How we test honestly
It is easy to make any method look good on past charts. These rules stop us from fooling ourselves.
Two words first
| Word | Simple meaning | Dollar example |
|---|---|---|
| R | The money risked on one trade, from entry to stop. | Our rule bots risk 0.5% of $10,000, so 1R = $50. A trade at +2R made $100. A trade at −1R lost $50. |
| Edge | The average result per trade over many trades. | +0.1R = +$5 a trade on average. That is our minimum before we pay for a challenge. |
A low win rate can still be fine. With a 2R target, winning 1 trade in 3 is break-even.
Our five rules
- Pick on 2025, judge on 2026. We choose the rules on 2025 prices only, then test them on 2026, which they have never "seen". Only the 2026 number counts.
- Strict accounting. We pay the spread on every trade. If one candle touches both the stop and the target, we count a loss.
- Testing many ideas is a trap. If you try 100 versions, one looks good by luck. So we also look at the average of all versions, not only the best one.
- 30+ trades before judging. Under 30 trades, the result is mostly luck. Losing 3 in a row is normal, even for a good method.
- Don't change rules after a few bad trades. That is how you fool yourself.
Quick check