Writing Entry/Exit Rules You Can Actually Test

The Five Rules Every Strategy Needs

Before any replay session, write these down:

1. Entry — the exact conditions that trigger a trade, on a named timeframe.
2. Exit — how a trade closes in profit (fixed target, trailing, opposite signal).
3. Stop — where the stop goes (structure-based, fixed pips, ATR multiple).
4. Size — risk per trade, from which position size is derived.
5. Management — breakeven moves, partial exits, or none.

Each as an if-then: “IF the M15 closes above the prior session high AND the H1 trend filter is up, THEN enter long at the close, stop below the breakout candle’s low, target 2R, no management.” That’s testable. “I buy breakouts when they look strong” is not.

Operationalizing the Vague Bits

Discretionary strategies lean on words a computer would choke on: “strong,” “clean,” “obvious.” You have two honest options per term:

• Operationalize it — translate to something checkable. “Strong momentum” → “the breakout candle’s body is ≥1.5× the average of the last 10 bodies.” Now it’s a rule.
• Admit it as judgment — keep the discretion, but apply it consistently and tag it in the journal so you can later check whether your “A+ only” filter actually added expectancy. Manual backtesting legitimately tests judgment (that’s its edge over automation) — as long as the judgment is the same judgment you’ll use live.

What you can’t do is leave it undefined and decide trade-by-trade as results roll in — that’s how rules quietly drift to fit the data (overfitting).

The “Stranger Test”

The single best check on rule quality: could someone else read your document and take the same trades you would? If they’d diverge, your rules contain hidden judgment you haven’t surfaced — fine to keep, but name it. This test also future-proofs you against yourself: the version of you running the backtest and the version trading live three months later should agree on what a setup is.

Common Rule-Writing Failures

• Entry without invalidation — defining the setup but not what kills it leads to held losers.
• Exits vaguer than entries — most traders specify entries precisely and “manage” exits by feel; the exit is half your expectancy, so spec it equally.
• Untestable conditions — “when the news is bullish” can’t be replayed objectively; either proxy it (a price-based filter) or scope it out.
• Rules that need the future — “enter at the candle that turned out to be the low” is look-ahead bias hiding in the rule itself.

Write the five rules, pass the stranger test, then run the backtest. Rules good enough to test are usually rules good enough to trade — the writing is part of the strategy development.

Frequently Asked Questions

Can a purely discretionary strategy be backtested at all?

Yes — that's exactly what manual backtesting is for. The requirement isn't zero judgment; it's consistent, surfaced judgment: you apply the same discretionary filter every time and tag it in the journal, so afterward you can measure whether the discretion helped. Undefined, trade-by-trade improvisation is what can't be tested.

How specific is specific enough?

Pass the stranger test: another trader reading your rules takes materially the same trades you would. If they'd diverge on entries, the entry rule has hidden judgment to surface; if they'd diverge on exits, your exit is underspecified. Specificity is sufficient when the divergence is small and the remaining judgment is named.

Should entry and exit rules be equally detailed?

Yes, and most traders under-specify exits. Your exit logic determines the size of winners versus losers — half of expectancy — yet it's commonly left to feel while entries get paragraphs. If anything, spend more care on exits; they're where vague rules leak the most edge.

What if my best trades come from breaking my rules?

Then your rules are incomplete, not your discipline excessive — the profitable 'rule break' is an unwritten rule you're applying intuitively. Surface it, define it, and test it as its own condition. If it survives testing, it joins the rule set; if it was luck, the test reveals that too.