Torch #1 in FRC list

[bastiball]

https://computerchess.org.uk/ccrl/404FR ... 4#Torch_v4

Btw I'll add some games for Stockfish its other opponents too if you guys wondering.

Right now, I can finally say Torch #1 in FRC

[towforce]

Have I missed something? I thought Torch was no longer publicly available, and that it's now an internal chess.com project?

[Uri Blass]

Have I missed something? I thought Torch was no longer publicly available, and that it's now an internal chess.com project?

You missed nothing.
It is not publicly available but they sent it to testers for rating list.

Note sure if it is best in FRC because they tested stockfish17.1 and not latest developement version of Stockfish.

[Hai]

Have I missed something? I thought Torch was no longer publicly available, and that it's now an internal chess.com project?

You missed nothing.
It is not publicly available but they sent it to testers for rating list.

Note sure if it is best in FRC because they tested stockfish17.1 and not latest developement version of Stockfish.

Opponent ELO Diff Results Score LOS Perf
Stockfish 17.1 4072 +12 −12 (−14) 246 − 254 (+56−64=380) 49.2% 246.0 / 500 99.3% −18

Rank Name Rating Score Average Opponent Draws Games LOS
1 Torch v4 4086 +11 −11 65.1% −95.3 61.6% 4100 99.3%
2 Stockfish 17.1 4072 +12 −12 60.3% −63.3 69.8% 2800 100.0%

[AndrewGrant]

Does seem that the result is correct, or at least within reason, with some games I played against 17.1 after the CCRL list was updated.

This was not really an intentional thing. Happy accident.

I don't think there is a lot of FRC data in our nets, and we've certainly not talked about it ever since the project was no longer running. But I guess there must have been some data somewhere, since its well known that with a little effort, you can make leaps in FRC. I was able to establish that with Ethereal's 13.00 release long ago, offering two different nets.

I imagine the SF team will just fix this for SF 18, as they don't like seeing themselves portrayed as anything other than #1 by a large margin. If they don't, then its a cute little asterisk next to Torch.

[Uri Blass]

Does seem that the result is correct, or at least within reason, with some games I played against 17.1 after the CCRL list was updated.

This was not really an intentional thing. Happy accident.

I don't think there is a lot of FRC data in our nets, and we've certainly not talked about it ever since the project was no longer running. But I guess there must have been some data somewhere, since its well known that with a little effort, you can make leaps in FRC. I was able to establish that with Ethereal's 13.00 release long ago, offering two different nets.

I imagine the SF team will just fix this for SF 18, as they don't like seeing themselves portrayed as anything other than #1 by a large margin. If they don't, then its a cute little asterisk next to Torch.

I think they already not number 1 by a big margin in everything related to chess for a long time.

For example they do not seem to be interested to be better than LeelaQueenOdds in beating humans without a queen and it is clear that Leela is better.