How to check whether a trading signal record can be falsified

The defining property of a claim worth reading is that it could have turned out otherwise. Most published trading records fail this test before any of their numbers are examined, because the way they are constructed makes an unfavourable result impossible to produce. That is not usually fraud. It is a structure that quietly removes every route to being wrong.

The four structures that make a record unfalsifiable

1. The call has no invalidation

"Bullish above support, watch for continuation" cannot be scored. There is no level at which it was wrong and no time by which it had to be right. Every subsequent price path can be narrated as consistent with it. A scoreable call names a direction, an entry region, a target and the level at which the idea is dead — the same discipline described in why every trade idea needs an invalidation level.

2. The outcome is selected after the fact

If a provider publishes twenty ideas and reviews four, the review is about the reviewer's memory, not the ideas. This is the most common failure and the hardest to detect from outside, because the deleted calls leave no trace. The only structural defence is a record where every published call appears, including the ones nobody wants to discuss.

3. The scoring rule is decided per trade

If a call can be scored as a win because price touched the first target, or as a loss because it later reversed, or as scratch because the trader would have exited manually, the rule is being chosen to fit the result. A fixed rule, written before the call and applied mechanically — first level touched wins, expiry after a stated time, no discretionary exits — removes that freedom.

4. The denominator is missing

A percentage without an n is not a measurement. Three decided calls out of three produces a headline that looks identical to 300 out of 300 and means almost nothing, as the sampling arithmetic in losing streaks and drawdown maths shows. Every honest record prints the sample size beside every rate, including inside each breakdown rather than only in the total.

The seven questions to ask of any published record

1. Was every call timestamped and published before the outcome was known, in a place that cannot be edited later?
2. Does each call carry an explicit entry, target and invalidation level?
3. Is the scoring rule stated in advance and applied identically to every call?
4. Are all calls included, or only a selection — and how would a reader tell?
5. Is the sample size shown next to every percentage, including each sub-breakdown?
6. How are undecided outcomes handled: expired, still open, or quietly dropped?
7. Are fees, funding and slippage excluded — and is that exclusion stated plainly?

Question six catches more bad records than the other six combined. A call that neither reached its target nor its invalidation before expiring is the largest bucket in most real records. If a published rate is computed only over decided outcomes and the expired bucket is not shown alongside, the headline is arithmetically true and practically misleading.

Screenshots, and why they prove nothing

A profit-and-loss screenshot answers none of the seven questions. It has no denominator, no invalidation, no scoring rule and no record of the positions that were not screenshotted. It is also trivially producible: on a platform allowing high leverage, a small number of accounts running opposite positions guarantees that at least one of them ends the week with an impressive image.

The same applies to a chart with an entry arrow drawn after the move. The question to ask of any marked-up chart is simply: where was this published before the candle formed, and can that be checked? If the answer requires trust rather than a timestamp, the chart is decoration.

How this product answers those questions

Applying the list to ourselves, since it would be absurd not to. Every call the engine publishes carries a direction, an entry, a target and a stop at the moment it is issued. Each one is timestamped, is scored against a fixed rule, and appears on the public track record whether the outcome was favourable or not — including the expired bucket, which is currently the largest one. Sample sizes are printed beside the rates rather than in a footnote.

What that record is not: it is the measured history of a published model's calls, gross of fees, funding and slippage, with no capital deployed behind it. It is a description of what a published model did, not a statement about what any reader would have experienced or what the model will do next. Those are different claims and only the first one is checkable.

The reason to build it this way is not modesty. A record that can come out badly is the only kind that carries information when it comes out well, and a small sample stated honestly is more useful than a large claim stated vaguely. The same standard is worth applying to your own results — see how to build a trading journal for the fields that make a personal record scoreable rather than anecdotal. What is published, and on what delay, is set out on the pricing page.