Recalculating old puzzles


In a comment to problem 1478 @am34 writes

All 6 players have played 15-19 on move #1 here which is instantly recognized as a clear win according to Kingsrow but it is incorrectly labelled as missing a win here. I expect the same is true for 1-6, which is also a winning move. This is a fantastic site but I think the scores for each move should be recalculated using an engine with an endgame database as this is not the first time I’ve come across this issue.

I’m moving the discussion here for more space, better formatting, and email notifications.

Thank you for pointing this one out! The current problem generator uses a seven piece win-loss-draw database, and I’m hoping the next version will use the ten piece one. But that doesn’t help old puzzles like this one. I agree, it needs to be recalculated so that those moves get labeled as alternates or it gets disabled for having too many alternates. And there are lots of other old problems that could be improved this way.

The reason I haven’t done it is that I haven’t figured out how. I don’t know what to do with attempts or comments that become irrelevant. Lets say 1.15-19 becomes an alternate. It doesn’t seem fair to just delete those old “mistakes”, they are part of a user’s rating history. And who knows what players would have chosen if they had gotten to try again. But if they aren’t deleted then users will see those mistakes when they are no longer possible which is confusing. I’m not sure what to do here.

So I’ve disabled this one for now. If/when you bump into others, please point them out as well. And if you have any ideas on what the right way to recalculate old puzzles is, I’m all ears. I’ve asked Richard from chesstempo what he does. He’s been generous in the past, here’s to hoping he has the time to answer a question about a site that’s not even his.

And thank you again for pointing this one out.


You can get the 10 piece end game WLD database for free from Kingsrow.


@Lloyd, thank you, I think I have it somewhere. The problem is the engine doesn’t give any analysis after finding a draw in the database, it just says ‘draw’. And so the principal variation lines in the puzzle don’t get populated to show why it’s a draw. I want to force the engine to build these lines before I use the larger database in the puzzle generator. The seven piece has this problem too but the engine has to search farther before finding it so more positions get analysis.