Backtest evidence

Is the textbook 5% disparity reachable, and does the reading return to zero because price comes back?

MA Disparity Mean Reversion: 4,320 Settings Tested

Finding

It is not reachable, and the average closes the gap rather than price. Across 2024 and 2025 on 15-minute, hourly and four-hour USDJPY bars, not one bar printed a 25-period disparity of 5% or more; the largest reading in two years was 1.94% on 15-minute bars, 2.78% hourly and 4.54% on four-hour. A 25-period disparity kept nine bars in ten inside 0.27% on 15-minute bars, 0.56% hourly and 1.05% on four-hour. Splitting each round trip back to zero into the part price contributed and the part the average contributed, the price share was negative in 13 of 18 cells and the average's share ran from 61.6% to 139.6% — price kept moving away while the average walked up to it. The zero cross is a close crossing the moving average and nothing else: over three timeframes and five periods there were zero sign disagreements, and from period 5 to 200 across all five averaging types the trade counts and pips were identical.

Key results

Bars where a 25-period disparity reached 5%
none in two years on any timeframe; maxima 1.94% / 2.78% / 4.54%
Range holding nine bars in ten at period 25
0.27% on M15, 0.56% on H1, 1.05% on H4
Where the return to zero came from
price contributed negatively in 13 of 18 cells; the average contributed 61.6% to 139.6%

Scope

  • USDJPY. 2024, 2025, the first half of 2025 and the second half
  • 15-minute, hourly and four-hour bars
  • 5 periods (5, 10, 25, 75, 200) x levels (5 percentage values, 5 ATR multiples) x exit levels (3 percentage, 3 ATR) x 5 averaging types = 4,320 runs, with the averaging comparison as a separate 360
  • 0.3 pip spread, 0.1 lots, filled at the close

Method

Measured with Formiq's backtester. Fills use the close; no stop or target in the main sweep. A shuffled close-only series serves as a control, and the disparity zero cross is compared directly against a close-versus-average crossover.

Limitations

  • One pair. The unit is a fraction of price, so the 0.27% / 0.56% / 1.05% figures are USDJPY's for these two years and do not transfer to a pair trading at a different level.
  • The shuffled control is a close-only series. It discards each bar's high and low, so it cannot be used for rules that read wicks.
  • Fills use the close with zero slippage. A setting that trades 2,700 times a year on 15-minute bars would execute worse than this in practice.
  • The 5% figure comes from daily equity charts and does not transfer to short foreign-exchange timeframes.