Accusations of cheating in the highest tournament levels of chess

[Narz]

I suppose this is another reason why the current chess GM crowd is filled with clowns/characters; it wasn't this way before computers obliterated everyone

Chess players have always been weird and always will be weird.

Magnus is far less odd than Fischer.

 

[stinkubus]

More complex games than Chess have already been solved. The only reason a complete solution would not be published it because it would be considered a waste of computing time if you've already got an engine that performs well enough in monte carlo simulations.

 

[Kyriakos]

More complex games than Chess have already been solved. The only reason a complete solution would not be published it because it would be considered a waste of computing time if you've already got an engine that performs well enough in monte carlo simulations.

Imo it should, since inevitably there will be the one person who will memorize all typical (near average number of moves played) games and then show irl how pointless this is. You'd then at best be stuck with people playing the knight for 3-4 moves in succession to try to throw him off

 

[stinkubus]

I really doubt there are any humans on earth that would be capable of memorizing a GTO chess strategy. Too many game states.

The fact you'd have to mimic the chess engine precisely just to draw over and over again is quite literally what it means for a game to be solved.

 

[TheMeInTeam]

It's already been mentioned ITT, but I have no idea why competitive chess persists when the game has, for all intents and purposes, been "solved". What can a human player do at this point that would be novel or interesting?

there aren't many games which aren't already "solved" or could be in principle by ML algorithms. i don't see what makes chess special, and the same applies to starcraft 2/go/any platformer/rocket league/any shooter and so on.

sc2 ml ai won't just beat you, it will absolutely dumpster you. if you did nothing but practice sc2 for 60 hours per week for the next year of your life and played the best sc2 agent again...it will still absolutely dumpster you. how is this meaningfully different from chess then in this context?

complex tbs might need more compute, but i don't otherwise see what makes anybody think that these are sacred/above the ability for machines to also "solve" in the way you're using it.

Imo it should, since inevitably there will be the one person who will memorize all typical (near average number of moves played) games and then show irl how pointless this is. You'd then at best be stuck with people playing the knight for 3-4 moves in succession to try to throw him off

while chess has fewer possibilities than other games, my understanding is that memorizing that many board states and perfectly recalling them is still far beyond human ability.

 

[Kyriakos]

I don't know, I suppose if there are relatively few "average number of moves, or around that" chess games (a very quick googling tells me those would be in the order of 10.000), one can imagine that a few very particular people could use mnemonics to do that.

I am not suggesting that someone tries to memorize or partly memorize all possible games, regardless of how practical it is they take place.

 

[Narz]

More complex games than Chess have already been solved. The only reason a complete solution would not be published it because it would be considered a waste of computing time if you've already got an engine that performs well enough in monte carlo simulations.

This isn't true.

We've solved only up to 7-pieces. It will never be solved from the start. Maybe 8-piece tablebases in our lifetime. For it to be solved we'd need 32 piece tablebase.

It is estimated there are to 1040 (10 to the 40th power) moves in chess. There are even more possible variations of chess games than there are atoms in the observable universe.

 

[Zelig]

It's already been mentioned ITT, but I have no idea why competitive chess persists when the game has, for all intents and purposes, been "solved". What can a human player do at this point that would be novel or interesting?

Why does competitive running exist when fast movement has been solved by cars, or at worst, bicycles?

 

[Kyriakos]

Why does competitive running exist when fast movement has been solved by cars, or at worst, bicycles?

You still only have humans using those running moves and techniques (and drugs ). When machines with legs run like that and outcompete, it will be where we now are with chess.

 

[stinkubus]

Why does competitive running exist when fast movement has been solved by cars, or at worst, bicycles?

I have no idea why competitive running exists at all. As a spectator sport it's about as exciting as watching paint dry.

 

[Kyriakos]

Magnus's statement today:

Ok, so no proof offered, and an open accusation of cheating. I don't see how this looks good, but the chess organizations have to react now.

