AlphaZero Chess is not that strong ...

[CheckersGuy]

Of course. It is what they say, and there is no reason at all to doubt them.

Lol, you actually believe Google (multibillion dollar company that has exclusive monopoly on advertising and most of data produced by human kind) is ethical and doesn't lie.
I guess just now you are impatiently expecting Santa Clause to arrive with his reindeers and bring you your Xmas present .

DeepMind was already a very well established research company before it was bought by google.
You really think once they were bought they just started framing people ?
Is everything a conspiracy theory to you ?
Furtheremore, DeepMind's research can be reproduced and therefore verified. People are now copying AlphaGoZeros approach with good results and will likely do so in the future with AlphaZero

[Milos]

People are now copying AlphaGoZeros approach with good results and will likely do so in the future with AlphaZero

Could you point me out those good results. What score did they achieve against some 9 den (doesn't have to be AGZ level really)?

Btw. are you hoping to get a position at DeepMind when you are so passionately brown-nosing them?
Sorry to disappoint you, but you seems to be too old for them .
Btw. this result has nothing to do with DeepMind coz it would be simply impossible without 5000TPUs (almost all of the existing TPUs), so it is Google advertisement and certainly they were the ones running the show. DeepMind is like YouTube, only the name remained as a brand, everything else is Google.

[mclane]

Its very obvious that the bean counter who computed 70.000.000 NPS
and came 23-27 searches deep were completely smashed away by the 80.000 NPS b-strategy neural network.

This is what all humans see, all humans who play chess, no matter which elo they have, see how stupid the bean counter stockfish looked.
With its pieces nearly in the beginning position while alpha zero had everything developed and lots of space, stockfish had to move backwards and almost had Zugzwang positions were it was helpless.

Ask the grandmasters who comment or the world chess champions who commented.

Stop trolling . What is the thing in the games you don’t understand ?

Alpha zero played very idealistic and speculative chess.
It played for a plan while stockfish had no clue what was going on.


The approach to create chess strength by alpha beta failed.
The paradigm came to an end.
There is now a new boss in town. And it plays a completely different chess.

[abulmo2]

The results of the match AZ vs SF8 is translated in +100 Elo for AZ.
The weakened part of Stockfish for this match is already known :
1) Only 1 GB hashtable (64GB to 128GB would have been more way more suitable)

Unproven. Most of the study published here on Talkchess are in conditions too far from the one used by the AlphaZero team to reach any conclusion. They show that the optimal hash size is not the maximal one, and that the elo gain is not much (something like +6 (+/- 5) for 1GB vs 128 MB).

2) 1 minute by move fixed time control

- Other time controls would probably have given an advantage to AZ. It is much easier to distinguish easy positions from hard ones with MCTS+NN than with alphabeta.
- The article also deals with Shogi & Go, where fixed move time control is current.

3) No opening book for Stockfish (AZ saved his opening knowledge from its previous games)

AZ did not use an opening book either. The NN may have a better knowledge of the opening, but it does not store any position and cannot be compared to an opening book.

4) Only version 8 (the current development version is already about 40 Elo above)

- Version 8 is the latest official version available from Stockfish site.
- The 40 Elo were measured on slower machine at faster time control, and are probably much less at the time control and computing speed used by the AlphaZero team, because of diminishing returns.

With this 4 points upgraded to a regular level, the current Stockfish version would been already 100 Elo stronger.

Unproven and exagerated. My guess is more something like:
1) +5 elo
2) -10 elo (with AZ tuned for the same time management)
3) 0 elo (with AZ using an opening book too)
4) +20 elo
So a total of +15 elo, invisible on a 100 game tournament.

[Milos]

You really think once they were bought they just started framing people ?

Who was talking about framing ppl, in which context that one came to your mind???
I was talking about cherry picking results, doing many things that are not reported in the paper and many other unethical things that many ppl in research even at universities do. Especially when they know that not a single result of theirs can be verified and replicated.
Actually you believing me being a conspiracy theorists for well-known stuff even in academia shows just how naive you are.

[Ovyron]

We can see that in the game of go, AlphaZero was still learning, but at chess seemed to have arrived to a plateau. Maybe the maximum strength?

No, we're talking about 1 minute per move chess, and 3400 ELO at 1/minute move is laughably weak.

For instance, let's suppose that you invite Eros Riccio to a match against Stockfish at 1/minute move (the same A0 played), while Eros gets his usual time control, say, 30 day starting time, with 30 day added every 10 days.

Eros's handicap will be that he isn't able to upgrade his machine, or download new engines, he's going to be frozen in time with today's technology.

