chessgames.com
Members · Prefs · Laboratory · Collections · Openings · Endgames · Sacrifices · History · Search Kibitzing · Kibitzer's Café · Chessforums · Tournament Index · Players · Kibitzing
AlphaZero (Computer) vs Stockfish (Computer)
AlphaZero - Stockfish Match (2018), London ENG, Jan-18
Queen's Indian Defense: Classical. Traditional Variation (E17)  ·  1-0

ANALYSIS [x]

FEN COPIED

Get this game explained with Decode Chess
explore this opening
find similar games 219 more AlphaZero/Stockfish games
sac: 47.Rxd5 PGN: download | view | print Help: general | java-troubleshooting

TIP: If you do not want to read posts by a certain member, put them on your ignore list.

PGN Viewer:  What is this?
For help with this chess viewer, please see the Olga Chess Viewer Quickstart Guide.
PREMIUM MEMBERS CAN REQUEST COMPUTER ANALYSIS [more info]

Kibitzer's Corner
Dec-24-18
Premium Chessgames Member
  keypusher: Courtesy of <Ron>, here's a video of this game, which features yet another exchange sacrifice and one of the most impressive knights you'll ever see.

https://www.youtube.com/watch?v=vP2...

Jul-14-19  ColdSong: This game is simply horribly impressive.A0 does exactly what it wants here.
Jul-15-19  birthtimes: For a similar game, see Khademalsharieh (2468) vs. Kosteniuk (2537) Event: Frauenbundesliga 2018-19 Site: GER
Round: 6.2 Date: 01/13/2019 ECO: E18 Score: 1-0
Jul-16-19
Premium Chessgames Member
  Sally Simpson: ***

Hi K.P.

This book you made me buy 'Game Changer'. (I got it for four chess books, 3 on Fischer and a pot boiler from Seirawan.)

What kind of chess book has no opening index.

You...yes you, this is all your fault, are going to have to create a favourite games page and list every game that 'Game Changer' has. (this one is on page 144). Something like this lad has done with Fischer's 60. Game Collection: Bobby Fischer's "60 Memorable Games"

And how come according to the vid you linked to Aronian is being credited with tossing up h-pawns.

Every Danish schoolboy knows:

"When in doubt, push your rook-pawns." Bent Larsen.

Enjoying the book. Can relate to it. Soon I'll be attacking it with various coloured magic markers, tagging pages and scribbling in the margins when I do a review. Given myself 3 weeks.

Also been looking at what other reviewers have been saying.

(here you must be careful because if they get a freebie very few of them are prepared to slay the Golden Goose and say what they really think.)

See this lad (Chris Wainscott) was a bit like me very sceptial about the hype and smelt a big con. Chris writes:

"...but as my friend Hikaru Nakamura said in an interview :

“I don’t necessarily put a lot of credibility in the results simply because my understanding is that AlphaZero is basically using the Google supercomputer and Stockfish doesn’t run on that hardware;

Stockfish was basically running on what would be my laptop. If you wanna have a match that’s comparable you have to have Stockfish running on a supercomputer as well.”

That comment made up his mind that it was all hype.

But later he realises he may have misjudged it and...

'Then, like a beam of light shot straight into my soul I come home one day to find the book Game Changer in my mailbox.'

He's got a freebie...(remember I knocked one back.)

He then listed all 18 chapter headings and nothing else adding:

"After playing through the games of AlphaZero I really can’t rave about the games enough."

One rave would have been enough to prove he had actually read a bit of the book.

We then get palmed off to a video about 'Game Changer' from Mathew Sadler!

http://ontheroadtochessmaster.com/r...

I can promise you mine will not be like that.

"...like a beam of light shot straight into my soul.." do you think they will put that in the blurb on the back of the reprint.

***

Jul-16-19
Premium Chessgames Member
  keypusher: <Sally> Quite right about the index, or rather about the lack of one. I should have mentioned that. Doing a game collection does seem like reasonable penance.

I lost my (first edition) copy of 60MG long ago, must shell out for another one.

