Do candlestick and level readings outperform indicators when both are measured the same way?
Price Action or Indicators? 3,264 Backtests
Finding
Neither family leads. Traded on the same bars with the same exit and cost, the count profitable in both years is 10 of 16 trend readings, 7 of 22 price action readings and 6 of 24 oscillators on the hourly chart; on the four-hour chart price action reaches 8 of 22 and oscillators 8 of 24 while trend falls to 2 of 16. The leading family changes with the timeframe. The reading with the highest 2024 net finished 47th, 59th and 53rd of 68 in 2025, losing money on all three timeframes.
Key results
- Profitable in both years, hourly chart
- trend 10/16, price action 7/22
- Where 2024's best reading finished in 2025
- 47th, 59th and 53rd of 68
- Conditions that name no side
- 2 of 70
Scope
- USDJPY, 2024 and 2025, plus the two halves of 2025
- 68 readings x 3 timeframes x 4 windows x (3 holding horizons + a zero-spread repeat) = 3,264 runs
- 0.3 pip spread, 0.1 lots, exits at 5, 10 and 20 bars
Method
All 70 conditions in the tester's catalogue, each left at the values it ships with and traded on its own. The only exit is a fixed number of bars: no stop, no target, no opposite signal, because a third of the set never produces a second side to close against. Two of the 70 are filters that name no side and cannot open a position alone, so 68 were traded.
Limitations
- Two years of USD/JPY. Nothing here establishes future profit or that the result repeats on another pair.
- Each condition was measured once, at the values it ships with. This is not the result of tuning.
- Signals and fills both use the same bar's close. The spread is fixed; slippage and swap are excluded.