Payoff ratio

What it is

Payoff ratio is the average winning trade divided by the average losing trade.

If winners average +3 and losers average −1, the payoff ratio is 3.

It is the second half of win rate. Neither one means anything alone, and quoting only one is how most strategy pitches avoid saying whether the thing makes money.

Why it matters

Together they give expectancy per trade:

expectancy = win_rate × payoff_ratio − (1 − win_rate) × 1

The same expectancy can come from very different pairs. Winning 30% of the time at a payoff ratio of 4 is roughly as good as winning 80% at 0.5:

0.30 × 4 − 0.70 × 1 = 0.50
0.80 × 0.5 − 0.20 × 1 = 0.20

Both are legitimate strategies. What is not legitimate is quoting the 80% win rate and staying quiet about the 0.5 payoff — the first number sounds like skill, the second is the one that decides the outcome.

Why it matters for AI quant

A model asked to maximise accuracy will naturally push win rate up and payoff ratio down: predicting the common case is easy, and the common case is usually a small move. The trade-off is invisible unless both numbers are reported.

This is the same mechanism behind overfitting to a metric — you get exactly the number you asked for.

How to check

Ask for the ratio with the sample size. A payoff ratio of 5 built on two winning trades is not a fact about the strategy; it is a fact about two trades.

Then ask the follow-up: what happens to the ratio if the largest single winner is removed? If it collapses, the edge rests on one event.