Chess System Tal Extreme, draw rate very low

[chrisw]

In one of CCRL running leagues (15 games each played so far), the average draw rate of all the other programs is about 10-12 draws per 15 games. Eg very few win/loss results.
Compare and contrast Chess System Tal Extreme on the same tournament, only 2 draws in 15 games. 13 are wins/losses.
Only a few games of data but if those stats continue it would not only be remarkable but counter to current theories on the nature of chess.

[Graham Banks]

After the last update. Still pretty good.

[chrisw]

After the last update. Still pretty good.

What stats do you have for other stuff in that Elo range?

[chrisw]

After the last update. Still pretty good.

I count 63 draws in 112 games for the “other” engines = 55% draw rate.
And 3 draws in 16 games = 20% draw rate for Chess System Tal Extreme.

That’s a dramatic difference. Would still be interested to know your stats for that particular division in general.

[Graham Banks]

After the last update. Still pretty good.

What stats do you have for other stuff in that Elo range?

I can do the same for some other engines close in strength once I'm up again.

[Graham Banks]

[Aleks Peshkov]

Nowadays engines subconsciously tuned like or against Stockfish/Leela meta. If Tal style become dominant paradigm other engine authors will prepare countermeasures.

[mclane]

Yes it's mainly incest

[Peter Berger]

Yes it's mainly incest

I don't see how this really makes sense when it is about the corrolaries.

You could always do a Stockfish extreme that starts from a higher ELO starting value and will do similar things in a slightly better way. It is just that the developers haven't bothered yet.

Personally I think that the main reason for the ever higher drawing rates in computerchess are the UHO openings used in all kinds of testing. I understand how they make testing way easier but they also avoid to make people really think about how to get wins from equal starting positions, which they would do if they really had to.

Smyslov used to say: I make 40 good moves, if my opponent does the same, the game will end in a draw. He became world champion, so he clearly had a point. Given his flawless technique (for human standards) he'd probably have enjoyed the competition to be done with UHO openings, too.

There is no doubt that Stockfish is the strongest engine around. It just doesn't care about winning the most games from equal starting positions and/or against weaker opposition as it isn't the competition it is taking part in.

Which is a pity IMHO. CSTal extreme is certainly way more entertaining to watch than "classical" engines of about the same strength.

[mclane]

All those elo list mainly measure similarity.
In the old days of dedicated chess computers e.g. often at the top engines had to play against mephisto amsterdam. Then against mephisto roma68000, but also against roma68020, then there was almeria68000 and almeria68020 etc etc.
So if your novag engine (lets say diablo68000 - kittingers first effort on a 16 bit cpu) for whatever reason is not good against richard lang, you do not have to stand 1 richard lang engine in the „test“ but zillions of langs. All very similar.

And today we mainly see engines similar to stockfish, the strongest open source engine .
And then the engines produce draw scores of 70% ! And people conclude: chess is finished. Stockfish plays like god.
Lets change the game to Go or whatever more complex game.

The closed sources of the early days gave us a big variance of different , really different engines.

Mchess was completely different then Fritz.
And the king was also very different from rebel.
Genius was different from them. And chess tiger was also different.
Gandalf was unique. Cstal had a different approach.
Mephisto III completely unique approach.



These DIFFERENT engines made computerchess interesting.
Today we have a very high draw rate because of ??