Could all 2500+ (combined) play AlphaZero to a win given 20 tries? • page 2/3 • General Chess Discussion • lichess.org

I don't actually think the rating difference between the engine and the players would be as relevant as you think. Magnus Carlsen is considered to have an accuracy of around 98%. So for a given position, the vast majority of the time he will find the best move anyway. Presumably, with a team of players of similar levels and a considerable amount of time to study the position, the team accuracy is probably going to be >99%. Again, I'm not saying they'll draw every game, but in a given position, there is not always going to be some super difficult-to-calculate line that the players will never be able to find. The overwhelming majority of the time, they'll find the best move.

I, on the other had, probably have an average accuracy of maybe 80% in long time controls, and my knowledge of theoretical lines is pretty thin. The difference between 80% accuracy and 98% accuracy vs. 98% to 99.xx% is going to result in many more losses for me than for Magnus playing a higher rated engine.

[edit]Moreover, there's maybe a strategic point worth considering, that teamGM knows which theoretical lines tend to be very drawish, and can try to steer the game in that sort of direction on purpose, making it easier on themselves to achieve the result they want.

@Itsmidnight I don't actually think the rating difference between the engine and the players would be as relevant as you think. Magnus Carlsen is considered to have an accuracy of around 98%. So for a given position, the vast majority of the time he will find the best move anyway. Presumably, with a team of players of similar levels and a considerable amount of time to study the position, the team accuracy is probably going to be >99%. Again, I'm not saying they'll draw every game, but in a given position, there is not always going to be some super difficult-to-calculate line that the players will never be able to find. The overwhelming majority of the time, they'll find the best move. I, on the other had, probably have an average accuracy of maybe 80% in long time controls, and my knowledge of theoretical lines is pretty thin. The difference between 80% accuracy and 98% accuracy vs. 98% to 99.xx% is going to result in many more losses for me than for Magnus playing a higher rated engine. [edit]Moreover, there's maybe a strategic point worth considering, that teamGM knows which theoretical lines tend to be very drawish, and can try to steer the game in that sort of direction on purpose, making it easier on themselves to achieve the result they want.

JohnLuke090904

#12

This actually happened with Garry Kasparov before. A bunch of GMs came together and battled him in an epic correspondence match. Kasparov ended up winning so I think that tells a lot about how this might go.

Bugday

#13

Maybe they can draw one game... Maybe. But engine would crush them in other games.

ryan121

#14

@JohnLuke090904
Are you talking about Kasparov vs The World? Can you tell which grandmasters participated in Team GM vs Kasparov?

@JohnLuke090904 Are you talking about Kasparov vs The World? Can you tell which grandmasters participated in Team GM vs Kasparov?

JohnLuke090904

#15

@ryan121 I am talking about that. And the entire Russian grandmaster league contributed as well as many others.

turtleleo77

#16

sf 13 is much better than alphazero i belive. alphazero is old

ryan121

#17

@JohnLuke090904
The final move is decided by voting of the world. These voters aren't qualified. If the GM team can decide independently, Kasparov can't beat them.

@JohnLuke090904 The final move is decided by voting of the world. These voters aren't qualified. If the GM team can decide independently, Kasparov can't beat them.

JohnLuke090904 edited

#18

@ryan121 The GMs actually did influence all the moves because the ordinary people weren’t participating. Edit: 50,000 participants but I assume that most know chess well and were consulting with the GMs

clousems

#19

My guess: Bobby Fischer, fearful of being spied on by the computer, smashes it to smithereens with a baseball bat after a couple games. GMs win on time.

ryan121

#20

@JohnLuke090904
After the game, there were controversies. That somehow an Inferior move was selected among the candidate moves. if a Carlsen vs Team GM takes place, Voting should be eliminated to prove that Carlsen (or Alpha Zero, Kasparov) won Fairly. Team GM should be rewarded financially for their winning efforts.

@JohnLuke090904 After the game, there were controversies. That somehow an Inferior move was selected among the candidate moves. if a Carlsen vs Team GM takes place, Voting should be eliminated to prove that Carlsen (or Alpha Zero, Kasparov) won Fairly. Team GM should be rewarded financially for their winning efforts.