So, given these conditions, what would be the expected result of such a match?

Eros is defeating the very best correspondence chess Centaurs, that are also getting 30 days every 10 moves, so it's not surprising to expect that it'd defeat 1/min move Stockfish 100-0.

100 wins, 0 draws, 0 loses.

Sure, the match would take 30 years, and it's meaningless because Eros is able to use any engine, book, and latest Dev Stockfish to choose his moves, but that's not the point...

The point is that chess moves exist that are able to defeat Stockfish at 1/min move 100% of the time, and A0 wasn't able to find them in 72 of these games.

It's clear that, at least at 1/mine move conditions we're really far away from maximum strength, and the question is if one can create an entity that plays the moves Eros takes days, or weeks to find, while being limited to 1/min move.

And what ELO such an entity would have?

7000 doesn't sound as crazy considering this 100-0 result stands even if you allow Stockfish 1day/move, because with its stupid search it's going to decide on a move very soon and stick with it for the rest of the hours, due to branching factor, doing something useless the rest of the time (and that's why analysis methods like iDeA or DPA exist so the engine would build a more useful analysis tree.)

Heck, you may not need Eros, let RomiChess train against Stockfish for a day for each of its moves against Stockfish at 1/min move, and it might defeat it soundly, just because 1/min move is such a short time frame and it can be easily exploited by a x1440 time advantage, we just need an entity that can find those moves within 1 minute, but what is stopping us?

[duncan]

Czub, you are a proven troll that can't play chess. This is a known fact.
So I'll just leave you in your delusional bubble. Sorry for disturbance, my bad.

whose chess engine has a higher elo, yours or his capa ?

[Milos]

3) No opening book for Stockfish (AZ saved his opening knowledge from its previous games)

AZ did not use an opening book either. The NN may have a better knowledge of the opening, but it does not store any position and cannot be compared to an opening book.

A0 is fully deterministic. What would you say about the match (100 games) of two classical engines on single core played exclusively from chess starting position?
You wouldn't mind? If that is really the case, you have no clue about computer chess.

- Other time controls would probably have given an advantage to AZ. It is much easier to distinguish easy positions from hard ones with MCTS+NN than with alphabeta.

Could you please show us how is this distinguishing of easy positions done in MCTS+NN and how much Elo does it bring?
I can tell you how much Elo does 1move/min costs SF playing against same SF with TC 40/40 - over 30Elo.

My guess is more something like:
1) +5 elo
2) -10 elo (with AZ tuned for the same time management)
3) 0 elo (with AZ using an opening book too)
4) +20 elo
So a total of +15 elo, invisible on a 100 game tournament.

Your guess is simply irrelevant coz it is based on alternate reality.

[mclane]

Its very obvious that the bean counter who computed 70.000.000 NPS
and came 23-27 searches deep were completely smashed away by the 80.000 NPS b-strategy neural network.

This is what all humans see, all humans who play chess, no matter which elo they have, see how stupid the bean counter stockfish looked.
With its pieces nearly in the beginning position while alpha zero had everything developed and lots of space, stockfish had to move backwards and almost had Zugzwang positions were it was helpless.

Ask the grandmasters who comment or the world chess champions who commented.

Stop trolling . What is the thing in the games you don’t understand ?

Alpha zero played very idealistic and speculative chess.
It played for a plan while stockfish had no clue what was going on.


The approach to create chess strength by alpha beta failed.
The paradigm came to an end.
There is now a new boss in town. And it plays a completely different chess.

Czub, you are a proven troll that can't play chess. This is a known fact.
So I'll just leave you in your delusional bubble. Sorry for disturbance, my bad.

Any chess player in the world sees what was going on.
It seems to me the only person in this forum who is not willing to see the truth is you.
Alpha zero beated the search tree beasts.
The bean counters.
Yesterday Komodo , Houdini and stockfish were the best chess program chess programming of the last 50 years was capable to build, and suddenly today, within a day after the event, these whole 50 years was reset by a machine that taught itself how to play chess.

And nothing can turn the wheel back.

It’s a new paradigm and this paradigm is not called a strategy but b strategy.


You can of course continue to call me an idiot, but the idiot are the people who replay the 10 games we have and don’t see the plot/Clou.

[Milos]

Any chess player in the world sees what was going on.

You are not a chess player, at least not in a conventional sense (knowing rules of chess, playing games and liking chess, isn't a sufficient condition for one to be called a chess player, even though it might be a necessary one). Since your initial premise is wrong, the whole conclusion is also wrong. Q.E.D.