Reply
 
LinkBack Thread Tools Display Modes
  #1   Report Post  
Old December 7th 17, 09:43 AM posted to rec.games.chess.misc
external usenet poster
 
First recorded activity by ChessBanter: Apr 2007
Posts: 3,170
Default Child Prodigy defeats World Champion.

Stockfish lost to AlphaZero, the new kid on the block.

Stockfish scored +0 -28 =72.

In 100 games AlphaZero scored 25 wins and 25 draws with White, while with Black it scored 3 wins and 47 draws.

It didn't lose a game, with the final score 64:36.
  #2   Report Post  
Old December 7th 17, 07:10 PM posted to rec.games.chess.misc
external usenet poster
 
First recorded activity by ChessBanter: Jan 2008
Posts: 569
Default Child Prodigy defeats World Champion.

On Thursday, December 7, 2017 at 1:43:05 AM UTC-7, Offramp wrote:
Stockfish lost to AlphaZero, the new kid on the block.


I thought that the commercial chess engine Komodo was stronger than
Stockfish.

John Savard
  #3   Report Post  
Old December 7th 17, 07:23 PM posted to rec.games.chess.misc
external usenet poster
 
First recorded activity by ChessBanter: Jan 2008
Posts: 569
Default Child Prodigy defeats World Champion.

On Thursday, December 7, 2017 at 11:10:11 AM UTC-7, Quadibloc wrote:
On Thursday, December 7, 2017 at 1:43:05 AM UTC-7, Offramp wrote:
Stockfish lost to AlphaZero, the new kid on the block.


I thought that the commercial chess engine Komodo was stronger than
Stockfish.


The paper referred to Stockfish as the TCEC 2016 champion. I see that in
2017 at the TCEC, Houdini was the champion, and Komodo the runner-up. It
was in 2015 and 2014 that the situation I remembered - Komodo the champ,
and Stockfish the runner-up - existed.

John Savard
  #4   Report Post  
Old December 7th 17, 08:05 PM posted to rec.games.chess.misc
external usenet poster
 
First recorded activity by ChessBanter: Mar 2015
Posts: 32
Default Child Prodigy defeats World Champion.

On 07/12/2017 09:43, Offramp wrote:

Stockfish lost to AlphaZero, the new kid on the block.

Stockfish scored +0 -28 =72.

In 100 games AlphaZero scored 25 wins and 25 draws with White, while
with Black it scored 3 wins and 47 draws.

It didn't lose a game, with the final score 64:36.


Now that AlphaZero is all the rage, it's time to put some things into
perspective:

1. AlphaZero ran on superior hardware which should be worth much more
than a +102 Elo performance. So we cannot say that AlphaZero is the
stronger chess engine.

2. The testing conditions are unclear. I have the suspicion that they
were not fair for Stockfish. Hikaru Nakamura also expressed his doubts,
so it isn't just me.

3. There is nothing new about neural networks or machine learning. Even
AlphaZero's blank slate approach has been tried before. What is new is
the gigantic amount of hardware thrown at the problem.

4. Currently it is not imaginable how AlphaZero could be made to run on
commodity desktop hardware. Perhaps in 10 years GPUs might be fast
enough to run the software. The training phase of the neural network,
however, will require a supercomputer for quite some time to come.

Cheers,
Rainer
  #5   Report Post  
Old December 7th 17, 08:17 PM posted to rec.games.chess.misc
external usenet poster
 
First recorded activity by ChessBanter: Jan 2008
Posts: 569
Default Child Prodigy defeats World Champion.

On Thursday, December 7, 2017 at 12:05:48 PM UTC-7, Rainer wrote:

Now that AlphaZero is all the rage, it's time to put some things into
perspective:


1. AlphaZero ran on superior hardware which should be worth much more
than a +102 Elo performance. So we cannot say that AlphaZero is the
stronger chess engine.


2. The testing conditions are unclear. I have the suspicion that they
were not fair for Stockfish. Hikaru Nakamura also expressed his doubts,
so it isn't just me.


This is indeed quite true.

So it is not accurate to claim that AlphaZero is superior as a chess
program to Stockfish - run a version of that program adapted to use
your computer's hardware, and it would play much worse than Stockfish.

So why is this exciting?

For one thing, Stockfish was evaluating 70 million positions a second,
while AlphaZero was evaluating 80 thousand positions a second.

