Trading Signal Accuracy: How to Measure It Honestly
Updated: 2026-09-17
Every signal service advertises accuracy. Almost none explain how it was measured. Understanding the arithmetic behind a win rate is the single most useful skill for judging any signal source, including ours.
The break-even maths
With a typical binary payout of 80%, a winning trade returns 0.8 of the stake and a losing trade costs the full stake. Break-even therefore sits at roughly 55.6% of trades won.
At a 92% payout break-even falls to about 52%; at 70% it rises to about 59%. This is why payout matters as much as accuracy, and why comparing win rates between brokers without comparing payouts is meaningless.
A service claiming 90% accuracy is claiming a return that would compound an account beyond the size of the market within a year. Treat the number as advertising, not data.
How big a sample has to be
Ten trades tell you nothing: a coin flip produces seven wins out of ten about seventeen percent of the time. Thirty trades start to be indicative. One hundred trades give a usable estimate with a margin of roughly ten percentage points.
The sample must be consecutive. Discarding trades you "would not really have taken" turns a measurement into a story.
Record every signal you received, including the ones you skipped and why. Skipped losers are the most common source of inflated personal statistics.
Building a signal log
Minimum columns: date and time UTC, pair, direction, expiry, confidence score, whether you took it, outcome, and stake. A spreadsheet is enough.
After fifty entries, group by confidence score and by session. Most traders discover their edge is concentrated in one session and disappears in another — that finding is worth more than any new indicator.
Review monthly, not daily. Daily review encourages changing the method after normal losing streaks, which is how consistent systems get abandoned.
What we publish and what we do not
Signal AI publishes a confidence score with every signal and does not publish a headline accuracy figure, because a single number across all pairs, sessions and expiries would be misleading.
Results vary by pair, by session and by how strictly you follow the expiry. Your own log on your own conditions is more informative than any number on a marketing page.
Trading short-term instruments carries a high risk of losing money. Accuracy measurement is about managing that risk honestly, not about removing it.
Free access inside Telegram — no download needed.
Open Signal AI in Telegram