Backtest evidence

Should a crossover use the simple or the exponential moving average?

SMA vs EMA: Which Moving Average Is More Profitable? 672 Tests

Finding

Neither came out ahead. Across 672 head-to-heads on USDJPY — 56 period pairs, three timeframes, four windows, identical settings on both sides — the EMA finished ahead in 318 (47.32%), and split into twelve timeframe-and-window cells it won between 15 and 39 of 56 in each, all near the half-way line. The difference is not small, though: swapping the method at fixed periods moved net pips by a median of 491.1, and against a control that moves the period by one bar the swap moved the result further in 593 of 780 rungs (76.03%). Which way it moves is not predictable — the same method won both 2024 and 2025 for 19 of 56 period pairs on 15-minute bars (33.93%), below a coin flip. Their centres of gravity are identical, 9.5 bars back at period 20. The EMA weights the newest bar 1.905 times as heavily and took fewer trades in 405 of 672 head-to-heads.

Key results

Head-to-heads where the EMA finished ahead
318 of 672 (47.32%)
Net pips moved by swapping the method
median 491.1; further than a one-bar period change in 593 of 780 pairs
Same method winning both years
19 of 56 on 15-minute, 26 of 56 hourly, 25 of 56 four-hour

Scope

  • USDJPY, 2025-01-01 to 2025-12-31, with 2024 and the two halves of 2025
  • 15-minute, hourly and four-hour bars; 24,903 15-minute bars in 2025, 6,226 hourly and 1,610 four-hour
  • The 56 pairs of 8 fast and 8 slow periods where fast is shorter, compared one-to-one under both methods: 3 timeframes x 4 windows = 672 head-to-heads, plus 780 control pairs
  • 0.3 pip spread, no slippage, 0.1 lots, charged on both fills

Method

Measured with Formiq's backtester. A long fills at the close where the fast average crosses above the slow one and a short where it crosses below, reversing on the opposite cross. Each period pair is run under both methods and compared one-to-one; as a control, a continuous ladder that moves the period one bar at a time is run under both methods, 780 pairs in all.

Limitations

  • One instrument, USDJPY, and two years. Nothing here promises the same on another pair or another year.
  • Only simple and exponential were compared. Weighted, smoothed and Hull are in the crossover article.
  • The weights, the centre of gravity and the step formulas follow from arithmetic and hold on any data. The distances, crossing counts and results are USDJPY over two years.
  • Entries and exits are taken at the bar close, and the spread is held at 0.3 pips throughout.
  • The seeding measurement is read at one moment, 1 January 2025. Another date gives a different size of error, though the decay rate is date-independent.