Win Rate

The share of closed trades that ended in profit, as a percentage. 40 wins out of 100 trades = a 40% win rate. On its own it says nothing about how much each win or loss was worth.

In plain English

Win rate (WR) is the most intuitive metric and the most misleading. A high win rate feels like skill, but a strategy that wins 90% of the time can still go broke if its rare losses are enormous. And a strategy that wins only 30% of the time can be excellent if its winners dwarf its losers — the positive skew of a trend following strategy.

So win rate is only half the picture. It has to be read next to the payoff ratio (average win size ÷ average loss size) or, more simply, next to profit factor. Together they tell you whether the strategy makes money; win rate alone never does.

Win rate is also a measured quantity with its own uncertainty. On few trades it is barely pinned down. The honest version is its confidence interval (the band of true win rates consistent with the trades seen), not the bare percentage.

Why it matters for this fleet

Across the 210 EMA-cross variants, win rate ranges from 0% to 66.7%, with a median of 23.2% — and 209 of 210 win less than half their trades. That is not a flaw; it is the signature of trend-following. These strategies take many small losses while waiting for a few large trends. The edge, where it exists, lives in the size of the winners, not their frequency.

The single config that wins more than half its trades is the exception that proves the rule.

Examples from the live fleet

• id478 (EMA 50/200 · BTC · 1d · 2× · long) — win rate 66.7%, the only row above 50% in the entire fleet. It is a 3-trade strategy (2 wins, 1 loss). Its confidence interval is ±36.5 percentage points — the true win rate could be anywhere from 21% to 94%. The one "high win rate" config is a coin-flip-sized sample.
• id523 (EMA 21/50 · SOL · 1h · 2× · long) — win rate 31.9% on 436 trades, yet a profit factor of 1.46: it loses two trades out of three and still makes money, because the average win is about 3.1× the average loss. Textbook positive skew.
• id628 (EMA 9/21 · BTC · 1m · 2× · short) — win rate 10.6% on 10,574 trades. A precisely-known win rate (±0.6pp), and precisely catastrophic — a 145-trade losing streak and a −98% drawdown.
• id511 (EMA 21/50 · BTC · 1h · 2× · long) — win rate 24.9%, pinned tightly (±3.9pp). Knowing the win rate precisely still doesn't make it profitable — its edge is indistinguishable from zero (sharpe significance). Confidence in a number is not the same as an edge.

How to read it honestly

1. Never judge a strategy by win rate alone. Pair it with the payoff ratio or profit factor. A sub-50% win rate is normal — even desirable — for trend-following.
2. A win rate on few trades is barely known. Quote the confidence interval, not the bare percentage. id478's 66.7% (3 trades) and id511's 24.9% (469 trades) are not remotely comparable in trust.
3. A precisely-known win rate is not an edge. id511 proves it: ±3.9pp precision, zero establishable edge.

Related

• profit factor
• positive skew
• confidence interval
• expectancy
• number of winners
• trend following

Sources

• growth/content/dossiers/ema-cross/1-analysis.md (run 83 analysis)
• growth/content/dossiers/ema-cross/1-dataset.csv (the 210-row result set)