What is a signal worth? Win rate, profit factor and honest math

"We win 85% of our trades!" is the favorite line of people hoping you can't do math. Let's do the math.

Win rate alone is an empty number

Win rate is the percentage of trades that close positive. It sounds like the definitive number. It isn't, because it ignores how much you win and lose on each one.

A system that wins 85% of the time making €10, and loses €100 on the other 15%, is a money-destroying machine: 85×10 = +850; 15×100 = -1,500. In other words, it loses €650 per 100 trades while winning 85% of them.

Conversely, a system that wins only 40% of the time, but makes €300 when it wins and loses €100 when it fails: 40×300 = +12,000; 60×100 = -6,000. It makes €6,000 while being right less than half the time.

The pair that tells the truth: win rate + payoff

• Payoff (average win / average loss): what a win is worth in multiples of a loss.
• Profit factor: total won divided by total lost. Above 1, the system created value; 1.5+ over a long track record starts to be respectable; numbers far above that over short periods are usually just a small sample.

Rule of thumb: distrust any advertised win rate without the payoff next to it. It is half an equation sold as a result.

The third element: the sample

10 trades tell you nothing; luck dominates small samples. Before taking metrics seriously, ask:

• How many trades in total?
• Over how long, and across how many different market regimes?
• Do the numbers include costs (commissions, slippage)?

The honest math checklist

• Win rate accompanied by payoff (or profit factor)
• Maximum drawdown next to the return
• Sample: number of trades + period + regimes
• Costs included
• All of it verifiable in a public record, not a screenshot

Whoever has these numbers published doesn't need to shout percentages. The data speaks softly, and that is what serious systems sound like.

---

Kamla's metrics (win rate, profit factor, drawdown) are calculated on the public trade record, and show as "accumulating" while the sample is small, because premature numbers are marketing, not math. See them at kamla.ai/performance.