Has CCRl test AllieStein YET ?

[Modern Times]

It is pretty simple really. Let me spell it out for you. When testing of v0.4 began, v0.5 did not exist.

[Chessqueen]

It is pretty simple really. Let me spell it out for you. When testing of v0.4 began, v0.5 did not exist.

It is pretty simple really, a buggy version ends up behind Houdini and both Komodos, Is it NOT obvious why Buggy Version 0.4 ended in such place as it did with your test. I wonder why Version 0.5 is leading, let me see probably because it is NOT Buggy like version 0.4. Of course there are only a few games played so far, and of course a few games also apply to Houdini and both versions of Komodo ==> https://tcec.chessdom.com/

[Modern Times]

It ended up behind Komodo and Houdini primarily because it has not been trained on chess960, and that cost it as much as -100 Elo. It would have done better at standard chess.

[Chessqueen]

It ended up behind Komodo and Houdini primarily because it has not been trained on chess960, and that cost it as much as -100 Elo. It would have done better at standard chess.

To train AllieStein in Chess960 would take several months if NOT years since there are so many opening that can come out of 960 possible positions instead of just one like in standard chess, but anyway so far it is the only NN engine that can play Chess960. I wonder that special code Stockfish has that performs so well in Chess960 ?
https://tcec.chessdom.com/