Positive Skew (fat right tail)

A return profile where most trades are small losers but a few rare winners are huge — the few big wins carry the entire profit.

In plain English

"Skew" describes the shape of a strategy's distribution of trade outcomes. A positively skewed strategy:

• Loses (or barely scratches) on the majority of trades.
• Makes its money from a minority of trades that are very large.
• Has a Win Rate (WR) well below 50% but is still profitable, because the average winner is several times the size of the average loser.

The classic shape of trend following: you take many small probing losses waiting for a real trend, and when one arrives it pays for all of them and more. The single most useful number for spotting it is the avg-win ÷ avg-loss ratio — well above 1 means the winners are doing the heavy lifting.

Why it matters for this fleet

Every ema cross strategy is positively skewed. This drives two non-obvious consequences:

1. Low win rate is not a red flag. A 28% WR strategy can be excellent if its winners are ~2.5× its losers. Judging these by WR alone is a mistake — use profit factor and the avg-win/avg-loss ratio.
2. A take-profit is poison. A take-profit (take profit stop loss) caps the winners — it amputates the fat right tail that is the edge. Capping a positively-skewed strategy's winners while still taking all the small losses turns a profitable system into a losing one.

Examples from the live fleet

• id523 — EMA 21/50 · SOL · 1h · 2× · long: win rate (the share of trades that close in profit) is 31.9% — below half, so you lose more than two trades out of three — yet profit factor (gross profit divided by gross loss) is 1.46. How? The average win is about 3.13× the average loss (the payoff ratio — avg win ÷ avg loss). A few big winners carry the profit-and-loss while most trades are small losers. That is textbook positive skew, and it is exactly why this row clears the edge bar despite the low hit rate.

The same shape — sub-50% win rate, payoff ratio well above 1 — repeats across the ema cross fleet. Judging any of these rows by win rate alone would mislead you; the avg-win/avg-loss ratio and profit factor are what reveal the skew.

Caveat — the "a take-profit destroys the skew" example can't be shown. The classic demonstration is to bolt a take-profit (take profit stop loss) onto a positively-skewed strategy, watch the win rate rise but the payoff ratio collapse, and see profit factor fall below 1 — proof that capping the winners amputates the fat right tail that is the edge. The dossier #1 baseline has no take-profit or stop-loss on any row, so there is no bracketed twin to compare against. The mechanism stands; the worked counter-case will land when a TP/SL dataset exists.

Related

• trend following
• take profit stop loss
• profit factor
• expectancy
• win rate
• number of winners

Sources

• wiki/qa-sessions/2026-06-01-session.md#q2 (first asked here)
• /api/analytics (avgWin / avgLoss / winRate fields)