Alphazero news

[S.Taylor]

Yes, SF8 on second place

It might not beat the komodo which was out then or a bit later.

Komodo totally dominated SF, 3-4 years ago or more.

[sovaz1997]

Yes, SF8 on second place

It might not beat the komodo which was out then or a bit later.

Komodo totally dominated SF, 3-4 years ago or more.

I am not sure that Komodo was superior to the Stockfish. Have any exact data?

[Guenther]

Yes, SF8 on second place

It might not beat the komodo which was out then or a bit later.

Komodo totally dominated SF, 3-4 years ago or more.

I am not sure that Komodo was superior to the Stockfish. Have any exact data?

I did a quick chart with CCRL data. I hope it is somehow readable. I didn't remember immediately how to fit best
two entities into the same time scale in one graph, when preparing the data.

Note that I omitted the versions with 4cpu because this would be confusing and those who had no ratings in both
categories 40/40 and 40/4 - for the graph I used an average rating of both categories.
Also there are more Komodo entries due to mostly no testing of dev versions of SF.
Last caveat, this is only about Komodo and SF (as the post suggested) and Houdini and Rybka are not represented.

Anyhow one still can conclude that Komodo really was above SF between 2011-08 and 2013-10, while since 2013-11 SF always
was in front of Komodo beginning with the SF DD release.

Yet, I wouldn't call this 'totally dominating' for that period.

https://docs.google.com/spreadsheets/d/ ... O-001ibS9k

http://rwbc-chess.de/pics/chart.png

[Laskos]

Yes, SF8 on second place

It might not beat the komodo which was out then or a bit later.

Komodo totally dominated SF, 3-4 years ago or more.

I am not sure that Komodo was superior to the Stockfish. Have any exact data?

I did a quick chart with CCRL data. I hope it is somehow readable. I didn't remember immediately how to fit best
two entities into the same time scale in one graph, when preparing the data.

Note that I omitted the versions with 4cpu because this would be confusing and those who had no ratings in both
categories 40/40 and 40/4 - for the graph I used an average rating of both categories.
Also there are more Komodo entries due to mostly no testing of dev versions of SF.
Last caveat, this is only about Komodo and SF (as the post suggested) and Houdini and Rybka are not represented.

Anyhow one still can conclude that Komodo really was above SF between 2011-08 and 2013-10, while since 2013-11 SF always
was in front of Komodo beginning with the SF DD release.

Yet, I wouldn't call this 'totally dominating' for that period.

https://docs.google.com/spreadsheets/d/ ... O-001ibS9k

http://rwbc-chess.de/pics/chart.png

TCEC Superfinals? Have to check. Until Lazy SMP was introduced in Sf, Komodo beat systematically Sf.

[Henk]

Too lazy to read this thread. But stockfish 10 unstable in giving scores when evaluating positions during opening phase. Jumping for instance from say 0.2 to 0.8 and back too often. Don't know if that holds for Komodo too. But it is quite annoying.

[Guenther]

Last caveat, this is only about Komodo and SF (as the post suggested) and Houdini and Rybka are not represented.

Added Houdini, Rybka, Fruit and Shredder to the graph...
(Shredder 9-64 data is extrapolated to CCRL from GURL data, because it is not available there)

The graph starts now at early 2005.

[corres]

Too lazy to read this thread. But stockfish 10 unstable in giving scores when evaluating positions during opening phase. Jumping for instance from say 0.2 to 0.8 and back too often. Don't know if that holds for Komodo too. But it is quite annoying.

I propose you to switch off the contempt totally.
For this you should modify the source of Stockfish because dynamic contempt is non-switchable.

[jp]

after 93 games at 5+5, test30 is stable at -70 elo vs SF10 & -21 elo vs SF10+book (perfect2017/2018). will attach pgn's later, going to try to get a decent sample of SF10 vs SF10+book results.

Any more news, yanquis?

[WuShock]

https://www.newyorker.com/science/eleme ... .0-styling

[Leo]

https://www.chess.com/news/view/updated ... game-match

Stockfish 8 ???
They really fear real Stockfish ...

Unfortunately we weren't able to get our time machine up and running before we finished writing the paper in order to test against engines that hadn't been released yet. But don't worry, we are working hard on that!

While I sympathize with that statement, releasing A0 source code and networks for anyone to test sounds better.
Many will not be satisfied with in-house testing with supposedly fair conditions.

That would be good, but it would also be a lot of work for us (AZ is tightly-coupled with DM and Google's systems) for not really much value to the scientific community. We feel that it's our ideas and algorithms that are important, not our implementation. That's why we have published all the algorithms we developed in detail, with almost-runnable pseudo-code, so that they can be replicated easily. And who knows, maybe someone will come up with a better implementation than we did! I am personally really excited about what Leela has already achieved, and hopefully it won't be long before it will be at the level of AlphaZero!

It looks like it will be a long time.