Stockfish plays practically perfect chess from 120 + 1.2

[Jouni]

I played 100 games matches between SF 17.1 and SFdev. No book. Draw rate with 60 + 0.6 was 98% and with 120 +1.2 100% . No duplicate games with one core. It's over. Chess is solved.

[jefk]

strictly speaking engine vs engine results with 100 pct draw wouldn't mean
yet that the game is a draw (with perfect play); but with the historical
rating progress of the engines the cumulative evidence (*) becomes overwhelming
and it must be concluded that chess is a draw with perfect play. Which theoretically
doesn't mean it's 'solved', yet imo it does mean an 'ultraweak solution' (**).

(*) your result with 120+1 time control, all draws, was already known in
Iccf correspondence chess of course, eg.:
https://www.iccf.com/event?id=105028

(**) with this knowledge, setting up projects going to 'weakly solve' chess
(like it was done with checkers) is futile, ie would be a waste of resource(s)
(better look at internatnl draughts).

[Peter Berger]

Which theoretically
doesn't mean it's 'solved', yet imo it does mean an 'ultraweak solution' (**).

Given that you use these terms so often, I would assume that you had looked up what they actually mean by now. There is nothing all that ultra-weak to an "ultra-weak solution" of a game.

There is of course some empirical evidence that the game of chess is most probably a draw. But I haven't seen anything coming remotely close to even being a serious attempt of solving it.

I'll save you some work, as even the Google KI can give a reasonable definition:

"Ultra-weakly solved" is
a term from game theory that describes a game where it is theoretically proven whether the first player will win, lose, or draw from the starting position, assuming perfect play from both sides. This solution does not provide the actual winning strategy; it simply establishes the game's theoretical outcome from the initial setup. This is the most basic type of game solution, distinct from weakly solved games, which provide a strategy for the initial position, and strongly solved games, which provide a strategy for all possible positions.

[Jouni]

BTW I just read that DU with 64 core Ryzen has played 75 000 games in PlayChess.com and lost 2 .

[jefk]

There is nothing all that ultra-weak to an "ultra-weak solution" of a game.

no but that's just the name, as it was mentioned in a Phd Thesis by Victor Allis (*)
supervised by vd Herik. Where it was stated that if the theoretical outcome
can be determined (eg. as a draw) then you have an 'ultra-weak' solution.
Some may suggest that 'determining' this is not sufficient, it should be proved;
in a rigorous, (oldfashioned?) mathematical way. Well i simply disagree, based on definition
(and it's origin) in mentioned Phd thesis (the term itself was actually invented earlier
by another computer guy, someone for the rest not related to game theory).
The need for a (old fashioned math) 'proof' also is -erroneously- suggested by google (in
your quote) and by wikipedia. Both are incorrect, because the term was defined
in mentioned Phd thesis and there the author only talks about 'determining' (the
theoretical outcome). Ofcourse if rigorously proven (as eg. with the game of hex), it
would be maybe even less 'ultra-weak' (in the verbal sense) then when the outcome
is 'only' determined (beyond all reasonable doubt). NB actually, with all the
cumulative evidence, imo it's indeed proved that the result with best play is a draw.
Then people could argue about mathematical methods, what the requirements are for
a 'proof' etc. Been there, done that. But i have no need to go into this further in this
forum As i found out some time ago this (- computer chess, mainly engine programming-)
forum is not the place to delve deeply into game theory; there were some discussions,
lot's of negative (or hateful) reactions, and as result i now know (quite a lot) about such
terms (and their background) of course. No need trying to lecture me here, especially
if you don't have a Phd in math (preferably, game theory). And if you think google
and wikipedia are right, and dr Allis wrong, go mail him i suggest (but he doesn't
work in this field anymore, nowadays).
For the rest, the topic of this tread were some games by Stockfish, which all
ended in a draw, and a(n) (ironic?) comment by jouni that 'chess is solved'.
Not really the place to delve further into intricacies regarding game theory.
(*) https://cris.maastrichtuniversity.nl/ws ... SET1.0.pdf

[Peter Berger]

No need trying to lecture me here, especially
if you don't have a Phd in math (preferably, game theory).

I don't mean to lecture you, but I want to ask:

where in this text does the author write what you claim he does? I just skimmed through it and I don*t see him challenging any classical definitions. Can you help me?

[jefk]

section 1.5 (p 25 in the pdf link i specified):

"ultra-weakly solved For the initial position(s), the game-theoretic
value has been determined"

The term 'uws' originates from a certain Paul Colley (p24) a person
being irrelevant for the rest regarding the further context and meaning
this term (uws) later acquired.

[Peter Berger]

section 1.5 (p 25 in the pdf link i specified):

"ultra-weakly solved For the initial position(s), the game-theoretic
value has been determined"

The term 'uws' originates from a certain Paul Colley (p24) a person
being irrelevant for the rest regarding the further context and meaning
this term (uws) later acquired.

Thank you very much for providing the link, I have read the relevant part of the text. I don't think it says what you think it does. Anyway, I really don*t mean to lecture you at all - we can just agree to disagree for tonight.

[jefk]

[quote]I don't think it says what you think it does[/quote]

well my quote was literal. for 100 pct

Maybe your interpretation is different, up to you.

[jefk]

jouni wrote

DU with 64 core Ryzen has played 75 000 games in PlayChess.com and lost 2 .

then it would be interesting to see why those 2 games were lost.
Probably anomalies (time forfeit, engine crash, book error etc)?)
Normally the game should end in a draw of course, unless maybe it's
bullet game like 1-0 or so .