Backtest evidence

Did RSI mean reversion or an RSI midline trend rule hold up better on USD/JPY?

Best RSI Settings, Tested 330 Ways on USD/JPY

Finding

On hourly USD/JPY, the long fade averaged a 65.3% win rate but was profitable in both 2024 and 2025 in only 31 of 150 settings. The midline cross averaged 23.7% and was profitable in both years in 17 of 30 settings. The standard period-14 30/70 long fade lost 250.7 pips in 2025.

Key results

Hourly long fade
Mean win rate 65.3%; 31 of 150 settings profitable in both years
Hourly short fade
Mean win rate 61.2%; 9 of 150 settings profitable in both years
Hourly midline cross
Mean win rate 23.7%; 17 of 30 settings profitable in both years
Standard RSI 14, 30/70 long fade
25 trades, 60.00% win rate and -250.7 pips in 2025; 22 trades, 63.64% and +26.1 pips in 2024

Scope

  • Instrument: USD/JPY
  • Periods: 2024-01-01 to 2024-12-31 and 2025-01-01 to 2025-12-31
  • Timeframes: M15, H1 and H4
  • Parameter grid: 330 settings per timeframe and year

Method

Formiq backtests at bar close with a fixed 0.3-pip spread and no slippage. The test compares 150 long fades, 150 short fades and 30 midline crosses on M15, H1 and H4, run separately over 2024 and 2025. Article claims are checked by research/rsi-verify.run.ts against the generated sweep output.

Limitations

  • The test covers one currency pair and two calendar years; it does not establish a universal RSI setting.
  • Entries and exits use bar closes with a constant spread and zero slippage, so live fills will differ.
  • The tested grid is finite and many fade settings have small trade counts.