If you choose AI based only on win rate, you’ll lose. The first number I look at

The ranking for ‘GENSEN-AI’ has 19 items per model. Many people ask, “There are too many. Which ones should I actually look at?” so I’ll write the order I personally follow. Win rate is not the first consideration.

First, look at the number of settlements.This is the denominator. Even with a 90% win rate, if there are only 10 settlements, it’s indistinguishable from a coin flip landing heads 9 times out of 10. Such events can occur by chance and are within normal variation. If you have 100 or 200 settlements, only then does the win rate gain statistical meaning. A high score with a small denominator is not proof of ability but a sample deficiency.

Next, look at the average pips per trade.As written in previous articles, win rate can be designed with risk-reward adjustments. The key is whether, after equalizing all settlements, there is a net positive per trade. This is the core measure.

Finally, look at win rate to understand personality.Are you the type who takes small profits with high win rate, or the type who takes larger profits with lower win rate? This is not about superiority; it’s about confirming the trader’s temperament.

In summary, the order is “denominator → per-trade average → win rate.” When viewing performance on the sales page or when validating it yourself, this order remains the same. Our validation example (settlements 227, per-trade +2.3 pips, win rate 75.3%, numbers from validation) is also presented as a 3-point set in that order so you can read it in sequence. If you start with win rate, you’re stepping onto the seller’s stage of presentation.