I'm wondering if the following scenario has ever been played out:

Once side of a chess game is played by multiple humans at once, collaboratively determining what the next move will be. This could, perhaps, be enforced by some sort of game logic where the move with the highest percentage is the chosen-move. Let's pretend that there could be as many as 1000 humans contributing to the decision making. The other side is played by a single chess computer, and a reasonably-good one at that.

Has anyone ever attempted something like this? If so, what was the outcome? Does adding more human players to make decisions for one side of the game increase the likelihood of that side defeating a computer opponent?

  • 2
    Yes, me and two of my brothers has played against Rybka. We lost all 4 games. Based on our study adding more humans had no increase in likelyhood of winning (when our father joined the game, we still lost). – Salvador Dali Jul 5 '14 at 3:58
  • 1
    The human side could play so that the top player is basically playing alone and the other players simply check for blunders before the move is made. That should improve the human side's performance compared to the top player playing alone, because the effect of the other players can only be positive (if we don't take some psychological aspects into account). The questions are, is this improvement significant at all, and does the human side have a better strategy? – JiK Jul 10 '14 at 10:24
  • Yes, that's the root question I'd be interesting in figuring out the answer to if such an experiment has ever been carried out. The metrics you propose are very interesting and may indeed help to improve the performance of the human side. Who knows! – Ravenstine Jul 10 '14 at 15:23

Six months ago a match was held on ICC: Komodo, one of the strongest engine, faced hundreds of people combined. Komodo won.


  • "Komodo won." Who would have guessed... – Qudit May 13 '19 at 6:59
  • 1
    The problem with that "hundreds of people combined" is that many of them did not make a great contribution. I have my doubts of what would happen if you could put all top 10 players to cooperate – David May 14 '19 at 0:24

Here's a source about how GM David Howell and three others took on AlphaZero in a game. They played two games, both drawn.

Caveat: AlphaZero isn't like conventional engines. Conventional engines have a "contempt" score that tweaks how it plays against weaker opposition. With a high contempt score, conventional engines are very likely to have beaten GM David Howell and his three teammates.


Yes, usually in Freestyle tournaments (which normally consist of mixed human and computer teams). There is some detail on them here, dealing particularly with the PAL/CSS tournaments.

This Freestyle site is worth a look too, though the focus is specifically Centaur (i.e. human & computer) entries. There's a little more on this topic in the Wikipedia article for Advanced Chess.

Be careful of sites that say Freestyle is just like Correspondence chess, there is a key difference. In Freestyle you can use any aspect of computer assistance, including engine analysis. In correspondence (using the ICCF rules) you cannot analyse games or positions, but using opening books or looking up databases is acceptable (it is considered the same as reading books or exploring openings). The point being that you need to be the one making the assessment and the decision.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.