Page 1 of 1

New Engine! But disappointing WAC results

PostPosted: 11 Nov 2018, 21:56
by AxolotlFever
Hello everbody,

Over the last couple of months I have been writing my own chess engine, and tested all changes against the newWAC suite, as well as a few others. I have however hit a barrier, at a WAC score of 245 / 300 correct, at 10 seconds a move. I have tried many things to improve this score, but I cannot seem to make it happen. Strangely, when I increase the complexity of my evaluation function, this score tends to drop (with all evaluation factors "turned on", I get a score of 222/300).

When I read stories of other people's engines I see scores of above 290/300, and so I am really worried there is a problem which is causing the lower score.

Do you think there is a problem with my engine? As I wrote everything myself, and do not use magic bitboards or anything too fancy, could this just be a problem of speed / nps? My engine is written in Java if this makes a difference.

I am extremely grateful to anyone who takes the time to answer, please let me know if there is any information I should provide to make this situation clearer. Here is the link to the code on github, I hope it is ok to link this!

PS: in particular, WAC 2 is giving me trouble, and I do not get the solution in under 5 minutes :(

Thanks all!

Re: New Engine! But disappointing WAC results

PostPosted: 18 Nov 2018, 23:53
by jdart
The vast majority of the WAC positions are solvable by most engines in under one second. If you are not solving almost all of these you probably have a bug.

WAC 2 is a little harder than average. I think null move can affect this test. Try running with null move disabled and see if you get a different result. I notice you don't do null-search verification. You might consider adding that.


Re: New Engine! But disappointing WAC results

PostPosted: 19 Nov 2018, 22:35
by AxolotlFever
Hi Jon!

Thanks very much for your reply. Sadly I think there might be a bug somewhere, I will go on a hunt. I do know one Problem though, which is that my general "nps" is very very low. Possibly this is playing a role. I will look into your suggestions, thank you for making them.

King regards,