Win-rate tracking

Every ChartLens analysis with a numeric stop and target is logged the moment it is produced and checked hourly against exchange candles until it resolves as target hit, stopped, or invalidated. Nothing is removed afterwards, and the win rate shown in the app is the same one published on this site.

Why this feature exists

Every product in this category produces analysis that reads with the same confidence whether it turns out right or wrong, and none of them count. That is not a small gap — it is the entire reason a user cannot tell a good tool from a convincing one.

Counting is the only claim that a competitor cannot match by rewriting their landing page. To publish a record they would have had to be logging outcomes all along.

What counts, and what does not

An analysis enters the record only if it made a falsifiable claim: a tradeable symbol, a numeric stop and a numeric target. Two kinds of analysis are excluded:

“Wait” conclusions. If the model says there is no edge here, there is nothing to be right or wrong about.
Approximate reads. If the price axis could not be read, there are no numbers to check.

Both appear in your history labelled Not tracked. Excluding them keeps the rate honest in both directions — counting a "wait" as a win would be absurd, and counting it as a loss would punish the model for the most useful thing it does.

The three outcomes

Target hit — price reached the first target before touching the stop.
Stopped — price touched the stop first. Realised −1R.
Invalidated — price hit the stop before ever trading into the entry zone. The idea was never fillable, so it scores zero rather than counting as a loss.

Anything unresolved after 30 days expires at zero. A trade idea that has not done anything in a month has been answered.

Sample size honesty

No win rate is published — in the app or on this site — below 20 resolved analyses in a group. A 100% rate on three calls is not a track record, and printing it would undermine the exact claim this feature exists to support. Groups below the threshold show a dash and their current count.

What it does not prove

It measures whether the levels an AI produced were touched. It models no fees, no slippage, no partial fills and no position management, so it is not a claim about what a human trader would have earned. That distinction is stated on the record page itself rather than buried here.