Last updated: April 14, 2026

I Tested Moving Average Filters on 1,000 Strategies Across 1,000 Markets - Here's What the Data Shows

Moving averages are the most widely used indicators in trading - but almost nobody tests them properly. Most traders either optimise for one strategy on one dataset and call it confirmed, or dismiss them entirely as "too simple." Neither is the right answer. The real question is: when you test SMA and EMA filters across thousands of different strategies and thousands of different market conditions, do they actually hold up? That is what this study sets out to answer.

How BreakoutOS Tests Filters at Scale

The standard approach to indicator testing is flawed. You take one strategy, add an indicator, run an optimisation, and if the backtest improves - you declare it confirmed. The problem is that result is almost certainly overfitted to your specific strategy and your specific historical data. It tells you nothing about whether the indicator is genuinely useful.

BreakoutOS takes a different approach. The starting point is a single base strategy - in this case, a NASDAQ 60-minute long strategy with no filter. The entry rule is straightforward: previous day's low multiplied by a 0.8 factor of a 40-period average true range, with an end-of-day exit. Simple and clean.

From that base strategy, BreakoutOS extracts the strategy's DNA - its structural characteristics, parameter ranges, and market behaviour - and automatically generates 1,000 sibling strategies. These are structurally similar but distinct strategies, not copies. Then it runs every filter condition across all 1,000 strategies on 1,000 different market conditions.

The result: approximately 1,000,000 exhaustive iterations - completed in under one minute.

This is what separates a robust finding from an overfitted one. If a filter consistently improves performance across 1,000 different strategy variants and 1,000 different market regimes, you can trust that result. If it only looks good on one specific strategy at one specific parameter value, you are looking at noise.

The 8 Filter Conditions Tested

The study tested 8 distinct moving average filter conditions, covering the most common ways traders apply SMA and EMA to their strategies:

• SMA close-above (long): Only take long entries when close is above the SMA. Periods tested: 10, 20, 30 ... 200 (step 10, 20 variants).
• SMA close-below (long): Only take long entries when close is below the SMA. Same period range.
• EMA close-above (multiplier 2): Close above EMA. Periods tested: 2, 4, 6 ... 40 (step 2).
• EMA close-below (multiplier 2): Close below EMA. Same period range.
• EMA close-above (multiplier 5): Close above EMA. Periods tested: 5, 10, 15 ... 100 (step 5).
• EMA close-below (multiplier 5): Close below EMA. Same period range.
• Dual EMA crossover - fast above slow: Fast EMA (periods 2-40, step 2) above slow EMA (periods 5-100, step 5) at time of entry.
• Dual EMA crossover - fast below slow: Fast EMA below slow EMA at time of entry.

For each condition and each parameter combination, the test measured three things across all 1,000 strategies:

1. Net profit/drawdown ratio improvement - the primary measure of risk-adjusted performance
2. Average trade improvement - whether the per-trade quality increased
3. Bounce index - how many originally-failing strategies were rescued by adding the filter

SMA Results: Simple Is Robust

Simple moving average turned out to be a genuinely strong performer - and the reason is not just the improvement size but the consistency across parameters.

For the best-performing SMA condition, 540 out of 1,000 strategies improved their net profit/drawdown ratio. That is just over half. The overall average improvement to the net profit/drawdown ratio was 0.3% - modest on that metric. But the average trade improvement was 48%. That is a substantial number: strategies that previously had too small an average trade to be tradeable in real conditions were made viable just by adding a simple moving average filter.

The bounce index was also 48% - meaning nearly half of the originally-failing strategies among the 1,000 siblings were recovered by adding the SMA filter. That is a meaningful rescue rate.

The most striking finding, though, was about period selection. Across the SMA tests, it barely mattered which period you used. Whether you chose SMA 20, SMA 50, SMA 100, or SMA 200 - on average, across 1,000 different strategies and 1,000 different market conditions, average trade improved by approximately 50%.

When a result holds regardless of the parameter you choose, that is the definition of robust. SMA earned an 83% viability rating in this study - highly recommended as a filter for NASDAQ long breakout strategies. It requires only one optimisable input, which limits the scope for overfitting, and the improvement is consistent enough that you are not going to accidentally pick a bad period.

EMA Close-Above/Below: The Weakest Performer

The simple EMA conditions - close above or below a single EMA at the moment of entry - were the weakest performers in the study.

Viability was around 50%. That means you are essentially at a coin flip. Some parameter values improved results, some did not, and there was no consistent pattern across the 1,000 strategies and 1,000 market conditions.

The specific case that performed worst was the long NASDAQ strategy requiring close to be below EMA at entry. That condition showed no meaningful improvement and is not a robust filter.

The issue is not that EMA is a bad indicator - it is that using it as a simple price-above/price-below filter on a single line does not produce reliable edge when tested at scale. It can look convincing on one strategy with one optimised parameter. Tested across 1,000 variants and 1,000 market regimes, the edge disappears. This is the hallmark of an overfitted result.

The Winner: Dual EMA Crossover

The top-performing condition in the study was one Tomas did not fully expect to win: the dual EMA crossover, specifically the condition where the fast EMA is below the slow EMA at the time of entry.

For a long NASDAQ breakout strategy, entering when the fast EMA is below the slow EMA - a counter-trend regime condition - consistently outperformed all other moving average filter types across the 1,000 strategy tests.

The parameter analysis showed a clear sweet spot: fast EMA in the range of 10-20, slow EMA in the range of 70-90. That gives a slow-to-fast ratio of approximately 3:1 to 4:1. Across the optimisation map - plotting all combinations of fast period versus slow period - the improvement was broad and stable, not a narrow peak that would suggest overfitting.

What this means practically: for a NASDAQ long strategy, better entries tend to come when the short-term trend (fast EMA) is pointing below the longer-term trend (slow EMA) - a slight counter-trend setup at entry, even within an overall long trade. This is consistent with how breakout strategies work: the best breakouts often come from periods of compression or slight pullback, not from runaway trends.

As with the SMA results, the dual EMA crossover condition is confirmed on 1,000 different strategies and 1,000 different market conditions. That is the important thing - not that it looks good in this one backtest, but that it holds up at scale.

What These Results Mean for Your Strategy Building

There are three practical takeaways from this study.

First: SMA is underrated. A lot of traders dismiss simple moving average as too basic. The data disagrees. For NASDAQ long breakout strategies, SMA is a genuinely robust filter - 83% viability, consistent improvement across any period you choose, and a meaningful rescue rate for strategies that were not quite tradeable on their own. If you are not including SMA as a candidate filter in your strategy development process, you are leaving edge on the table.

Second: single EMA conditions are not reliable at scale. They can look good on one strategy with one parameter. They do not hold up across 1,000 variations. If your strategy relies on a close-above/below EMA filter that you optimised specifically for that strategy, treat that result with caution. It is likely overfitted.

Third: test everything at scale, not on one example. This is the core principle behind BreakoutOS's approach to filter testing. It does not matter if your indicator looks good on your strategy - what matters is whether it looks good across thousands of strategy variants and thousands of market conditions. If it does, you have something real. If it only works on your specific combination, you are trading on noise.

The next study in this series will compare CCI and RSI using the same methodology. Both are popular oscillator-based filters. The results are, according to Tomas, surprising.

Frequently Asked Questions

About the Author

Tomas Nesnidal is a breakout trading specialist, hedge fund co-founder, and creator of BreakoutOS. He has managed institutional portfolios using breakout strategies for over 15 years, trading from 65+ countries. He is the author of The Breakout Trading Revolution and co-founder of Breakout Trading Academy.