Bug: Insights calculates found mate combination incorrectly

So yesterday I played not less than genius (for my elo 🤣🤣🤣) game.

https://www.chess.com/game/live/164558926804?move=0

And was waiting for insights to show my first M5+. But it never did, so I clicked through details of each and discovered that this game was put under M3. The mate combination was M6 if the opponent covers and sacrifices queen or M5 if it doesn’t. Game review shows M6, but insights cut the combination to M3.

I was a little disappointed that insights did not recognise my training results.

would be nice if you could look into that

cheers

K

Insights runs at a different strength and it's possible that level didn't give the move to maximize the mate length at that strength

I tried just now the “fast” (the weakest) option in the app, it still found M6. Does that mean that insights runs in much weaker setting? But it has stat M5+ so I don’t quite get how it then works if it doesn’t see M6.

It also possible that the algorithm for the tactics prioritizes distinct lines. On the M6, there are multiple continuations. I don't know what specific strength Insights runs at, just that it's not the same as Game Review and there can be differences

Exactly — and that’s why I believe this is a bug.

If Insights only tracks the mate sequence along the *best defensive move*, and discards the entire sequence when the opponent plays a worse move, then the M5+ stat becomes logically incorrect.

A forced mate does not stop being forced just because the opponent failed to find the best defense. In my game, every continuation leads to M6 or M5, and even the weakest engine finds it. Only Insights cuts it to M3 because it abandons the original sequence when ...Qh4 wasn’t played.

That means Insights is not evaluating forced mates, but only a few predicted lines with best defense. This is why the mate length is shortened artificially.

This is exactly why I believe the logic needs to be reviewed.

I think there's a major rework going on, that's going to impact the Insights feature, so I don't think any specific new work or fixes on Insights itself is likely.

Thanks for the update.

Since the rework is not confirmed and you mentioned it only as a possibility, I would still appreciate if this specific logic issue could be forwarded to the development team. It’s a fundamental evaluation flaw, not a UI detail, and it directly affects the accuracy of M5+ statistics.

Even if Insights is being redesigned, this is exactly the kind of core logic problem that should be taken into account during the rework. I just want to make sure this issue is not lost or overlooked in the transition.

Please forward it to the dev team so they are aware of it.