Computer Chess Worlds (unofficial)

[Dubslow]

The most widely respected computer chess tournament in the world runs nearly continuously, as the current season has been for several months.

http://tcec.chessdom.com/live.php

In the grand finals, Stockfish 8 has a 6 point lead on Houdini with a 10-4-47 record, game 62 in progress, and 38 remaining to be played.

Notably, every single victory by either engine has been as white. Black hasn't managed anything more than a draw.

I think the forum's game against SF 7 is doomed... well maybe with a recent Komodo they can do well... hmm now I'm second guessing myself.

[Batalov]

There was an interesting casus there

Game 17 of the Superfinal has ended with a TB position win for white. The fact that the 50 moves rule would have entered in effect has caused considerable debate on the topic. Both engine authors have submitted their opinion to the TCEC team. Multiple experts have also given their thoughts on the topic.

The score of game 17 is now under investigation and the final decision by Anton Mihailov will come next week. The decision is taking time, as the primary goal as usual is the fairness and equal conditions for the participating engines.

Replay game 17 here http://tcec.chessdom.com/archive.php?se=9&sf&ga=17

[axn]

Originally Posted by Dubslow

Notably, every single victory by either engine has been as white. Black hasn't managed anything more than a draw.

That's because, the engines are not playing from move 1 but instead from move 'n', where the "book" moves are supplied by an IM. These have been selected to give white a starting advantage

Originally Posted by Dubslow

I think the forum's game against SF 7 is doomed... well maybe with a recent Komodo they can do well...

Not in the least bit. The game conditions are completely different. Most notably, SF is not using a TB. Forum is doing quite well, I think,

[Dubslow]

Originally Posted by axn

That's because, the engines are not playing from move 1 but instead from move 'n', where the "book" moves are supplied by an IM. These have been selected to give white a starting advantage

I realized they were using presupplied openings, but you're saying they're deliberately using suboptimal openings for black?

Edit: This page says to check the "Opening Book" link above to learn more about how they're "prepared", but in line with the rest of TCEC's online presence where contextual information beyond the games themselves are difficult to find, that link doesn't actually exist. Do you know the source?

It does seem like they're doctored against black though, because they all start a >.5 for white, which contrasts the standard ~.3 advantage white usually starts with...

[axn]

Originally Posted by Dubslow

Edit: This page says to check the "Opening Book" link above to learn more about how they're "prepared", but in line with the rest of TCEC's online presence where contextual information beyond the games themselves are difficult to find, that link doesn't actually exist. Do you know the source?

Go to the live page (http://tcec.chessdom.com/live.php), click on the "Opening Book" button at the top.

Originally Posted by Dubslow

It does seem like they're doctored against black though, because they all start a >.5 for white, which contrasts the standard ~.3 advantage white usually starts with...

That's what I mean. They've done this so that there are at least some non-draws (3hr / side, 44 cores, 6-man TB = drawfest).

[LaurV]

Yes, that is true, they are "doctored", but it doesn't matter as the engines play both black and white games from the same position. This is the best way to see if one engine is better that the other. If you start 'clean', engines with close strength will always draw, even if their strength is not the same. With this method they can differentiate the engines better. The stronger engine will draw when play black (if the position's unbalance is not exaggerated), and win when play white.

[MooMoo2]

Originally Posted by Dubslow

Notably, every single victory by either engine has been as white. Black hasn't managed anything more than a draw.

That matches what's going on in correspondence chess games. No pre-selected openings there:
http://serverchess.blogspot.com/2015...m-is-long.html

I think the forum's game against SF 7 is doomed...

Agreed!

[gd_barnes]

Originally Posted by MooMoo2

That matches what's going on in correspondence chess games. No pre-selected openings there:
http://serverchess.blogspot.com/2015...m-is-long.html


Agreed!

Why do you agree that the forum is doomed in our correspondence game against SF 7? Komodo 10, the strongest program in the world outside of SF (and perhaps even slightly stronger), shows the current position as 0.10 to 0.20 in favor of white...slightly better for black than white's typical opening advantage of 0.25 to 0.30.

If by doomed you mean that the game is doomed to a draw then I won't disagree but there is still plenty of chess left to play. We have not yet attempted to put any of our engines into a multi-core deep think mode yet.

[MooMoo2]

Originally Posted by gd_barnes

Why do you agree that the forum is doomed in our correspondence game against SF 7? Komodo 10, the strongest program in the world outside of SF (and perhaps even slightly stronger), shows the current position as 0.10 to 0.20 in favor of white...slightly better for black than white's typical opening advantage of 0.25 to 0.30.

If by doomed you mean that the game is doomed to a draw then I won't disagree but there is still plenty of chess left to play.

I don't think black will lose; I meant that the game is doomed to a draw. From looking at the statistics of various correspondence chess and high level engine vs. engine matches (like TCEC), it's very rare for black to win. The side playing black can't be just a bit stronger than white, it needs to be strong enough to overcome white's initial opening advantage and gain enough of an advantage to checkmate white (for example, black may be strong enough to have a king and bishop against a lone white king, but that isn't enough to win).

[Dubslow]

Originally Posted by axn

Go to the live page (http://tcec.chessdom.com/live.php), click on the "Opening Book" button at the top.


That's what I mean. They've done this so that there are at least some non-draws (3hr / side, 44 cores, 6-man TB = drawfest).

Ah, makes sense I guess. Thanks for pointing out the opening book link. Although it discusses season 8, I'm reasonably certain they've made only minor tweaks to the general setup for season 9 (e.g. some matches start with a 9 move opening instead of 8).

Originally Posted by axn

Not in the least bit. The game conditions are completely different. Most notably, SF is not using a TB. Forum is doing quite well, I think,

As discussed elsewhere, the stated goal for the forum game is to beat Stockfish as black, and that is doomed I believe. I agree that the forum is very comfortably on track to secure a draw, but a draw would fail to meet the stated goal of the game.

[LaurV]

Originally Posted by Dubslow

As discussed elsewhere, the stated goal for the forum game is to beat Stockfish as black, and that is doomed I believe. I agree that the forum is very comfortably on track to secure a draw, but a draw would fail to meet the stated goal of the game.

We agree with that, 101%. However we still want to play that game. It is stimulating for us, even if now the engine does most of the work.

In a future game we have to play better, and use Rybka in an earlier stage.