so a negative rolling line on its own is weak evidence.

The chart below shows 400 made up trades from that system, with the rolling 50 average underneath. It finishes up 70.6R against an expected 68R, yet the rolling line still drops below zero in three separate stretches, as low as minus 0.32R.

img_13006_r1.png (105.46 KiB) Viewed 21 times

That is why a threshold on the rolling line alone can trigger reviews too often. Pairing it with the rate regime check above, or with a count of rule breaks in the same window, helps tell a bad patch from a changed market.

A practical tweak is to test the threshold on simulated runs before using it live. If a healthy system would trip it several times over a few hundred trades, the threshold is measuring noise rather than decay.

[FXS Prop Desk]

Recognising decay early is the hard part. A strategy in a normal drawdown and one that has stopped working look the same for a while.

Comparing current results with the historical distribution helps. If the current drawdown is deeper or longer than anything in the backtest or the live record, decay becomes more likely, and reducing size or pausing makes sense.