Triple Exponential Moving Average: The Best Moving Average I Have Ever Found
I tested every type of moving average I could find – simple, exponential, weighted, linear regression, alligator, crossovers – across thousands of breakout strategies. One moving average destroyed everything else in the ranking. The triple exponential moving average is the most robust, most reliable, and most improving MA filter I have ever encountered.
This is not opinion. This is what the data says after running 100 AI-generated filter conditions through BreakoutOS on NASDAQ 60-minute breakout strategies, using rigorous in-sample and out-of-sample testing.
Watch the full video below for the complete breakdown, then keep reading for the detailed rankings and how to implement the winning conditions.
How I Tested 100 Moving Average Conditions
If you followed my earlier research where I asked AI to generate 100 filter conditions, you know the setup. I used Gemini AI to create 100 unique filter conditions written in EasyLanguage, organized into six categories: moving averages, oscillators, price action, strength, volatility, and volume.
This video focuses specifically on the moving average category. Here is the testing framework:
- Generate baseline strategies – I prototyped thousands of breakout strategies on NASDAQ 60-minute charts using the Mr. Breakouts Formula without any filters applied.
- Apply each MA condition – Every moving average filter was layered on top of the baseline strategies and tested independently.
- Split the data – Each strategy was evaluated on an in-sample period and an out-of-sample period. If a filter only works in-sample, it is curve-fitted garbage.
- Measure uplift – BreakoutOS uses a proprietary “uplift index” that measures improvement across multiple metrics simultaneously: average trade, win percentage, net profit, and net profit-to-drawdown ratio.
- Rank by total score – I combined in-sample performance, out-of-sample performance, and a robustness index into a single total score for each condition.
This approach eliminates the biggest mistake most traders make: testing one indicator on one strategy and declaring it the best. When you test across thousands of strategies, the noise cancels out and the real edges reveal themselves.
The Moving Average Categories I Tested
The 100 conditions covered a wide variety of moving average types and applications. Here are the main groups within the MA category:
- Price above/below a single MA – The classic approach. Is price above or below a simple, exponential, or weighted moving average?
- MA slope – Measuring whether the moving average itself is rising or falling. Compare the current value to 1 bar ago.
- Fast MA above/below slow MA – Crossover logic. A shorter-period MA compared against a longer-period MA to define trend direction.
- Alligator joints – Multiple MAs with different periods and offsets, measuring convergence and divergence.
- Triple exponential MA – The distance between price and a triple-smoothed EMA, with a threshold for entry.
- Linear regression – Using a linear regression line as a directional filter.
- Weighted MA – The weighted moving average variant with heavier emphasis on recent prices.
Each type was tested with optimizable parameters – period lengths, thresholds, lookback windows – so the playing field was level. Every condition got the same chance to prove itself.
In-Sample vs. Out-of-Sample: Which Moving Averages Survive?
Here is where it gets interesting. Many conditions that look excellent in-sample completely fall apart out-of-sample. That is the difference between a real edge and an illusion.
The in-sample top performers were:
- Price above/below a single MA – strong in-sample uplift across most strategies
- EMA slope – consistent improvement in average trade and win percentage
- Fast MA above/below slow MA – crossover logic showed solid in-sample results
- Alligator joints – convergence/divergence patterns helped filter bad trades
But in-sample means nothing if it does not survive the out-of-sample test. When I looked at out-of-sample performance, the ranking shifted dramatically:
- Triple exponential moving average – surged to the top
- EMA slope – held its ground, confirming its strength
- Linear regression – showed genuine forward-looking improvement
Notice what happened: some of the in-sample winners disappeared entirely. The classic price-above-MA condition, which looks great historically, did not hold up nearly as well on unseen data. The conditions that survived are the ones that measure something deeper about market structure – not just position relative to an average, but momentum and smoothed trend direction.
The Complete Moving Average Ranking
Here is the full ranking combining all three dimensions – in-sample performance, out-of-sample performance, and robustness index – into a single total score:
| Rank | Moving Average Condition | In-Sample | Out-of-Sample | Robustness | Total Score |
|---|---|---|---|---|---|
| 1 | Triple Exponential MA (#77) | Strong | Best | Best | Highest |
| 2 | EMA Slope (#79) | Strong | 2nd | High | 2nd |
| 3 | Linear Regression | Moderate | 3rd | High | 3rd |
| 4 | Weighted MA | Moderate | Moderate | High | 4th |
| 5 | Price Above/Below Single MA | Best | Moderate | Moderate | 5th |
| 6 | Fast MA / Slow MA Crossover | Strong | Weak | Low | 6th |
| 7 | Alligator Joints | Strong | Weak | Low | 7th |
The triple exponential moving average did not just win. It dominated. It scored highest in out-of-sample performance and highest in robustness, meaning the improvement you see historically is the most likely to carry forward into live trading. That combination is extremely rare.
