The Workshop · lesson 8
Stop-losses tested worse than no rule. What tested better.
Most people think the entry is the trade. On this desk the entry is the cheap part; the exit is where the number is made. Here is a toy market to see why. Each trade is a run with a small drift and a lot of noise. Pick an exit, run five hundred of them, and compare.
| exit | average | finish up | worst 1-in-20 | days held | per day |
|---|
Three exits. Hold 21 days collects the whole drift and then whatever comes after it — half these runs reverse, so it also collects the reversal: the worst tail of the three, and three weeks of capital for every trade. The price stop cuts that tail, but it fires on noise: raise the noise and watch its hit rate fall below the hold's as runs that would have finished up get stopped on a wobble, with the average about the same. The state exit leaves when the reason for the trade is gone — in the toy, when the run's own trend has turned — and keeps the drift, skips most of the reversal, and hands the capital back in twelve days instead of twenty-one, because it is asking "is the run still a run?" rather than "did the price wobble?"
Fact, from the desk's replay. On fresh strength runs, leaving when Alpha closes back under 60 kept about 80% of the 21-day hold's average in less than half the time — twice the return per day held — with a worst one-in-twenty of −9% instead of −22%. In the 2022 grind, where every fixed hold lost, the same exit removed two-thirds of the loss. Price stops on the same entries cut the worst case and the hit rate together and left the average where it was. Trailing stops and profit targets changed nothing. On pairs, a stop-loss tested worse than having no rule at all.
A stop asks the price to promise not to fall. Prices do not make promises; a run that ends at +8% routinely passes through −3% on the way. A state exit asks whether the thing you bought — a name in a run, a spread that is widening, a sector hand-off — is still true. When it is, a dip is noise. When it is not, a dip is the exit and so is a rally. The desk's exits are all of this kind: the run's state, the spread's state, the leg's state, or a calendar cap for when the state never resolves.
Every trade line names its exit and was tested only with it — the number beside the trade assumes you take it when it fires. The tracker tells you the moment it does, by email or in the browser if you want, and the closed view shows what it cost you when you did not. That last number is lesson 10.
Keep this: exit on the state of the trade, not the state of your account. A rule you set in advance is the only exit you will actually take.
The Workshop is education, not advice. Replayed numbers are averages over many trades, after average trading costs (2 bps a side on a stock, 1 on a fund, included for fairness; yours vary by broker and liquidity) unless stated, and are not forecasts; the live record is on the engine page. Open the deskManualThe EngineChangelog