<Ayler Kupp> has written a great deal about A0's hardware advantage. He does seem to have a point (I'm not competent to judge) but the chess is the chess, however it was produced.

Jul-16-19
Premium Chessgames Member
  AylerKupp: <Sally Simpson> Yes, <keypusher> is right. I have done "some" posting about AlphaZero computational performance advantage. But first let me congratulate you (if that's the right word) for getting a copy of "Game Changer". You may not agree with all of it but I think that, overall, your impression will be favorable.

Let me correct Nakamura's initial impression of the first AlphaZero vs. Stockfish match. AlphaZero was not running on a Google supercomputer, it was using a special chip, a Tensor Processing Units (TCUs) which was designed to optimize the performance of the types of calculations used in neural network-based processing by performing many operations in parallel . This is like using a Graphics/General Processing Units (GPUs) to assist the CPU in similar calculations as is done in LeelaC0. So it was a fairly small, desktop-like system with two cards, one for each TPU and not a large multi-core processor. Four first generation TPUs were used in the first AlphaZero vs. Stockfish match.

Stockfish supposedly ran on a system consisting of 64 threads although I believe that this means 32 cores with hyperthreading enabled, a not-recommended configuration. The second match ran on a system comparable to the system used in recent TCEC tournaments with 44 cores. So, unless Nakamura has a heck of a laptop I would not consider Stockfish's hardware to correspond to that. But I somewhat agree with Nakamura's comment that "If you wanna have a match that's comparable you have to have Stockfish running on a supercomputer as well."

Then again, it depends on what you are interested in comparing. If you want to determine the better chess playing <system> (hardware + software) then that's clearly AlphaZero. But if you want to determine the better chess playing <software> than that's Stockfish, although perhaps not as clearly, at least to some.

I initially estimated AlphaZero's hardware computational advantage to be in the order of 100X but I later revised it to 80X when I found out that AlphaZero used first generation TPUs rather than the second generation (and faster) TPUs that I initially assumed. That would be analogous to considering Komodo to be a better engine than Stockfish if it ran on a system with two 44-core processors and Stockfish was restricted to running on a 1-core processor. So I was not surprised at all that AlphaZero defeated Stockfish decisively in either of their matches, and it's "interesting" that Sadler did not address the issue of AlphaZero's computational performance advantage in "Game Changer". Then again, it depends on what you are interesting in comparing.

Similar conclusions can be reached with DeepMind's data that attempted to take into account AlphaZero's computational performance advantage by reducing the time it had available to calculate so that the number of calculations that both AlphaZero and Stockfish could perform in a given amount of time would be roughly equal.. Their data shows that if AlphaZero's available time was reduced by a factor of 10 it would still beat Stockfish in a 100-game match by a score of 55.5 – 44.5. But if AlphaZero's available time was reduced by a factor of 30 then Stockfish would beat AlphaZero in a 100-game match by a score of 57 – 43. And if AlphaZero's available time was reduced by a factor of 100 then Stockfish would beat AlphaZero in a 100-game match by a score of 67 – 33.

Interpolating the results I estimate that if AlphaZero's available time was reduced by a factor of 80 to approximate the conditions of their first match then Stockfish would beat AlphaZero in a 100-game match by a score of 66 – 44. So I think that I was being generous to AlphaZero in calling the results of matches with AlphaZero enjoying a 80X computation performance advantage as "inconclusive".

Still, that's no reason to reduce our enjoyment of the book as long as you don't take its conclusion about the matches and AlphaZero's performance against Stockfish while enjoying an 80X computational performance advantage too seriously. Although I'm not sure I would agree with <keypusher>'s last statement in his post about "the chess is the chess, however it was produced." If Nakamura was a 80X better player than I am (and that's probably underestimating Nakamura's capabilities), I'm not sure that there would be too much interest in the results of a 100-game match between the two of us.

