Komodo 7

[lkaufman]

We are planning to release Komodo 7 sometime this week. It will be the same as or nearly the same as the Komodo 7x version now playing in the TCEC final, except that default settings of UCI options were modified for this event. Komodo 7 will be the first in the series with TableBase support (Syzygy). Estimated elo gains (including using 5 man TBs) over Komodo TCEC (the winner of the last TCEC final) are about forty elo points at five minutes plus 3 seconds according to an independent test and a bit more at bullet levels. As always, improvements include both search and eval changes. Price is set at $59.94, or $39.96 if you bought Komodo TCEC. Komodo TCEC will be marked down by $10.

[S.Taylor]

We are planning to release Komodo 7 sometime this week. It will be the same as or nearly the same as the Komodo 7x version now playing in the TCEC final, except that default settings of UCI options were modified for this event. Komodo 7 will be the first in the series with TableBase support (Syzygy). Estimated elo gains (including using 5 man TBs) over Komodo TCEC (the winner of the last TCEC final) are about forty elo points at five minutes plus 3 seconds according to an independent test and a bit more at bullet levels. As always, improvements include both search and eval changes. Price is set at $59.94, or $39.96 if you bought Komodo TCEC. Komodo TCEC will be marked down by $10.

Does it seem to you that it will be stronger than existing komodo and SF versions? or perhaps if SF did the same thing (Syzgy support), then they too would increase 40 elo?

[lkaufman]

Stockfish does already offer a Syzygy version. I think most estimates are that with 5 man TBs this is worth in the five to ten elo range, ten being our estimate. It might be worth more with 6 man TBs, I don't have them to test this.

[S.Taylor]

We are planning to release Komodo 7 sometime this week. It will be the same as or nearly the same as the Komodo 7x version now playing in the TCEC final, except that default settings of UCI options were modified for this event.

What was the nature and purpose of default settings being modified for tcec games? (to make it stronger at long timings?)

[lkaufman]

Drawscore was set to zero since our goal is just to win the match, not to rack up a big score. The default -7 is recommended for playing engines other than Stockfish and Houdini 4.
Table size is increased to 256 (from 64) because we think that is best with long time controls, many cores, and big hash tables. Perhaps 512 would be even better, no practical way to know.

[Ozymandias]

It might be worth more with 6 man TBs, I don't have them to test this.

But 6 men work, I take it.

[lkaufman]

Yes, they work fine and appear to add more elo points, according to limited testing.

[Werewolf]

Are you able to estimate the elo gain excluding Sygyzy tablebases?

If Komodo TCEC and 7 both had limited 5 man Nalimov, how much stronger would 7 be?

[TShackel]

Drawscore was set to zero since our goal is just to win the match, not to rack up a big score. The default -7 is recommended for playing engines other than Stockfish and Houdini 4.
Table size is increased to 256 (from 64) because we think that is best with long time controls, many cores, and big hash tables. Perhaps 512 would be even better, no practical way to know.

Hi Larry. Thanks for the update about Komodo 7.0. I'm glad you brought up the above topic about the draw score. Can you explain a little bit more what that means? Is negative more aggressive in komodo's case. What drawscore favors a draw and what drawscore is more aggressive to attempt to win?

For stockfish the positive contempt values favor more risky play whereas negative values favor draws. This drawscore in Komodo is interesting as I think it's a very drawish engine and anything to make it more aggressive would be interesting.

Thanks Larry.

Tim.

[TShackel]

Oh, and in light of our closed position discussions, do you think komodo 7.0 has any improvement in closed positions?

Thanks.

Tim.