The EMA slope came in second – and if you have followed my previous research, this should not surprise you. I have already shown in multiple videos that the EMA slope works across NASDAQ, Gold, Nikkei, and crypto. Seeing it confirm again on this independent test adds even more weight to that finding.
How the Triple Exponential Moving Average Works
The triple exponential moving average (TEMA) applies exponential smoothing three times. This removes more noise than a standard EMA while still responding relatively quickly to price changes. The result is a smoother line that tracks the underlying trend with less lag than you might expect from triple smoothing.
Here is the exact condition that won the ranking:
- Calculate the TEMA – Apply triple exponential smoothing with a period of 25 bars on the 60-minute chart.
- Measure the distance – Calculate the difference between the current close price and the TEMA(25) value.
- Normalize as a percentage – This is critical for professional use. Express the distance as a percentage of price so the condition works regardless of the instrument’s price level.
- Set a threshold – If the normalized distance is above a positive threshold, allow long entries. If below a negative threshold, allow short entries.
The beauty of this approach is that it captures both trend direction and trend strength in a single measurement. When price is far above the triple EMA, the market has strong upward momentum. When price is close to or below the triple EMA, momentum is fading or reversing.
This is fundamentally different from a simple “price above MA” filter. The distance component adds a quality dimension – it is not enough for price to be above the average. It needs to be meaningfully above it, which filters out the choppy, indecisive periods where breakouts tend to fail.
How the EMA Slope Condition Works
The second-place finisher deserves its own explanation because it is equally practical and has been validated across more markets than any other condition I have tested.
The EMA slope condition is simple:
- Calculate an EMA over a given period. The best-performing periods in this test were between 20 and 50 bars.
- Compare the current EMA value to the value 1 bar ago. If the current value is higher, the slope is positive (rising). If lower, the slope is negative (falling).
- Normalize as a percentage to make the condition work across different price scales.
- For long trades, require a rising slope. For short trades, require a falling slope.
What the slope measures is acceleration, not just direction. A market can be in an uptrend but decelerating – the moving average is still above yesterday’s value but by a shrinking amount. The slope filter catches this and keeps you out of trades where momentum is fading.
This is why it keeps winning across different markets and timeframes. Momentum is a universal property. Whether you are trading NASDAQ or Gold, a breakout backed by accelerating momentum has a higher probability of following through than one without it. The EMA slope measures exactly that.
What Both Winners Improve
The most important finding is not just that these conditions rank highest – it is what they improve. Both the triple exponential moving average and the EMA slope delivered gains across every major metric, in both in-sample and out-of-sample testing:
- Average trade size increased – Each trade captured more profit on average, meaning fewer wasted entries.
- Win percentage improved – More trades ended profitably, reducing the psychological burden of long losing streaks.
- Net profit increased – Overall returns went up, not down, despite filtering out a portion of trades.
- Maximum drawdown decreased – The equity curve became smoother with shallower pullbacks.
- Net profit-to-drawdown ratio improved – The risk-adjusted return – the metric that matters most for real-world trading – got better.
This is rare. Most filters are a trade-off. They improve one metric at the expense of another. A tight filter might boost win rate but kill net profit by removing too many trades. These two conditions improve everything simultaneously, which is why they score so high on the uplift index.
If you are building breakout trading strategies and not using either of these as a filter, you are leaving significant edge on the table.
How to Apply This to Your Trading
Based on the results, here is what I recommend:
- Start with the triple exponential moving average. Use a TEMA period of 25 on a 60-minute chart. Measure the distance between close and TEMA as a percentage. Optimize the threshold to find the level that filters choppy conditions without removing too many valid trades.
- Add or test EMA slope as a second filter. Use an EMA period between 20 and 50. Compare the current value to 1 bar back. Set a minimum slope threshold expressed as a percentage.
- Always validate out-of-sample. If a filter only improves your in-sample results, it is not a real edge. Split your data and make sure the improvement carries forward.
- Do not stack too many filters. One or two well-tested filters will outperform a chain of five mediocre ones. Complexity is not your friend – simple filters that measure real market behavior are what actually work.
- Test across multiple strategies, not just one. A filter that only helps one specific strategy is probably curve-fitted to that setup. Use the uplift approach across a large sample to confirm the edge is real.
The triple exponential moving average is now at the top of my filter toolkit. It earned that spot the only way that matters – through rigorous, large-scale testing with proper out-of-sample validation. The data does not lie.
Ready to test breakout strategies with filters like the triple EMA? BreakoutOS lets you prototype filter combinations visually, then validate them with the Breakout Filters Robustness Tester – all without writing code. See the tools at breakoutos.com