Equity curve trading (equity curve filter)
Equity curve trading switches a strategy on or off, or changes its size, according to the state of the strategy's own equity curve, for example trading only while the curve is above its moving average. It can shorten some drawdowns, but it also skips the trades that start recoveries, and it adds a parameter that must be tested out of sample.
Senzoukria · Glossary · Updated September 2026
How the filter works
Record the results the strategy would have produced on every signal, traded or not; this is the shadow equity curve. Compute a moving average of it, for example over the last 20 trades. Take real trades only while the shadow curve is above its average, and trade in simulation otherwise. Variants reduce size instead of stopping, or pause after a drawdown of a set depth.
The idea rests on one assumption: that a strategy's results are autocorrelated, so that bad periods tend to continue. If results are independent, the filter removes good and bad trades alike and only reduces the sample.
What it can and cannot do
- It can cut exposure during long losing regimes if losses really cluster.
- It always skips the first trades of a recovery, because the curve must climb back above its average before trading resumes.
- It cannot create an edge: a strategy with zero expectancy filtered by its own curve still has zero expectancy.
- It adds parameters, the average length and the rule, and those need their own out-of-sample test.
- It requires tracking every signal, including those not traded, which is easy to get wrong manually.
Testing it honestly
Compare the filtered and unfiltered versions on data not used to choose the filter settings, and look at more than the maximum drawdown: net result, number of trades and the time to recover. A filter that improves drawdown on one period and costs most of the net result on another is not an improvement. Checking the autocorrelation of trade results before building the filter shows whether there is anything for it to exploit.
In Senzoukria
The desktop has no built-in equity curve filter. In the automatic backtest the strategy's decide function receives recent bars, the position it holds and a state object it controls, but not its realized profit and loss; reproducing a filter means recomputing hypothetical trade results inside the script's own state, which the engine does not do for you. The Performance panel's equity and underwater charts, and its rolling expectancy over 30 trades, show whether a strategy's bad periods cluster enough to justify testing a filter.
Related
In the same section
This page in other languages
Frequently asked questions
- Does equity curve trading reduce drawdowns?
- Sometimes, when losses cluster in regimes. It also delays re-entry after a drawdown, so it can lower total profit and occasionally deepen drawdowns when the curve whipsaws around its average. The only way to know is an out-of-sample comparison.
- What moving average length should the filter use?
- There is no standard value. A short average reacts quickly and whipsaws; a long one reacts late. The length is a parameter and should be chosen on training data and judged on data it has not seen.