The Workshop · lesson 10
What discretion costs, measured — and six questions for any track record.
Nobody overrides a rule to make less money. Every override feels like judgment: take the profit while it is there, skip the next one after two losses, press the bet after a win, give the loser a little more room. Apply them to a thousand trades shaped like the desk's — 41% winners, wins about twice the size of losses — and see what judgment costs.
| by the rules | with your overrides |
|---|
Taking profits early is the expensive one, and the one that feels best: it raises the hit rate, which is why people do it, and it removes the large wins that a 41% strategy exists to collect. Skipping after losses costs nothing per trade and something in total — the skipped trades had the same expectancy as the others; you just took fewer of them, at random, at the moments you were least calm. Doubling after a win does not change the average; it changes the swing, and the swing is what you pay for in drawdown. Giving losers room is the quiet one: the rule-cut loss is the whole design, and a loser allowed to run is the only way this shape produces a large loss.
Fact. Every one of these is measurable in your own book, and the desk measures it: the tracker's closed view shows your result against the rule's on the trades you both took, the trades on the list you skipped and what they did, and the trades you took that were not on the list. Most people who look find that their discretion cost them a few percent a year and most of their worst drawdown — not because their judgment was bad, but because it was exercised at the worst moments, which is the only time it is tempting.
Yours, a fund's, a newsletter's, this desk's. 1. Over what window, and who chose it? A record that starts after the bad year is a brochure. 2. Return per unit of swing, not return: what is the Sharpe, and what was the worst drawdown, in percent and in months? 3. How much of it was the tide — what is the beta, and what does the record look like in the years the market fell? 4. After what costs, and with what fills? 5. Is it live, and is the live part shown separately from the replay? A replay is a claim; a live record is evidence; a live record you can verify is proof. 6. What were the rules, and did they change — and if they changed, is the change dated, explained, and tested against the version it replaced?
The answers are on the engine page, in this order: the window and every year in it, including the losing ones; Sharpe, drawdown and underwater time beside the return; beta and the bear year; costs charged; the live house book separate from the replay, with a daily hash you can check; and a timeline of every change to the rules with what it was tested against — including the findings that were set aside. If a record you are shown cannot answer the six, the answer is the record.
Keep this: discretion is a cost, and it is measurable. Measure yours before you trust it — and ask any record the six questions before you trust it either.
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