Stop discipline & mistake patterns

The stop-loss rule

The rule is blunt: sell any position that falls a fixed percentage (default 8%, configurable, overridable per position) below your cost — no exceptions, no “it’ll come back.”

The math behind the bluntness: losses compound against you asymmetrically. An 8% loss needs about a 9% gain to recover; a 25% loss needs 33%; a 50% loss needs 100%. The rule caps every mistake at the recoverable size. Its cost is real — you will sometimes be stopped out of a stock that then goes on without you — but it’s the premium on the only insurance that keeps one bad idea from undoing a good year.

The app surfaces the stop everywhere it’s relevant: the stop price on every open position, portfolio heat as the sum of all of them, and — after the fact — whether you actually honored it.

Mistake patterns

Once you have enough closed trades, the app replays each one against daily price history and looks for recurring, priceable habits:

• Ignored stops — how often price crossed your stop level and you held anyway, and what the extra slippage beyond the stop cost on average. This is the pattern that turns “I know I should honor stops” into “ignoring the stop cost me an average of $X, N times out of M.”
• Chasing extended — buys made beyond the buy zone of a recorded pivot, and how their outcomes compare to your in-zone entries.
• Giving back gains — positions that were up past the profit-taking zone and were then held back down into a smaller gain or a loss.

Each pattern is gated on a minimum testable sample — if you’ve only broken a rule twice, the app says “not enough data” rather than manufacturing a statistic. With AI enabled, the findings are phrased as short coaching notes; without it, you get the same numbers as plain lines. In both cases the numbers are computed first and the words are generated from them.

Caveats

• The replay uses daily bars, so “price crossed your stop” means the day’s range crossed it — intraday sequencing is not knowable from daily data.
• The patterns describe your recorded trades. Unjournaled trades, transfers, or partial fills entered as one line will blur them.
• A pattern is a hypothesis about a habit, not a verdict. Three instances is the start of a question, not proof.