Why Win Rate Alone Is Misleading

"Win rate" is the most quoted statistic in trading marketing because it sounds decisive: an "85% win rate" feels like a near-certainty. But on its own it is one of the least useful numbers you can be shown. Strip away the size of the wins and losses behind it and a glittering win rate can be hiding a strategy that loses money on every cycle.

Win rate measures frequency, not profitability

Win rate answers a single question: of the trades that closed, what share finished in profit? It says nothing about the size of those wins or the size of the losses. Two strategies can share an identical win rate and have opposite outcomes — one compounding steadily, the other bleeding capital — purely because of how big the average winner is relative to the average loser.

That gap is captured by expectancy: the average result you can expect per trade over a large sample.

Expectancy = (win rate × average win) − (loss rate × average loss). A strategy is only worth running when this figure is positive after realistic trading costs.

A worked illustration

The arithmetic makes the point quickly. Consider two strategies that take the same number of trades:

Strategy A wins far more often and loses money. Strategy B loses more often than it wins and is strongly profitable. A buyer shown only the win rate would pick exactly the wrong one. This is why a headline win rate, presented without the return distribution beside it, tells you almost nothing about whether a strategy makes money.

What a win rate hides

When you see only a win-rate figure, several things that determine real profitability are invisible:

• Average win vs average loss — the single most important missing piece, as shown above.
• The shape of the losses — a few enormous losers can sit behind a high win rate.
• Drawdown — the worst peak-to-trough decline you would have lived through.
• Costs — a gross win rate ignores fees and slippage, which turn marginal trades into losers.
• Sample and selection — a win rate from a tiny, hand-picked, or in-sample period is not evidence of a repeatable edge.

How Vector Ridge frames it

Because win rate alone is misleading, our conviction grades (A–D) are deliberately built on both sides of the equation. A model's grade is set by its own forward track record using win rate first and average net return second — so a grade can never be earned by a high hit-rate paired with poor returns. Frequency and size both have to hold up.

Two further choices keep the return side honest:

• Results are reported per trade and net of realistic trading costs — not gross, idealised, or cost-free figures.
• Each trade is weighted equally — no position sizing and no compounding assumptions layered on top to flatter the curve — and every trade is recorded, winners and losers alike, so the distribution behind any aggregate stays visible rather than smoothed away.

This is also why grades are only comparable within the same model. A short-horizon model and a multi-day model produce wins and losses on completely different scales, so their win rates are not interchangeable — an A is "best of its own lane," never a universal ranking.

What to ask before trusting a headline number

1. What is the average win versus the average loss? (Win rate without this is meaningless.)
2. Are the returns net of fees and slippage, or gross?
3. What was the worst drawdown, and are losses shown as well as wins?
4. Over how many trades, and was the period cherry-picked?
5. Is the record independently verifiable, or simply asserted?

That last point is the strongest defence against a misleading statistic of any kind. See how results are verified for why a track record that is locked on the record before each outcome is known is far harder to dress up, and how to evaluate a signal service for the full buyer's checklist.

Trading involves substantial risk of loss, and past performance is not a reliable guide to future performance. No win rate, expectancy figure, or grade is a guarantee of future results — see do you guarantee profits?