Forex filter rule

Checked October 2026

A filter rule buys after price rises a set percent off a low, and sells after it falls that percent off a high. Levich and Thomas included it in the same 1993 currency study. Backtoast ran the 1% filter. This version failed our test, losing 3.9% a year over 395 trades.

Backtoast

You pay $99 for two full tests of your own strategy, and a pass is an offer of 30% of the profit Backtoast makes trading it with its own money.

Forex filter rule

It's always in the market. A 1% rise off the low flips it long and a 1% drop off the high flips it short, so a small filter catches the start of a move and every false start too.

The published rule measures from the last peak or trough since its own last signal. The run couldn't do that, so it used the high and the low of the last 20 days.

Donchian 4 week rule is the futures cousin, flipping on a new four-week high or low.

When it works

A filter rule is trend following, so it pays when a currency moves one way for weeks on a central-bank rate cycle and gives back in range-bound markets, where every 1% swing flips it. It made +0.12% a trade in the last six months of our run, against a loss before that. The study tested filters from 0.5% to 5%, and a wider one flips less often, which a draft can test before you spend a full test.

Origins

Levich and Thomas found every filter size made money on every currency they tested from 1976 to 1990, before costs, with the 1% filter at 3.4% to 9.3% a year. Sweeney had shown filter profits in currencies in 1986. The run traded the same five dollar pairs as the moving average crossover, from late 2021 to October 2026.

The run is Backtoast's, from 29 October 2026, after costs.