Triple EMA Strategy: Three-Average Rules and Testing
A three-EMA strategy uses fast, medium, and slow exponential moving averages of the same price series. The slower average can filter a faster crossover. This is different from TEMA, a single indicator that combines successive EMA calculations to reduce lag.
Specify the three averages
The 9, 21, and 55 periods are one possible research configuration, not an established optimum. Unless the script requests another interval, all three averages use the current chart timeframe. Three different lengths do not create multi-timeframe analysis.
TradingView's EMA documentation explains the greater weighting of recent observations. An EMA still depends on historical prices and can cross repeatedly in a range.
Example rule set to test
| Component | Explicit hypothesis |
|---|---|
| Long entry | Fast EMA crosses above medium EMA, with both above slow EMA |
| Short entry, if enabled | Fast EMA crosses below medium EMA, with both below slow EMA |
| Signal timing | Evaluate on a confirmed chart bar |
| Long signal exit | Fast EMA crosses below medium EMA |
| Short signal exit | Fast EMA crosses above medium EMA |
This table is a proposed rule set, not a validated result. Decide whether a reversal closes the current position only or also opens the other direction. A state such as fast above medium must not be confused with a new crossing event.
Add execution and risk rules
Choose a stop and target independently of the indicator names. If testing a stop 1.5 ATR from entry, define when ATR is sampled and whether the distance stays fixed or trails. For a long position, a protective stop is below the entry; for a short, it is above. A target of twice the initial stop distance is a planned reward-to-risk ratio, not the realized payoff ratio.
Position size depends on the selected loss budget, stop distance, instrument value, and permitted size increments. Gaps and slippage can exceed the intended loss. Define fees, spread assumptions, and order timing before comparing reports.
Test whether the third average adds value
Run the two-average and three-average versions on the same dataset with identical exits, costs, and size rules. Record trade count, net results, drawdown, average win/loss, and exposure. A filter can remove profitable trades as well as losing ones; fewer signals do not automatically mean higher quality.
Do not choose the best result from many periods and then call that same sample independent validation. Keep an untouched interval and inspect performance across differing conditions. A fixed number of demo trades is not proof of readiness for live execution.
Limits
The averages lag and can whipsaw. An additional ADX filter may change behavior, but a reading above 25 does not guarantee a profitable trend. Generated Pine code still requires syntax and chart-behavior checks.
The former AAPL 68% versus 51% win-rate comparison and SPY 64% win rate/11.8% drawdown lacked a reproducible script and trade list. Those results have been withdrawn; the page now describes a testable strategy structure rather than an alleged performance record.
