The Workshop · lesson 4
A strategy that is wrong 60% of the time can beat one that is right 70%.
Ask someone what makes a good trader and they say "being right." Ask a professional and they say "being paid more when right than it costs when wrong." Those are different jobs. Set the two dials and run two hundred trades.
The number that matters is expectancy: hit rate × average win − (1 − hit rate) × average loss. Forty percent with wins twice the size of losses is +0.2 per unit risked — a business. Seventy percent with wins half the size of losses is +0.05 — barely. Seventy percent with the occasional large loss is how accounts die: the wins are frequent enough to feel like skill right up to the loss that erases them.
Fact, from the desk's own replay. The strength-run shape on TradeMath finishes up 41% of the time. Its average win is about 5.8%, its average loss about 3.0%, and its profit factor (all wins ÷ all losses) is 1.33. It loses more often than it wins and makes money, because the losses are cut by a rule and the wins are allowed to run.
Not because of a stop-loss price — those tested worse than no rule at all on this desk, because a spread breathes with noise and a stop just locks the breathing in. The losses are small because the exit is a state: the moment the reason for the trade is gone (Alpha back through its line, three narrowing days), the trade is over, win or lose. Wins are not cut by the same rule because the state persists while they run.
Every trade line shows its hit rate beside its average, never alone, and the tracker's Done list keeps your own — with the average win and loss beside it. Watch your own payoff ratio for a month. It tells you more about your trading than your win rate ever will.
Keep this: you do not need to be right often. You need to be paid well when you are, and to leave fast when you are not.
The Workshop is education, not advice. Replayed numbers are averages over many trades, after costs unless stated, and are not forecasts; the live record is on the engine page. Open the deskManualEngineChangelog