The victory of Stockfish 14 over Lc0 69626 is fully deserved.
The concreteness of play through precise processes of calculation of the variants, which the Alpha-Beta algorithm allows, is better for sure than the "fuzzy" and "promising" lines as Lco is able to calculate through PUCT algorithm.
In any case, the final result has determined by openings that has scarce strategic value: too many openings had imbalanced tactics just at the exit of the book lines.
The final result 19 to 7 is too liar as regard the actual value of the two programs, even regardless of the Lco ghastly time management. (Stockfish time management was fantastic, he never played the first move out of the book too quickly)
During the training of AlphaZero, and presumibly Lc0 too, we know from Nature paper that Sicilian is immediately discarded, so Lco has so little experience of autoplay in this opening.
I am confident about one thing. In case of as many Spanish as the Sicilians had been played (how many were played? I counted only 2 Tchigorin against about 30 Sicilians) the final result would be tight.
Or, more trivially, if Sadler and Noomen had selected a number of 1 e4-e5 openings equal to the number of 1 e4-c5 openings we would not have witnessed this massacre.
I think that It is not fair that openings with a rich historical tradition such as
- Spanish (2)
- Scottish (0)
- Russian (0)
- Italian (0)
have in chess a value equal to about 1/15 of the value attributed by Sadler and Noomen to the Sicilian opening.
There has never been a TCEC so biased in favor of Stockfish. Crowds of Dutch and East Indians we are witnessed, but not even a West Indian that we know is the opening that Alpha Zero and presumibly LC0 plays with white at the best.
Will we see, in the future, in TCEC22, a more balanced choice of 1 e4-e5 versus 1 e4-c5?
TCEC21 is over
Moderators: hgm, Rebel, chrisw
-
- Posts: 389
- Joined: Tue Oct 08, 2019 11:39 pm
- Full name: Tomasz Sobczyk
Re: TCEC21 is over
The openings were chosen randomly. It's natural that a tournament appears biased towards the better player.
dangi12012 wrote:No one wants to touch anything you have posted. That proves you now have negative reputations since everyone knows already you are a forum troll.
Maybe you copied your stockfish commits from someone else too?
I will look into that.
-
- Posts: 2129
- Joined: Thu May 29, 2008 10:43 am
Re: TCEC21 is over
I believe the superfinal used just 50 positions (in sequence) for the 100 games.
https://tcec-chess.com/articles/TCEC_Openings_FAQ.html
ECO code distribution
ECO A: 10 lines
ECO B: 16 lines
ECO C: 10 lines
ECO D: 8 lines
ECO E: 6 lines
Move length distribution
3 moves: 5 lines
4 moves: 7 lines
5 moves: 4 lines
6 moves: 12 lines
7 moves: 2 lines
8 moves: 7 lines
9 moves: 5 lines
10 moves: 3 lines
>10 moves: 5 lines
http://blogchess2016.blogspot.com/2021/ ... by-gm.html
Personally, I would like see less human interference...give the engines more say in the opening.
(>10 moves?) that's taking a huge chunk of the game away from the engines...
-
- Posts: 1142
- Joined: Thu Dec 28, 2017 4:06 pm
- Location: Argentina
Re: TCEC21 is over
They weren´t random at all.
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls
-
- Posts: 1142
- Joined: Thu Dec 28, 2017 4:06 pm
- Location: Argentina
Re: TCEC21 is over
FRC is the future. No bias, just random.Kanizsa wrote: ↑Sun Aug 08, 2021 1:16 pm The victory of Stockfish 14 over Lc0 69626 is fully deserved.
The concreteness of play through precise processes of calculation of the variants, which the Alpha-Beta algorithm allows, is better for sure than the "fuzzy" and "promising" lines as Lco is able to calculate through PUCT algorithm.
In any case, the final result has determined by openings that has scarce strategic value: too many openings had imbalanced tactics just at the exit of the book lines.
The final result 19 to 7 is too liar as regard the actual value of the two programs, even regardless of the Lco ghastly time management. (Stockfish time management was fantastic, he never played the first move out of the book too quickly)
During the training of AlphaZero, and presumibly Lc0 too, we know from Nature paper that Sicilian is immediately discarded, so Lco has so little experience of autoplay in this opening.
I am confident about one thing. In case of as many Spanish as the Sicilians had been played (how many were played? I counted only 2 Tchigorin against about 30 Sicilians) the final result would be tight.
Or, more trivially, if Sadler and Noomen had selected a number of 1 e4-e5 openings equal to the number of 1 e4-c5 openings we would not have witnessed this massacre.
I think that It is not fair that openings with a rich historical tradition such as
- Spanish (2)
- Scottish (0)
- Russian (0)
- Italian (0)
have in chess a value equal to about 1/15 of the value attributed by Sadler and Noomen to the Sicilian opening.
