Does a Moving Average Filter Improve a Breakout Rule
Adding a moving average to a breakout rule is one of the first things most people try. The reasoning is intuitive. Trade in the direction of the trend, and the trend is whatever side of the average price happens to be on. The rule becomes selective rather than mechanical, and selective feels more skilled. What the filter actually does is narrower and easier to examine than the reasoning suggests.
The Mechanism Is Subtraction
A direction filter does exactly one thing. It deletes some of the signals the base rule would have taken. It never creates a trade, never improves an entry price, and never changes how a trade that passes the filter behaves afterwards. Every difference in results comes from the subset that was removed.
That makes the evaluation simple in principle. Look at the trades the filter would have blocked and ask whether they were, as a group, worse than the ones it let through. If they were roughly the same, the filter has cut activity without cutting anything else, and a smaller sample of the same distribution is not an improvement. It is the same strategy at lower frequency, with more variance in the outcome over any given stretch.
Which Average, and When Was It Chosen
The period is the part that quietly does the work. A short average sits close to price and will often be on the same side as the breakout by construction, which means it blocks very little and the filter is close to decorative. A long average moves slowly, holds its side through several sessions, and will block a meaningful share of counter trend breaks, some of which would have been the best trades of the month.
The uncomfortable part is how the period usually gets picked. Someone tries several, keeps whichever produced the best looking history, and then describes the choice as trend following logic. The logic was fitted after the fact. A period chosen that way carries no information about the future, and its apparent edge is a description of the sample it was selected on.
What the Filter Duplicates
The opening range already contains directional information. Where price sat inside the range, whether the session opened away from the previous close, and which edge was tested repeatedly all speak to the same question the average is being asked. A filter that agrees with the range most of the time is adding a second copy of something you already had, and the two will disagree mainly in ambiguous conditions where neither is reliable.
This is worth checking directly rather than assuming. Count how often the filter and the range shape point the same way. If they almost always agree, the filter is not filtering. If they frequently disagree, the filter is overriding the structure the rule was built on, which is a much bigger change than it was probably intended to be.
What Would Count as Evidence
A filter that helps should show its help in the blocked group. Those trades should be visibly worse, not marginally worse, and they should be worse across different stretches of history rather than concentrated in one unusual period. If a single volatile month accounts for the entire benefit, the filter learned that month and nothing more.
It should also survive a change in the period without collapsing. If the filter helps with one setting and hurts with a neighbouring one, the setting is doing the work rather than the idea, and settings that precise do not survive contact with a new market condition. Robustness across nearby values is weak evidence, but it is real evidence, and precision at a single point is not.
Living Without It
The base rule without a filter takes more trades and takes some obviously poor ones. That is visible and irritating in a way that a missed winner is not, because the loss is on the statement and the missed trade is only a chart you looked at later. The asymmetry in how those two feel is a large part of why filters get added and rarely get removed.
None of this argues that a moving average is useless. It argues that the burden of proof sits on the addition, and that the proof has to come from the trades it blocked rather than from the appearance of the chart. A filter kept because it feels disciplined is a filter nobody has actually tested.