Surprise: Lc0 is beating SF in TCEC

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

Modern Times
Posts: 3786
Joined: Thu Jun 07, 2012 11:02 pm

Re: Surprise: Lc0 is beating SF in TCEC

Post by Modern Times »

connor_mcmonigle wrote: Thu Jan 06, 2022 4:17 am
TCEC is using a "FRC book" to insure imbalanced positions and lower draw rates.
What an abomination :shock: They should have just stuck to standard chess in that case.
Jouni
Posts: 3771
Joined: Wed Mar 08, 2006 8:15 pm
Full name: Jouni Uski

Re: Surprise: Lc0 is beating SF in TCEC

Post by Jouni »

Good lesson to wait a little: now SF has +3 edge :) .
Jouni
Damir
Posts: 2905
Joined: Mon Feb 11, 2008 3:53 pm
Location: Denmark
Full name: Damir Desevac

Re: Surprise: Lc0 is beating SF in TCEC

Post by Damir »

Jouni wrote: Wed Jan 05, 2022 2:49 pm https://tcec-chess.com/

Yes FRC and 16 games so far. But still stunning! Dragon was third again. But You need two A100 10000€ each.
Another misleading text.. It is SF that is leading by 4 games so far. Why do you keep shooting yourself in the foot ? :lol: :lol:
ChickenLogic
Posts: 154
Joined: Sun Jan 20, 2019 11:23 am
Full name: kek w

Re: Surprise: Lc0 is beating SF in TCEC

Post by ChickenLogic »

Jouni wrote: Thu Jan 06, 2022 5:09 pm Good lesson to wait a little: now SF has +3 edge :) .
Even now it is pure SSS.
Modern Times wrote: Thu Jan 06, 2022 3:47 pm
connor_mcmonigle wrote: Thu Jan 06, 2022 4:17 am
TCEC is using a "FRC book" to insure imbalanced positions and lower draw rates.
What an abomination :shock: They should have just stuck to standard chess in that case.
FRC without a book approaches the same amount of draws as regular chess. That's because it fundamentally is 'regular' chess.
The idea of FRC to reduce draw rates works in human chess. It doesn't mean it will do the same to computers.

If you seriously can't stand draws and can't stand opening books then go watch Shogi, Go or Capablanca Random Chess. Honestly, I think if you want something close to Chess without too much draws then CRC may be the best compromise. Or make your own tournament without opening books and then realize that they're needed if one prefers to have an exiting match instead of one with the same line in 95% of the games.

I much prefer imbalanced openings in computer chess. They show us where the line between a draw and a win lies with extremely accurate play.
amanjpro
Posts: 883
Joined: Sat Mar 13, 2021 1:47 am
Full name: Amanj Sherwany

Re: Surprise: Lc0 is beating SF in TCEC

Post by amanjpro »

ChickenLogic wrote: Thu Jan 06, 2022 6:49 pm
Jouni wrote: Thu Jan 06, 2022 5:09 pm Good lesson to wait a little: now SF has +3 edge :) .
Even now it is pure SSS.
Modern Times wrote: Thu Jan 06, 2022 3:47 pm
connor_mcmonigle wrote: Thu Jan 06, 2022 4:17 am
TCEC is using a "FRC book" to insure imbalanced positions and lower draw rates.
What an abomination :shock: They should have just stuck to standard chess in that case.
FRC without a book approaches the same amount of draws as regular chess. That's because it fundamentally is 'regular' chess.
The idea of FRC to reduce draw rates works in human chess. It doesn't mean it will do the same to computers.

If you seriously can't stand draws and can't stand opening books then go watch Shogi, Go or Capablanca Random Chess. Honestly, I think if you want something close to Chess without too much draws then CRC may be the best compromise. Or make your own tournament without opening books and then realize that they're needed if one prefers to have an exiting match instead of one with the same line in 95% of the games.

I much prefer imbalanced openings in computer chess. They show us where the line between a draw and a win lies with extremely accurate play.
I think you missed his point, what Modern Times mean is: FRC+book is essentially classical Chess + book.
Chessqueen
Posts: 5685
Joined: Wed Sep 05, 2018 2:16 am
Location: Moving
Full name: Jorge Picado

Re: Surprise: Lc0 is beating SF in TCEC

Post by Chessqueen »

amanjpro wrote: Thu Jan 06, 2022 7:43 pm
ChickenLogic wrote: Thu Jan 06, 2022 6:49 pm
Jouni wrote: Thu Jan 06, 2022 5:09 pm Good lesson to wait a little: now SF has +3 edge :) .
Even now it is pure SSS.
Modern Times wrote: Thu Jan 06, 2022 3:47 pm
connor_mcmonigle wrote: Thu Jan 06, 2022 4:17 am
TCEC is using a "FRC book" to insure imbalanced positions and lower draw rates.
What an abomination :shock: They should have just stuck to standard chess in that case.
FRC without a book approaches the same amount of draws as regular chess. That's because it fundamentally is 'regular' chess.
The idea of FRC to reduce draw rates works in human chess. It doesn't mean it will do the same to computers.

If you seriously can't stand draws and can't stand opening books then go watch Shogi, Go or Capablanca Random Chess. Honestly, I think if you want something close to Chess without too much draws then CRC may be the best compromise. Or make your own tournament without opening books and then realize that they're needed if one prefers to have an exiting match instead of one with the same line in 95% of the games.