There has never been a TCEC so biased in favor of Stockfish. Crowds of Dutch and East Indians we are witnessed, but not even a West Indian that we know is the opening that Alpha Zero and presumibly LC0 plays with white at the best.
Will we see, in the future, in TCEC22, a more balanced choice of 1 e4-e5 versus 1 e4-c5?
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls
-
- Posts: 89
- Joined: Sat Sep 13, 2014 4:12 pm
- Location: Zagreb, Croatia
- Full name: Branko Radovanović
Re: TCEC21 is over
What would you consider "random"?
-
- Posts: 1142
- Joined: Thu Dec 28, 2017 4:06 pm
- Location: Argentina
Re: TCEC21 is over
Picking from a book?Branko Radovanovic wrote: ↑Mon Aug 09, 2021 9:27 pmWhat would you consider "random"?
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls
-
- Posts: 883
- Joined: Sat Mar 13, 2021 1:47 am
- Full name: Amanj Sherwany
Re: TCEC21 is over
I don't think its the variant that matters here. If you use a book, then you have some influence on the result... If we are only interested in finding the strongest engine, then matches should be bookless, if we want to also enjoy the games then it is debatableCMCanavessi wrote: ↑Sun Aug 08, 2021 8:11 pmFRC is the future. No bias, just random.Kanizsa wrote: ↑Sun Aug 08, 2021 1:16 pm The victory of Stockfish 14 over Lc0 69626 is fully deserved.
The concreteness of play through precise processes of calculation of the variants, which the Alpha-Beta algorithm allows, is better for sure than the "fuzzy" and "promising" lines as Lco is able to calculate through PUCT algorithm.
In any case, the final result has determined by openings that has scarce strategic value: too many openings had imbalanced tactics just at the exit of the book lines.
The final result 19 to 7 is too liar as regard the actual value of the two programs, even regardless of the Lco ghastly time management. (Stockfish time management was fantastic, he never played the first move out of the book too quickly)
During the training of AlphaZero, and presumibly Lc0 too, we know from Nature paper that Sicilian is immediately discarded, so Lco has so little experience of autoplay in this opening.
I am confident about one thing. In case of as many Spanish as the Sicilians had been played (how many were played? I counted only 2 Tchigorin against about 30 Sicilians) the final result would be tight.
Or, more trivially, if Sadler and Noomen had selected a number of 1 e4-e5 openings equal to the number of 1 e4-c5 openings we would not have witnessed this massacre.
I think that It is not fair that openings with a rich historical tradition such as
- Spanish (2)
- Scottish (0)
- Russian (0)
- Italian (0)
have in chess a value equal to about 1/15 of the value attributed by Sadler and Noomen to the Sicilian opening.
There has never been a TCEC so biased in favor of Stockfish. Crowds of Dutch and East Indians we are witnessed, but not even a West Indian that we know is the opening that Alpha Zero and presumibly LC0 plays with white at the best.
Will we see, in the future, in TCEC22, a more balanced choice of 1 e4-e5 versus 1 e4-c5?
-
- Posts: 1142
- Joined: Thu Dec 28, 2017 4:06 pm
- Location: Argentina
Re: TCEC21 is over
If you play from the start position you get 100 draws. That's how it is.amanjpro wrote: ↑Mon Aug 09, 2021 10:24 pm I don't think its the variant that matters here. If you use a book, then you have some influence on the result... If we are only interested in finding the strongest engine, then matches should be bookless, if we want to also enjoy the games then it is debatable
I do agree, however, that using man-made 14 moves openings are quite stupid. That's like making the engines only play the endgame, and probably lines that the engines themselves would never ever play.
Follow my tournament and some Leela gauntlets live at http://twitch.tv/ccls
-
- Posts: 883
- Joined: Sat Mar 13, 2021 1:47 am
- Full name: Amanj Sherwany
Re: TCEC21 is over
Well, put Stockfish against Zahak, both bookless, we will see if we get 100 draws...CMCanavessi wrote: ↑Mon Aug 09, 2021 10:46 pmIf you play from the start position you get 100 draws. That's how it is.amanjpro wrote: ↑Mon Aug 09, 2021 10:24 pm I don't think its the variant that matters here. If you use a book, then you have some influence on the result... If we are only interested in finding the strongest engine, then matches should be bookless, if we want to also enjoy the games then it is debatable
I do agree, however, that using man-made 14 moves openings are quite stupid. That's like making the engines only play the endgame, and probably lines that the engines themselves would never ever play.
Thing is, these two monsters are almost unbeatable without human interference, so yeah I agree, bookless matches are probably going to be a drawfest.
But I enjoy TCEC that way better than a hypothetical bookless version, I don't care who wins, I care about seeing a chess that I might enjoy (well, to be honest I don't think I enjoy their games that much anyways, I'm too weak a player to understand their games)