How Do I Tell If a Trading Mentor Is Any Good?

Grade the advice, not the presentation

Confidence, screen time and a good chart do not carry information about expectancy. Two numbers do, and both are arithmetic anyone can run on the mentor's own claims.

Breakeven win rate is 1 / (1 + R). At 1:1 you need half your trades to win to stand still. At 2:1, 1 in 3. At 0.5:1, 2 in 3. So a mentor teaching quick profit-taking at half the risk is teaching a method that must be right two times in three before a dollar exists — which may well be achievable, but it is now a stated requirement rather than a vibe.

Expectancy is (win% x avgWin) - (loss% x avgLoss). Run it on the numbers they advertise. A claimed 80% win rate at 0.5R against a 1R loss is 0.8 x 0.5 - 0.2 x 1 = 0.20R per trade, which is genuinely good. The same 0.5R method at a 70% win rate is 0.7 x 0.5 - 0.3 x 1 = 0.05R — a twentieth of a unit per trade, thin enough that costs decide the outcome. Ten points of win rate is the whole business.

What a good mentor makes possible

The test is not whether they are right. It is whether being wrong would be visible. Advice stated as "I am long here, invalidated below 412, first target 419" can be graded by anyone at any later date. The same view stated as "watching for strength above the range" cannot be graded by anybody, including the person who said it — there is no version of the next two weeks that contradicts it.

Ungradeable advice is not necessarily dishonest. It is simply not evidence, and it cannot accumulate into a record no matter how long you follow it.

How long before you can judge?

Longer than the trial period. Twenty trades tells you almost nothing: at a genuine 60% win rate the 95% interval over 20 trades spans roughly 38.5% to 81.5%. A hundred trades narrows it to about 50.4% to 69.6%, which is finally a range with an opinion in it. Structure the decision around that: judge the method on the arithmetic first, then let a real sample accumulate before you conclude anything from the results.

Keep your own record while you learn

The reason to keep it is not the mentor. It is that "did this help?" is a question about your trades, and the honest version needs your entries, stops and exits as they stood before each outcome — not as you remember them after a good week. A record written afterwards converges on your current opinion of the method, which is precisely the thing under test.

One failure mode deserves naming because it looks like evidence. A strategy tuned until its backtest looks excellent is a known and quantified problem: the probability that a backtest is overfit rises with the number of configurations tried, and almost nobody reports how many they ran.^[1] If a mentor's method arrived with a beautiful historical curve, the question is how many versions preceded it.

The same scepticism has a stated threshold. Harvey, Liu and Zhu argued that finance has tested so many predictors that a new claim should clear a t-statistic above 3.0^[2] — a bar no mentor's results will be presented against, which tells you how to weight them. Weigh the direction of the advice too: across 66,465 households, the most active traders earned 11.4% a year against a 17.9% market,^[3] so a method that asks for more activity starts from behind.

kappi is a trade recorder: you commit a trade before the fact, it is sealed on your device for a time-capsuled delay you choose, then kappi publishes it on a Merkle-anchored log. The record is yours to share, with whoever you choose. $15/month, no free tier.