Searching 7,500 configurations for a prop-evaluation base, and what a high win rate actually costs

The goal was a configuration suited to a funded-account evaluation rather than to maximum profit: high win rate, a stable reward:risk, drawdown small enough to size contracts against, and consistency across the year. Roughly 7,500 configurations were swept across entry modes, filter stacks, bracket geometry, sessions, budgets, anchors and the reversal engine.

Scoring treats the account the way an evaluation does — a $100,000 account, a $6,000 target and a $3,000 drawdown limit — because contracts are sized to the limit, which makes drawdown the binding constraint rather than a tiebreaker.

Win rate is bought, one-for-one, with reward:risk

Across every bracket cell tested, the two moved in exact opposition:

The strongest edge in the entire sweep sits at a 42% win rate; the weakest sits at 80%. Every high-win-rate configuration is the same shape — a tight target against a wide stop — which wins often and loses rarely but enormously. That is not a strategy property, it is arithmetic: the target and stop set the hit rate, and the edge is roughly what it is either way.

The high-win-rate configurations all failed out of sample

Fifteen candidates were selected on the trailing 365 days and then scored on 2024-01 → 2025-08, which none of them had been optimised against.

All fifteen were profit-factor-negative out of sample. The drawdown roughly doubled, which is the more dangerous failure: a contract count sized on the backtest would have understated real risk twofold.

The MGC-tuned default fails the same test — 1.0% simulated pass rate out of sample against 85.8% in training.

What survived, and how little of it there was

Every cached configuration was then scored on four non-overlapping periods and ranked by its worst period rather than its best.

37 of 7,500 were profitable in all four. Every single one uses the Pullback (Counter-Candle) entry with the reversal engine on — the structure is more consistent than any parameter in it.

Ranking those by floor profit factor is misleading, though, because the highest-floor configurations barely make money:

The shipped configuration is the second: Pullback (Counter-Candle) with RSI, MACD-velocity, extension and minimum-VWAP-distance filters, reversal engine on, and a 3.25 ATR target against a 2.0 ATR stop. Its win rate moves 1.5 points across two and a half years — 47.4 / 47.1 / 45.9 / 46.5 — the most stable in the entire sweep, and it earns 3.4x the next candidate.

Contract sizing is not a free lever

Trade order was block-bootstrapped — five-trade blocks, so streaks survive the resample — and the account run forward until it hit the target or the limit. Under a trailing drawdown:

More size buys speed and costs probability, and past three contracts the pass rate keeps falling while the time saved flattens. There is no configuration in this search that supports four or more.

The limit worth stating plainly

Out-of-sample profit factor is about 1.05. An edge that thin needs thousands of trades to accumulate $6,000, so at one contract the simulated time to pass is three and a half years on 2024–25 data against roughly nine months on the trailing year. The same configuration, the same size, two answers an order of magnitude apart.

That is the regime dependence documented in the regime-versus-fit study, restated in evaluation terms: this configuration passes quickly if conditions resembling 2026 persist, and does not pass at all if 2024–25 conditions return. Nothing in this sweep removes that, and no amount of further tuning on the trailing year can.

Round two: the RSI entry, % Price brackets, and the stop-move matrix

Three gaps in the sweep above, all scored on the trailing 365 days only.

The win-rate trade-off was an artifact of excluding one entry

The finding reported above — that win rate and edge trade off one-for-one, best profit factor at a 42% win rate and worst at 80% — held across everything tested, but that search excluded the RSI Reset entry by scope. It does not generalise.

The tight-target/wide-stop shape is only a trap without the RSI filter stack. With it, the same geometry holds a profit factor near 1.45 rather than collapsing to 1.10. A high hit rate and a real edge are jointly available on this entry; they were not on the others.

% Price brackets roughly halve drawdown

Same entry, same filters, both at neutral untuned brackets:

ATR wins per contract and loses the account by 55%. The mechanism is that an ATR-denominated stop widens exactly when volatility spikes — when losses cluster — so the loss distribution carries a fat tail, while a percentage-of-price stop stays proportional to the instrument regardless of recent range. Tuned, % Price produced the only configurations in either sweep to reach four contracts, on drawdowns near $730.

Breakeven, trailing stops and the reversal engine

Every combination, across 26 base configurations — no stop-move, breakeven only, trail only, and breakeven-gated trail, each with the reversal engine on and off:

Breakeven and trailing are null on median, consistent with eleven earlier research stages that scored profit per trade — the result holds under a drawdown objective too. Combining them is worse than either alone.

The exception is narrow and real: a low trigger with a wide distance on a wide-stop configuration, worth about +$1,100. High-trigger trails never fire when the target sits 1 ATR away and return results byte-identical to no stop-move at all.

The reversal engine is strongly negative here — it loses on 45 of 52 pairings at a median cost near $5,000. That is the opposite of the Pullback configurations above, where every four-period survivor had it on. It interacts with the entry rather than being independently good.

The two configurations this produced

Both are RSI Reset with RSI, MACD-velocity, extension and candle-range filters and the reversal engine off; the balanced one adds a trail arming at 0.5 ATR and following 2.0 ATR behind.

The balanced configuration is the better vehicle. The extra thirteen points of win rate cost a lower profit factor, a deeper worst month, and two additional months to target — a hit rate bought with a 0.34 reward:risk makes the average loss roughly three times the average win, and that is what ends an evaluation. 103 cells cleared the same bar (11 of 12 months positive, two or more contracts, profit factor at or above 1.2), so both sit on a plateau rather than a spike.

These were selected on the trailing 365 days alone, at the user’s direction, and carry no out-of-sample validation.