Jul-16-19  northernfox: <AK> I am not an IT expert, but it does not seem right to me that your interpolated conclusion is that A0 would win a 100 game match by 66-44. A0 is good, but not that good!
Jul-16-19
Premium Chessgames Member
  Sally Simpson: ***

You lost your 60MG's...Mine has been to practically every country in Europe, including the old East Germany!

When I played postal chess one lad in prison had never seen a copy - I sent him mine - it's done three months inside and I got it back.

Hawk the 2nd shops they are starting to turn up, the descriptive version, not seen a 2nd hand algebraic yet.

You had better give me the link (or links ) where Ayler Kupp mentions Alf's hardware advantage. I'll try and grasp it - if not I'll ask questions. (I've tried looking for it but he seems to post in every thread about every subject....bless him.)

I'm going to have to rein myself in when doing page 98.....but I have not read the whole book so hopefully Mathew has reined himself by not claiming these are new concepts - No.12 is covered very well in 'Simple Chess by Stean. Most of the others by Steinitz/Tarrasch through to Reinfeld. and No.1 by Greco...when he was in nappies.

OK over to AlphaZero vs Stockfish, 2018

PS: if anyone asks what's on page 98 tell them to buy the book, best not to get involved with these people, if they don't have The Book then they won't understand - put them on ignore.

***

Jul-17-19
Premium Chessgames Member
  keypusher: <You had better give me the link (or links ) where Ayler Kupp mentions Alf's hardware advantage. I'll try and grasp it - if not I'll ask questions. (I've tried looking for it but he seems to post in every thread about every subject....bless him.)>

Hahaha, no thanks, Sally. You also have a search function. <AylerKupp> had some learned interlocutors, like <MrMelad> and <scholes>. I'm not qualified to have any part in their discussion.

Jul-17-19
Premium Chessgames Member
  Sally Simpson: ***

Hi K.P.

Don't need the links. I'll be using layman's terms.

BTW what do you think of Mathew's way of looking at a game.

"I like to play through games with the fast-forward button held down, which means that a full games flashes before my eyes in about 10 - 15 seconds.

It's enough to get a feel of the outline of the game and a sense of whether something unusual happened." (it's on page 33.)

I do that but use the arrow keys at 3-4 seconds between each move and STOP! when something nips my tits.

And I won't move on till I discover what it is my inner eye has spotted. Often it a two-mover, sometimes an unpunished blunder but it's usually something that needs looking at.

Where was I...

Think S.F. was playing at a reasonable high level and was not dumbed down. Here.

Stockfish vs AlphaZero, 2018 White to play.


click for larger view

White played 25.Qd1 instead of 25.Nb5.

(Pages 34 and 283 - 'Raking Bishops.')

So what happens if White saves the Knight.

Danny King here https://www.youtube.com/watch?v=-bn... gives a stuttering explanation and book review (not as smooth as his normal vid's possibly done on the morning after a heavy night. He gets better after a swig of coffee about 5:55 in.)

You mention S.F. liked White, then 0.00 (there is that zero zero zero again) and 16 minutes later it thought Black was winning.

That tells me S.F. was looking deep and I am now off to go through that game from the book. (should take me 10-15 seconds....no it will take me about 2 hours on a board.)

This is my last post here till I do the review. Already have 50% of it written in my head along with a few cartoons and pics.

Seen a few other reviews and people are copying each other or simply cut and pasting the hyped up advert.

If you are going to do one might as well make an effort. Give an opinion, show what you like and don't like.

No opening index - taking me ages to find a game. Mathew said they were inspired (there is that word again) by Fischer's 60 giving the games names.

OK but these names do not appear in the index - which BTW is a shambles. (go to the index and see how many names are meant to be on page 12. (14, Allen, Bennett...Tomasev) go to page 12 and none of them are there. Most crop up on page 17. Some names don't appear in the named index (Hoogland page 18) but that is Edward Winter territory. I'll mention it here but not in my blog.)

To give me more time I posted a blog yesterday https://www.redhotpawn.com/chess-bl... (1,200 hits already - these people are crazy, I would not read it and I write it!)

I'll be in touch.

