We recently ran a six-engine “confluence” futures strategy through a full tick-level backtest against real prop-firm rules. A confluence strategy only takes a trade when several independent signals agree, and the promise is obvious: more agreement, higher-quality trades. The goal was simple: find out whether there is a genuinely tradeable edge, and if so, exactly how to trade it. The bigger lesson turned out to be that most of a backtester’s job is killing false positives before they cost someone a funded account.
The 30-second trap
The first pass looked great. Several entry windows showed healthy profits. Then we applied the prop firm’s 30-second minimum-hold rule, which does not count trades closed in under 30 seconds. It turned out that 60 to 80 percent of the winning trades were closing within a few seconds, hitting a tight profit target almost instantly. Strip those out and the profit did not just shrink; it flipped to a loss large enough to breach the account’s drawdown limit. A high win rate built on sub-30-second scalps is not an edge. It is an artifact of a target that is too tight.
Bigger targets, one window
Only one slice of the day survived the rule honestly: the evening session. And the fix for the scalp problem was counterintuitive. We made the targets bigger, not smaller. Widening the profit target so trades had to actually develop cut the sub-30-second trades from about 22 percent to about 5 percent, and it raised net profit at the same time. We also found that active stop management, such as breakeven and trailing stops, hurt on the micro contract: it clipped winners short and brought the fast-exit problem back.
Frequency versus proof
Here is the tension every systematic trader runs into. The stricter the entry filter, the better each individual trade, but the fewer trades you get, and the harder it becomes to prove the edge is real rather than luck. Our strict filter fired only about twice a month, far too rare to validate. Loosening it to a three-signal agreement roughly doubled the frequency, and it brought the original evening window back to life.
The test that matters
The gold standard for trusting a backtest is out-of-sample performance with zero optimization. You pick one fixed configuration, run it once across data it was never tuned on, and see whether it holds. This one did. It was positive in every out-of-sample segment, with no breaches of the drawdown limit. A stretch that had looked like a failure in earlier, more heavily optimized tests turned out to be a false alarm caused by over-tuning, not by the market. That is the whole game: the edge survives when you stop letting the optimizer cheat.
Sizing into reality
A validated edge that earns fifty dollars a month is not a business. The last question was how large to trade it. Passing a prop-firm evaluation is a race between reaching the profit target and hitting the drawdown limit, and adding size speeds up both. Past a certain point you simply add risk of ruin without improving your odds of passing. The sweet spot landed at three micro contracts: a better-than-even chance of passing a 50K evaluation at a modest breach risk. The full-size contract was the wrong tool, because its minimum size is too large for the account’s drawdown limit. Sometimes the smaller instrument is the professional choice.
The configuration that survived
- Instrument: Micro E-mini Nasdaq (MNQ), not the full-size contract
- Bars: 70/4 Renko
- Signals: TH + PA + SJ only (KO, SU, NC disabled); all 3 must agree
- Entry: three-signal confluence with one-bar confirmation, single evening window (about 8:00 to 8:45 p.m. ET)
- Exit: a 60-tick target and a 200-tick stop, flat (no breakeven, no trailing)
- Size: three micro contracts on a 50K account
- Result: validated out-of-sample, positive in every segment, with roughly a 56 percent simulated evaluation-pass rate
The headline is not really the settings. It is the process: kill the false positives, treat the account rules as first-class constraints rather than afterthoughts, validate out-of-sample without optimization, and size for survival rather than for the fastest possible pass. That is the difference between a backtest that looks good and one you would actually fund.
This post describes historical backtest research for educational purposes. It is not financial or investment advice. Backtested results are hypothetical, do not represent actual trading, and do not guarantee future performance. Trading futures carries a substantial risk of loss.