 

[Snowygerry]

I have no idea why competitive running exists at all. As a spectator sport it's about as exciting as watching paint dry.

As a competition between humans obviously, as all sports do.

You and I can still have a enjoyable game of chess, even though there exist computers (and humans) elsewhere that can easily beat us both.

 

[Kyriakos]

I don't who he is, some jewish GM, but at least he is funny

The gist of what he is saying is that Magnus didn't present anything new, just suspicions without proof that Hans may have cheated more than he has previously admitted to. Also that Magnus only started complaining when he lost to Hans, previously he didn't have a problem with playing in tournaments with him, because he wasn't winning.
This is a very bad look, anyway.

 

[ywhtptgtfo]

sc2 ml ai won't just beat you, it will absolutely dumpster you. if you did nothing but practice sc2 for 60 hours per week for the next year of your life and played the best sc2 agent again...it will still absolutely dumpster you. how is this meaningfully different from chess then in this context?

SC2's Alpha Star did great in its signature run against TLO and Mana, but neither were even close to top tier. The AI also played with a zoomed out screen for most of the matches which negated the need to move the screen across the map.

It did win a few matches against the top player (Serral) but had also lost several matches against lower tier players like BeastyQT.

With that said, the AI was still quite beatable by the top players, although I am quite confident that it could eventually become unbeatable by humans had the experiment run longer. It's too bad Google terminated the project.

 

[TheMeInTeam]

SC2's Alpha Star did great in its signature run against TLO and Mana, but neither were even close to top tier. The AI also played with a zoomed out screen for most of the matches which negated the need to move the screen across the map.

later agent beat serral handily iirc. i suspect it's a matter of time before such engines become unwinnable. it was a similar story in chess...early iterations of the engines were comparable to kasparov, who could beat them sometimes. then they kept getting better at a pace best humans can't do. as you say, i expect this to happen in most games where a ml algorithm can learn them at all.

main limitation is anybody bothering with using the compute to do it. these are certainly quality ai, but it's the opposite problem of current civ 6 ai, which people complain about being bad. even stuff trained to the level of tlo/mana in sc2 are far beyond 99% of players in respective games. an ai that is at vaguely pro level would be too much for almost everyone, and is expensive to produce/retrain alongside patches. i expect it will be some time before this process gets cheap enough that devs prefer it to current practices, and are capable of tuning it to more "average" levels of ability or creating difficulties based on ai skill rather than bonuses. but maybe that eventually happens.

 

[ywhtptgtfo]

later agent beat serral handily iirc. i suspect it's a matter of time before such engines become unwinnable. it was a similar story in chess...early iterations of the engines were comparable to kasparov, who could beat them sometimes. then they kept getting better at a pace best humans can't do. as you say, i expect this to happen in most games where a ml algorithm can learn them at all.

Serral won one of them and other players had more success.

One thing with Starcarft 2 that's different to chess and Civ6 is that AI has additional advantage in the realm of physical parameters. They already auto-detect everything visible on the mini-map as well as cloaked units moving due to their perfect eyes

The AI has to be restricted to human-level APM. Otherwise, their micro could be unbeatable.

One good thing taken away from Alpha Star is that the SC2 meta changed a bit since the human players learned some strategies discovered by the AI (e.g. over-producing workers in a base)

 

[warpus]

To prevent future cheating all subsequent chess tournaments should be performed in the nude

 

[Kyriakos]

To prevent future cheating all subsequent chess tournaments should be performed in the nude

But have you seen how most of the players look?

 

[Kyriakos]

By the way, imo Hikaru should give it a rest. He isn't even playing in tournaments anymore (iirc?), he has no positive reason to try to destroy another player's career.

 

[Kyriakos]

Fide, the chess authority, made an announcement about forming a commission to look into the accusations.
Still, since Magnus has literally given no proof, I am not sure what they will be looking into - assuming it will go beyond checking older claims about online game non-official tournament cheating.