***

Jul-17-19
Premium Chessgames Member
  AylerKupp: <<Sally Simpson> Think S.F. was playing at a reasonable high level and was not dumbed down.>

First let me clarify that there were two sets of matches, one 100-game match played in late 2017 from the initial opening position and a set of matches of about 1,000 games played in early 2018. Some of theses games were played from specific starting positions. It's the second set of matches that are addressed in "Game Changer" by looking at a subset of the games as described in Chapter 18.

Stockfish was not dumbed down too much for the second set of matches; the time control was 3 hours/game plus a 15 sec increment per move, presumably starting at move 1. Stockfish used a similar system as was used for TCEC Season 10 in Oct – Dec 2017; a 44 core, 2.2 GHz Xeon 2699 v4 server. Stockfish was configured to use a 32 GB hash table and (presumably) 6-piece Syzygy tablebases. These is a reasonably powerful high-level system although systems with many more cores have been used in other engine tournaments. For the games starting from the initial position I don't believe that Stockfish was allowed to use an opening book.

For the late 2017 match, the one that generated all the excitement as a result of AlphaZero defeating Stockfish 66was definitely dumbed down if not downright crippled. The time control was 1 minute per move which effectively prevented Stockfish from using its time management function which would have allowed it to use less time in simpler positions or positions with forced moved and saved some time for more complex critical positions. Neither opening book nor tablebase support was allowed and a much smaller 1 GB hash table was used. The reason for these restrictions was that AlphaZero did not use opening books, had no tablebase support, and had no time management function.

When the subject of the substantial computational performance advantage that AlphaZero enjoyed was brought up its supported didn't think that it would be fair to penalize AlphaZero because Stockfish was not able to use TPUs. But somehow it was fair to penalize Stockfish for being able to use an opening book from the initial positions, tablebases, and a time management function. No reason was given why Stockfish was limited to a 1 GB hash table, the same size I use when running with 4 cores. But, frankly, given the 80X estimated performance advantage enjoyed by AlphaZero I doubt that it would have made much difference in the results, as the second series of matches showed.

NOTE: Create an account today to post replies and access other powerful features which are available only to registered users. Becoming a member is free, anonymous, and takes less than 1 minute! If you already have a username, then simply login login under your username now to join the discussion.

Please observe our posting guidelines:

  1. No obscene, racist, sexist, or profane language.
  2. No spamming, advertising, duplicate, or gibberish posts.
  3. No vitriolic or systematic personal attacks against other members.
  4. Nothing in violation of United States law.
  5. No cyberstalking or malicious posting of negative or private information (doxing/doxxing) of members.
  6. No trolling.
  7. The use of "sock puppet" accounts to circumvent disciplinary action taken by moderators, create a false impression of consensus or support, or stage conversations, is prohibited.

Please try to maintain a semblance of civility at all times.

Blow the Whistle

See something that violates our rules? Blow the whistle and inform a moderator.


NOTE: Please keep all discussion on-topic. This forum is for this specific game only. To discuss chess or this site in general, visit the Kibitzer's Café.

Messages posted by Chessgames members do not necessarily represent the views of Chessgames.com, its employees, or sponsors.
All moderator actions taken are ultimately at the sole discretion of the administration.

This game is type: CLASSICAL. Please report incorrect or missing information by submitting a correction slip to help us improve the quality of our content.

<This page contains Editor Notes. Click here to read them.>

Featured in the Following Game Collections[what is this?]
Knight outpost
from Alphazero brilliancies by Elesius
"Bold Sir Launcelot", p. 113
from Game Changer by keypusher

Home | About | Login | Logout | F.A.Q. | Profile | Preferences | Premium Membership | Kibitzer's Café | Biographer's Bistro | New Kibitzing | Chessforums | Tournament Index | Player Directory | Notable Games | World Chess Championships | Opening Explorer | Guess the Move | Game Collections | ChessBookie Game | Chessgames Challenge | Store | Privacy Notice | Contact Us

Copyright 2001-2021, Chessgames Services LLC