Page 1 of 3

IN 100 games if SF reaches 51 in TCEC it should be stopped

Posted: Fri Oct 11, 2019 10:16 pm
by Chessqueen
I do not know if it was luck that AllieStein v0.5-dev_7b41f8c-n11 got a better score than LCO but AS did NOT do as good as LCO against Stockfish 19092522, probably next time around Alliestein with an update might be as strong as Stockfish, unless there is something better than RTX 2080 waiting around the corner. Anyway In 100 games if SF reaches 51 it should be stopped, or they will continue it anyway ? www.tcec-chess.com/

Re: IN 100 games if SF reaches 51 in TCEC it should be stopped

Posted: Fri Oct 11, 2019 10:26 pm
by Graham Banks
Looks like a decisive match victory for Stockfish. That won't sit well with some. :P

Re: IN 100 games if SF reaches 51 in TCEC it should be stopped

Posted: Sat Oct 12, 2019 12:22 am
by daws
AS could still win if SF has 3 crashes but it would be a hollow victory. If we recall, Stockfish had two crashes in the premier division but still finished in the top spot.

Looks like reports of the "end of an era" for AB have been greatly exaggerated.

Re: IN 100 games if SF reaches 51 in TCEC it should be stopped

Posted: Sat Oct 12, 2019 12:59 am
by Dann Corbit
Chessqueen wrote:
Fri Oct 11, 2019 10:16 pm
I do not know if it was luck that AllieStein v0.5-dev_7b41f8c-n11 got a better score than LCO but AS did NOT do as good as LCO against Stockfish 19092522, probably next time around Alliestein with an update might be as strong as Stockfish, unless there is something better than RTX 2080 waiting around the corner. Anyway In 100 games if SF reaches 51 it should be stopped, or they will continue it anyway ? www.tcec-chess.com/
From TCEC 16 Rules and Information:
"Superfinal

The Superfinal consists of 100 games at TC 120+10, with 50 different openings, among them once the normal start position, so that each engine plays both black and white of the same opening position. The match will be presented with opening 1 used in games 1 and 2, then opening 2 used in games 3 and 4 etc.
If the match is theoretically won for one side before game 100, the match will still continue until all 100 games have been played."

SF has already won 51 games, and they are playing on, so the rules are being followed.
I like that, because the games produce really interesting data.

Re: IN 100 games if SF reaches 51 in TCEC it should be stopped

Posted: Sat Oct 12, 2019 2:54 am
by jp


SF has this position as White. +2.75! :) Maybe it needs to consult Honey. :wink:
AS has 0.08.

Re: IN 100 games if SF reaches 51 in TCEC it should be stopped

Posted: Sat Oct 12, 2019 9:26 am
by Zenmastur
daws wrote:
Sat Oct 12, 2019 12:22 am
AS could still win if SF has 3 crashes but it would be a hollow victory. If we recall, Stockfish had two crashes in the premier division but still finished in the top spot.

Looks like reports of the "end of an era" for AB have been greatly exaggerated.
In case you were in doubt, simply look at the number of outright blunders made by AS. A good example is move 33. ... Qc6???



Then it compounds the mistake by playing 34. ... Nd6???

Until they figure out how to stop NN engines from making moves like this, there is little hope of a new era, just a few interesting games.

Regards,

Zenmastur

Re: IN 100 games if SF reaches 51 in TCEC it should be stopped

Posted: Sat Oct 12, 2019 12:59 pm
by Ovyron
Zenmastur wrote:
Sat Oct 12, 2019 9:26 am
In case you were in doubt, simply look at the number of outright blunders made by AS. A good example is move 33. ... Qc6???

<snip>

Then it compounds the mistake by playing 34. ... Nd6???
I guess the questions are:

1.Does Leela play those moves? If not it's clear Leela should have gone to the final instead of Allie, which would indicate Allie is just better at beating weaker opposition.

2.Has this just been because of improvements by Stockfish? If so, NNs rapidly improved and eventually caught up and surpassed Stockfish's strength, but they plateaued* while Stockfish kept improving, so this would be THE REVIVAL OF AN ERA.

(*I reckon this is the first time in my life I use a word with 4 vowels in a row)

Re: IN 100 games if SF reaches 51 in TCEC it should be stopped

Posted: Sat Oct 12, 2019 3:18 pm
by DustyMonkey
monte carlo tree search produces blunders by design

while the root move may be "well sampled" and avoid hanging pieces,
the 2nd ply is less well sampled,
the 3rd ply even less well sampled,
the 4th ply poorly sampled,
the 5th ply even more poorly sampled,
...

This is the core fact of monte carlo tree search. I know some folks will jump in trying to deny it, or excuse it, or hand wave the problem "away" ...

The best argument trying to defend monte carlo on the matter is that the blunders aren't shallow, but it ignores the obvious. The game is more than one move. Giving up a little bit on each move, with high probability, compounds, again and again, on every move through a game. You can say "on move 20 the odds of MCTS blundering is low!" ... but you cannot say "over all moves of a game, the odds of MCTS blundering is low!" .. its not factual. At all.

Re: IN 100 games if SF reaches 51 in TCEC it should be stopped

Posted: Sat Oct 12, 2019 3:29 pm
by Zenmastur
Ovyron wrote:
Sat Oct 12, 2019 12:59 pm
Zenmastur wrote:
Sat Oct 12, 2019 9:26 am
In case you were in doubt, simply look at the number of outright blunders made by AS. A good example is move 33. ... Qc6???

<snip>

Then it compounds the mistake by playing 34. ... Nd6???
I guess the questions are:

1.Does Leela play those moves? If not it's clear Leela should have gone to the final instead of Allie, which would indicate Allie is just better at beating weaker opposition.

2.Has this just been because of improvements by Stockfish? If so, NNs rapidly improved and eventually caught up and surpassed Stockfish's strength, but they plateaued* while Stockfish kept improving, so this would be THE REVIVAL OF AN ERA.

(*I reckon this is the first time in my life I use a word with 4 vowels in a row)
The fact that ANY top NN angine could produce two MEGA-BLUNDERS in a row is sufficient evidence that we are not now, nor were we in the recent past in a "new era". Their technology isn't mature enough yet!

1.) I don't think what Leela would have done has any bearing on the subject. It came in third or fourth if I recall.

2.) SF hasn't made any great gains (just "normal" albeit slow progress) in playing strength in the recent past or so this site would have me believe.
https://nextchessmove.com/dev-builds

Regards,

Zenmastur

Re: IN 100 games if SF reaches 51 in TCEC it should be stopped

Posted: Sat Oct 12, 2019 4:14 pm
by Ovyron
Are you saying that Leela winning Season 15 was a fluke?