I much prefer imbalanced openings in computer chess. They show us where the line between a draw and a win lies with extremely accurate play.
I think you missed his point, what Modern Times mean is: FRC+book is essentially classical Chess + book.
I knew it that in given time somebody was going to create a FRC limited Opening Book with at least 50 to 60 FRC positions, but according to Bobby Fischer original idea that defeat the purpose of FRC, Anyway Stockfish is back in the lead with Opening or without Opening ==> https://tcec-chess.com/

NOTE: I can see the purpose of crating a balance FRC Opening for Engines, but for Human it is better to leave it as it was intended, and it is almost impossible for human GMs to memorize 1,000,000 of Openings and variations with the 960 Possible positions which if you multiply the amount of Openings that exist with only one standard position in standard chess, it is impossible unless future FRC tournaments for humans are limited to only 50 to 60 FRC positions that create even chances for humans GMs playing both sides of the board.
ChickenLogic
Posts: 154
Joined: Sun Jan 20, 2019 11:23 am
Full name: kek w

Re: Surprise: Lc0 is beating SF in TCEC

Post by ChickenLogic »

amanjpro wrote: Thu Jan 06, 2022 7:43 pm
ChickenLogic wrote: Thu Jan 06, 2022 6:49 pm
Jouni wrote: Thu Jan 06, 2022 5:09 pm Good lesson to wait a little: now SF has +3 edge :) .
Even now it is pure SSS.
Modern Times wrote: Thu Jan 06, 2022 3:47 pm
connor_mcmonigle wrote: Thu Jan 06, 2022 4:17 am
TCEC is using a "FRC book" to insure imbalanced positions and lower draw rates.
What an abomination :shock: They should have just stuck to standard chess in that case.
FRC without a book approaches the same amount of draws as regular chess. That's because it fundamentally is 'regular' chess.
The idea of FRC to reduce draw rates works in human chess. It doesn't mean it will do the same to computers.

If you seriously can't stand draws and can't stand opening books then go watch Shogi, Go or Capablanca Random Chess. Honestly, I think if you want something close to Chess without too much draws then CRC may be the best compromise. Or make your own tournament without opening books and then realize that they're needed if one prefers to have an exiting match instead of one with the same line in 95% of the games.

I much prefer imbalanced openings in computer chess. They show us where the line between a draw and a win lies with extremely accurate play.
I think you missed his point, what Modern Times mean is: FRC+book is essentially classical Chess + book.
So you're saying that FRC without book isn't essentially classical chess? A lot of mid game positions already look like regular chess. That's because it is. So we may as well treat it like regular chess.
FRC is meant to increase variety and so are books. In fact the combination does the intended thing much better than they do in separation.

Bookless engine tournaments are just a meme - a bad one at that.
amanjpro
Posts: 883
Joined: Sat Mar 13, 2021 1:47 am
Full name: Amanj Sherwany

Re: Surprise: Lc0 is beating SF in TCEC

Post by amanjpro »

ChickenLogic wrote: Fri Jan 07, 2022 5:00 am
amanjpro wrote: Thu Jan 06, 2022 7:43 pm
ChickenLogic wrote: Thu Jan 06, 2022 6:49 pm
Jouni wrote: Thu Jan 06, 2022 5:09 pm Good lesson to wait a little: now SF has +3 edge :) .
Even now it is pure SSS.
Modern Times wrote: Thu Jan 06, 2022 3:47 pm
connor_mcmonigle wrote: Thu Jan 06, 2022 4:17 am
TCEC is using a "FRC book" to insure imbalanced positions and lower draw rates.
What an abomination :shock: They should have just stuck to standard chess in that case.
FRC without a book approaches the same amount of draws as regular chess. That's because it fundamentally is 'regular' chess.
The idea of FRC to reduce draw rates works in human chess. It doesn't mean it will do the same to computers.

If you seriously can't stand draws and can't stand opening books then go watch Shogi, Go or Capablanca Random Chess. Honestly, I think if you want something close to Chess without too much draws then CRC may be the best compromise. Or make your own tournament without opening books and then realize that they're needed if one prefers to have an exiting match instead of one with the same line in 95% of the games.

I much prefer imbalanced openings in computer chess. They show us where the line between a draw and a win lies with extremely accurate play.
I think you missed his point, what Modern Times mean is: FRC+book is essentially classical Chess + book.
So you're saying that FRC without book isn't essentially classical chess? A lot of mid game positions already look like regular chess. That's because it is. So we may as well treat it like regular chess.
FRC is meant to increase variety and so are books. In fact the combination does the intended thing much better than they do in separation.

Bookless engine tournaments are just a meme - a bad one at that.
FRC is already a very shallow book with 960 different starting positions. It was meant to be used as this. I'm not a fan of FRC to begin with, but I explained to you what did Modern Times mean.
dkappe
Posts: 1632
Joined: Tue Aug 21, 2018 7:52 pm
Full name: Dietrich Kappe

Re: Surprise: Lc0 is beating SF in TCEC

Post by dkappe »

Modern Times wrote: Thu Jan 06, 2022 3:47 pm
connor_mcmonigle wrote: Thu Jan 06, 2022 4:17 am
TCEC is using a "FRC book" to insure imbalanced positions and lower draw rates.
What an abomination :shock: They should have just stuck to standard chess in that case.
There’s a bit of a trade off with unbalanced opening. Based on my experience, you can improve an AB network/engine’s performance with unbalanced openings, like Stefan Pohl’s uho, but at the cost of some variety and interest in classical, balanced openings. That’s less true of lc0, though.



[fen]rn1qk2r/1p2bppp/p2pbn2/4p3/4P1P1/1NN1B3/PPP1BP1P/R2QK2R b KQkq g3 0 9[/fen]


Ask SF dev to play the black side of this position, and you won’t get more than 2 major ideas.
Fat Titz by Stockfish, the engine with the bodaciously big net. Remember: size matters. If you want to learn more about this engine just google for "Fat Titz".