So AlphaZero somehow obtained a lot more chess knowledge than Stockfish
has. Studying what it came up with as settings for its neural network
would perhaps make it possible to upgrade Stockfish to do what Komodo
is good at: finding good positional moves in quiet positions.

Already, studying Alpha Go has led to improvements in conventional Go-
playing programs - the latest version of CrazyStone gained 100 Elo this
way. This happened by adding "deep learning" techniques to its
development.

John Savard


  #6   Report Post  
Old December 9th 17, 01:35 AM posted to rec.games.chess.misc
external usenet poster
 
First recorded activity by ChessBanter: Jan 2008
Posts: 569
Default Child Prodigy defeats World Champion.

As the success of AlphaZero at chess would spark interest in applying
similar techniques to chess engines running on ordinary hardware, I
thought I'd see what was already out there.

I saw KnightCap and NeuroChess mentioned in an article about
Giraffe.

Giraffe was the best program of this kind, but its development,
understandably, ceased when its programmer was hired for the AlphaGo
team by Google.
  #7   Report Post  
Old December 15th 17, 10:42 AM posted to rec.games.chess.misc
external usenet poster
 
First recorded activity by ChessBanter: Jul 2006
Posts: 81
Default Child Prodigy defeats World Champion.

Bs"d

When Stockfish analyzed 70 million positions a second,
while AlphaZero was evaluating 80 thousand positions a second, that would mean that AlphaZero can run on much more simple hardware.
  #8   Report Post  
Old December 15th 17, 01:15 PM posted to rec.games.chess.misc
external usenet poster
 
First recorded activity by ChessBanter: Mar 2015
Posts: 32
Default Child Prodigy defeats World Champion.

On 15/12/2017 10:42, Eliyahu wrote:

Bs"d

When Stockfish analyzed 70 million positions a second, while
AlphaZero was evaluating 80 thousand positions a second, that would
mean that AlphaZero can run on much more simple hardware.


Quoting from the DeepMind paper:
"AlphaZero uses a markedly different approach that averages over the
position evaluations within a subtree, rather than computing the minimax
evaluation of that subtree."

Comparing Stockfish's 70 million positions searched per second with
AlphaZero's 80 thousand is an exceptionally blatant case of apples and
oranges.

Each of the four TPUs used in the AlphaZero machine has the computing
power of several high-end GPUs.

Cheers,
Rainer
  #9   Report Post  
Old December 15th 17, 02:41 PM posted to rec.games.chess.misc
external usenet poster
 
First recorded activity by ChessBanter: Jan 2008
Posts: 569
Default Child Prodigy defeats World Champion.

On Friday, December 15, 2017 at 2:42:37 AM UTC-7, Eliyahu wrote:

When Stockfish analyzed 70 million positions a second,
while AlphaZero was evaluating 80 thousand positions a second, that would mean that AlphaZero can run on much more simple hardware.


Not necessarily, as AlphaZero may have - and, indeed, must have - subjected
those 80 thousand positions to a much deeper scrutiny.

What that figure means is that AlphaZero has an enormous amount of chess
knowledge. Possibly, the chess knowledge in its neural networks could be
abstracted out to a less slow evaluation function, leading to more powerful
chess programs.

John Savard
Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is Off
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
Sloan v Smith, Petition for Rehearing and Rehearing en banc samsloan rec.games.chess.politics (Chess Politics) 3 December 27th 09 07:42 PM
Deadly, Dangerous, Decisive Damiano's Defense Defeats Drei Denizens During World Open samsloan rec.games.chess.analysis (Chess Analysis) 9 July 30th 07 05:27 AM
Deadly, Dangerous, Decisive Damiano's Defense Defeats Drei Denizens During World Open samsloan rec.games.chess.computer (Computer Chess) 9 July 7th 07 10:36 PM
Deadly, Dangerous, Decisive Damiano's Defense Defeats Drei Denizens During World Open samsloan rec.games.chess.politics (Chess Politics) 10 July 7th 07 10:36 PM
8 year old Child Prodigy can't play in world chess championship because he is BLACK [email protected] rec.games.chess.politics (Chess Politics) 17 November 6th 06 06:18 AM


All times are GMT +1. The time now is 09:41 AM.

Powered by vBulletin® Copyright ©2000 - 2018, Jelsoft Enterprises Ltd.
Copyright 2004-2018 ChessBanter.
The comments are property of their posters.
 

About Us

"It's about Chess"

 

Copyright © 2017