Eight years of real trades. No hypotheticals.
Every number below comes from backtesting the exact rules on ES daily data from 2018 through 2026. Commissions and slippage included. Numbers updated 2026-04-08 after a simulator audit found a trail-stop fill bug that had been inflating results.
Until 2026-04-07, this page reported 113 trades, 67.3% WR, PF 5.86, $799 max DD. A Level 3 simulator audit (commit be6deeb) found that the SMA trailing stop in the H12 backtester was filling above the market on gap-down days, inflating the win rate and suppressing drawdown. Patched 2026-04-08.
The numbers below are from the patched simulator. The directional edge survives. The headline statistics don't.
36 winners. 49 losers. The average winner is +136.2 pts ($681). The average loser is -50.4 pts ($252). Winners are 2.7x larger than losers.
That's where the edge comes from. The strategy loses more often than it wins, but when it wins it wins big. Profit factor 1.99 means every $1 risked returns approximately $2.
The strategy was tested on three separate time periods. Rules fixed in advance, no adjustments between periods.
| Period | Win Rate | Profit Factor | Max DD | Result |
|---|---|---|---|---|
| 2020 - 2021 | 50.0% | 1.60 | $1,169 | PASS |
| 2022 - 2023 | 27.3% | 1.79 | $1,321 | PASS |
| 2024 - 2026 | 34.6% | 2.18 | $1,755 | PASS |
3 out of 3. Bull, bear, and chop. All profitable.
2020-2021 was the post-COVID rally. 2022-2023 included the worst bear market since 2008. 2024-2026 was a choppy recovery. The strategy made money in all three windows even with the corrected fills. The directional edge survives across regimes. What changed post-audit is the per-period drawdown — windows that previously looked tame now show $1,100-$1,800 max DD.
1,000 resampled equity curves from the 85-trade set. How bad can it get?
Position sizing from Monte Carlo: the median resampled drawdown is ~$2,480 per MES contract; the actual historical max DD is $2,699. Within a $3,000 personal-account drawdown budget, the strategy can only support 1 MES contract, and even then it would consume most of the buffer. This is the bug-correction story: pre-audit, the same strategy looked sizable at 2 MES against a $2,000 budget; post-audit, it barely fits at 1.
The trailing-stop buffer was swept across the full range from 3 to 20 points, well beyond a ±20% tolerance. Every variant remained profitable. Per-trade expectancy ranged from $110 to $145; max DD ranged from $2,045 to $2,743. The directional edge is not fragile to the exact buffer, but the DD envelope is large enough that every variant exceeds the personal-account budget.
| Points | Hold | $ (1 MES) |
|---|---|---|
| +791.5 | 23 days | $3,958 |
| +436.5 | 39 days | $2,183 |
| +379.8 | 43 days | $1,899 |
| +367.0 | 48 days | $1,835 |
| +267.3 | 46 days | $1,337 |
Top 5 total: $11,212 (92% of all profits). The strategy is heavily concentrated in a few big winners.
| Points | Hold | $ (1 MES) |
|---|---|---|
| -176.3 | 1 day | -$881 |
| -110.8 | 1 day | -$554 |
| -105.3 | 1 day | -$526 |
| -95.3 | 13 days | -$476 |
| -94.6 | 6 days | -$473 |
Without the top 5 winners, total PnL is $960 — barely breakeven. The whole edge is in the right tail.
The asymmetry is the strategy. Losers are small and fast (often 1 day). Winners are large and slow (23 to 48 days for the biggest). The SMA stop cuts losers quickly but gives winners room to run.
This is also the strategy's fragility. The top 5 winners account for 92% of all profits. Without those five trades the strategy is essentially breakeven. Multi-week trends like 2018-2019 and 2023-2026 produce them. Choppy years like 2021-2022 (combined: -$1,823) don't, because there's no sustained trend for the trail stop to follow.
Imagine you have 85 cards, each representing one trade's P&L. Monte Carlo takes those cards, shuffles them into a random order, and draws them one by one to build an equity curve. Then it does this 1,000 times.
Each shuffle produces a different sequence of wins and losses, and therefore a different maximum drawdown. By looking at the distribution of drawdowns across all 1,000 shuffles, you can estimate how bad things might get under different "luck" scenarios.
The median resampled drawdown is ~$2,480. The historical run came in worse at $2,699. That means the actual sequence the market produced was slightly unlucky relative to the trade-set's expectation, but not pathologically so. The MC distribution tightens the lesson: the drawdown is structural to the strategy's win-rate and R:R, not a quirk of trade ordering.
A common concern with backtesting: "What if the exact parameter is overfit?" The robustness test answers this by varying the trail-stop buffer across the full range.
Buffer values tested: 3, 5, 8, 12, 15, 20 points. If the strategy only works at buffer=12 and breaks at 10 or 14, that's overfitting — a narrow mathematical artifact rather than a real market behavior.
This strategy passes the directional test: every variant from $110 to $145 per trade. The edge exists across a wide range of buffer values because it captures a genuine phenomenon (multi-week equity trends), not a quirk of one specific number. But every variant also has max DD between $2,045 and $2,743 — so the DD constraint binds across the whole parameter neighborhood, not just at one setting.
Everything above uses MES (Micro E-mini), where each point is worth $5. The full-size ES contract is $50 per point. Same strategy, same rules, same edge. Everything scales by 10x.
| MES (Micro) | ES (Mini) | |
|---|---|---|
| Dollar per point | $5 | $50 |
| Margin required | ~$1,500 | ~$15,000 |
| Total PnL (85 trades, 8yr) | $12,172 | $121,720 |
| Per trade average | $143 | $1,432 |
| Average winner | $681 | $6,810 |
| Average loser | -$252 | -$2,520 |
| Max drawdown (MtM) | $2,699 | $26,990 |
| P50 DD (Monte Carlo) | $2,480 | $24,800 |
| Biggest single loss | $881 | $8,810 |
| Annual average | $1,522 | $15,215 |
Same strategy, same rules, same edge. The only difference is the dollar per point.
To trade ES you need a much larger account. Historical max DD on ES is $26,990 so you'd need at least $60,000-80,000 to absorb the worst case. ES margin alone is around $15,000.
This is the takeaway from the bug correction. The strategy still has a real edge ($143/trade after costs), but the DD makes it a poor fit for the $3,000 personal-account budget that motivated the original module. It belongs on the research bench, not the live roster. Module 6 covers what does belong on the live roster.
An $800 max DD strategy at 67% win rate is a different product from a $2,700 max DD strategy at 42% win rate. One bug, found by a Level 3 audit eight months after the original publication, turned the first into the second. This is why we have a 3-gate framework. This is also why we audit our own simulators.
You've been running the strategy for 20 trades. Results: 6 winners, 14 losers (30% win rate). The historical average is 42.4%.
Should you stop trading the strategy?
B) Correct. A 42% win rate over 20 trades can easily produce 30% (or even lower) by random chance. The 2022-2023 walk-forward window actually ran at 27.3% WR for 22 trades and still cleared $2,938. Short samples deviate from the long-term average, especially for low-WR / high-R:R strategies where a few big winners do all the work. You need 100+ trades before drawing conclusions about whether the edge has changed.
The numbers were a story. The audit was a better one. Module 6 covers what the framework looks like when it works as intended — the strategies that are live on Topstep XFA today.
Module 6: Execution & Discipline →