Bug in my engine

[Ronald]

Hi,

I expect you get a "position startpos moves e2e4..." etc. from lichess with every new position. If you replay it afterwards with a "position fen" you don't have the positions that happened before the current position. If that is the case there might be a bug in your repetition detection, which will be executed in lichess, but not with the position fen.

Ronald

[JimmyRustles]

Hi Ronald,

I've thought about that, but I paste the fen from before the repetition started so the threefold repetition should still work the same. Also, when I try it in winboard, I paste the full game and still can't reproduce the problem.

[Guenther]

Hi Ronald,

I've thought about that, but I paste the fen from before the repetition started so the threefold repetition should still work the same. Also, when I try it in winboard, I paste the full game and still can't reproduce the problem.

Steven, I could reproduce a few full test games with that bug yesterday and will post a link to it in a few minutes here.
The main post will be in the BanksiaGUI thread in the general forum, because it also refers to a bug in the current Beta.
I will also add the logs for download at my RWBC site.

@Ronald, I had already yesterday attached a post, which had the full move history, not only a FEN and still I got the bug behaviour
on my machine.

[Guenther]

Steven, I could reproduce a few full test games with that bug yesterday and will post a link to it in a few minutes here.
The main post will be in the BanksiaGUI thread in the general forum, because it also refers to a bug in the current Beta.
I will also add the logs for download at my RWBC site.

http://talkchess.com/forum3/viewtopic.p ... 07#p823107

[JimmyRustles]

Thanks a lot for doing the tests with Raven, Guenther. I suspect it's a problem with the TT and mate scores. I'm going to try not using the TT when the TT score is mate and see if it fixes the bug. If it works, it shouldn't affect playing strength much, I don't think.

[Ronald]

I took a quick look at the function isThreefold() and I saw that a true is returned after numrepeats >= 3. A threefold repetition is reached after 2 repetitions of a position however. It will probably not be the cause of the bug you are trying to solve, but I think it's not correct.

[JimmyRustles]

Thanks Ronald. The numrepeats variable should probably be called numoccurences. It counts the number of occurences of the position in the stack, not the number of repeats. So 3 occurences is equal to 2 repetitions which is why it's >= 3 and not 2.

[JimmyRustles]

Steven, I could reproduce a few full test games with that bug yesterday and will post a link to it in a few minutes here.
The main post will be in the BanksiaGUI thread in the general forum, because it also refers to a bug in the current Beta.
I will also add the logs for download at my RWBC site.

http://talkchess.com/forum3/viewtopic.p ... 07#p823107

Hi Guenther, I'm pretty sure I've fixed the bug now. I just tried it in 40 lichess games and the bug didn't happen in any of them.

Would you be able to try it with a couple of positions where it had the bug before and see if it's working now please?

I've compiled a version of Raven with the bug fix that you can use to test it, it's here: https://github.com/sgriffin53/raven/tree/master/temp

Thanks again for all the testing you've done. Hopefully this should fix the bug.

[Guenther]

Steven, I could reproduce a few full test games with that bug yesterday and will post a link to it in a few minutes here.
The main post will be in the BanksiaGUI thread in the general forum, because it also refers to a bug in the current Beta.
I will also add the logs for download at my RWBC site.

http://talkchess.com/forum3/viewtopic.p ... 07#p823107

Hi Guenther, I'm pretty sure I've fixed the bug now. I just tried it in 40 lichess games and the bug didn't happen in any of them.

Would you be able to try it with a couple of positions where it had the bug before and see if it's working now please?

I've compiled a version of Raven with the bug fix that you can use to test it, it's here: https://github.com/sgriffin53/raven/tree/master/temp

Thanks again for all the testing you've done. Hopefully this should fix the bug.

I checked a few positions and the buggy behaviour seems corrected.
To be sure I will play some test games too, but probably after xmas.

I see that your Lichess rating in bullet received some nice +50 already :)

[JimmyRustles]

Steven, I could reproduce a few full test games with that bug yesterday and will post a link to it in a few minutes here.
The main post will be in the BanksiaGUI thread in the general forum, because it also refers to a bug in the current Beta.
I will also add the logs for download at my RWBC site.

http://talkchess.com/forum3/viewtopic.p ... 07#p823107

Hi Guenther, I'm pretty sure I've fixed the bug now. I just tried it in 40 lichess games and the bug didn't happen in any of them.

Would you be able to try it with a couple of positions where it had the bug before and see if it's working now please?

I've compiled a version of Raven with the bug fix that you can use to test it, it's here: https://github.com/sgriffin53/raven/tree/master/temp

Thanks again for all the testing you've done. Hopefully this should fix the bug.

I checked a few positions and the buggy behaviour seems corrected.
To be sure I will play some test games too, but probably after xmas.

I see that your Lichess rating in bullet received some nice +50 already

Great. Thanks a lot. I'm so glad to have this bug finally fixed.