An experiment I would like to see.

[Don]

Have you ever played non-consultation chess? You alternate playing the move for your team with your partner, so it's 2 on 2. You are not allowed to consult with your partner.

Years ago we used to do that at the local chess club and the quality of play is horrible, presumably because your partner does not think the same way as you nor does he see the same things you see.

How would this work with 2 equally strong programs that have different playing styles? Would they play worse that either one alone?

[Laskos]

Have you ever played non-consultation chess? You alternate playing the move for your team with your partner, so it's 2 on 2. You are not allowed to consult with your partner.

Years ago we used to do that at the local chess club and the quality of play is horrible, presumably because your partner does not think the same way as you nor does he see the same things you see.

How would this work with 2 equally strong programs that have different playing styles? Would they play worse that either one alone?

Nice question. Here I beg to suppose that the bots are different from humans. Maybe just a little weaker than individual engines.

[Don]

Have you ever played non-consultation chess? You alternate playing the move for your team with your partner, so it's 2 on 2. You are not allowed to consult with your partner.

Years ago we used to do that at the local chess club and the quality of play is horrible, presumably because your partner does not think the same way as you nor does he see the same things you see.

How would this work with 2 equally strong programs that have different playing styles? Would they play worse that either one alone?

Nice question. Here I beg to suppose that the bots are different from humans. Maybe just a little weaker than individual engines.

My best guess is the same as yours, they would play a little weaker.

When we did this the problem was that humans specialize a lot. Each partner has his own way of handling a particular situation, especially with respect to how the game starts. An obvious example is that some players never play 1. d4 and some never play 1. e4. If these two partner up, somebody is going to be playing out of his element.

Even though it's not so obvious with computers, I suspect that it's similar. If program A plays a move program B would not play, it might be because program A understands some concept that program B does not understand and that will also affect the follow-up.

A simplified example: suppose program A understands weak pawns and program B has no concept. The move A selects might be part of a 3 move plan to weaken the opponent pawns. But program B is oblivious and may not correctly follow through on the idea. Even though both program may be equally strong (but for different reasons) they might frequently be working against each other.

Maybe it's even simpler than that. Suppose you think the bishop belongs on c4 and I think it belongs on b5. I play to b5, and on your next move you move it c4! A good program might know better, but each of us has somewhat different notions of what is good and bad. So each of us might not be working with the decisions the other makes, but instead be trying to undo those decisions.

[rbarreira]

That makes sense, but doesn't this argument cut both ways? What if the two programs, having different strengths, prevent each other from continuing down bad paths?

[Don]

That makes sense, but doesn't this argument cut both ways? What if the two programs, having different strengths, prevent each other from continuing down bad paths?

That could happen but I think it's like most things - the bad tends to spoil the good and not the other way around.

We found that if you put a 2300 player with a 1500 player, the level of play is much closer to 1500 than it is 2300. Of course in this thought experiment we assume the players are equal (and they can be adjusted to be so with time handicapping) but I think the principle is similar.

It's a little different if consultation or voting systems between a larger team of players is allowed of course - then players can more easily cover for the players who do not understand the position as well.

There is well known proverb that goes like this: "There is a frustrating of plans where there is no confidential talk, but in the multitude of counselors there is accomplishment."

I never expected to get the answer to this from a scripture But there is indeed frustration involved when 2 players are taking turns moving without being able to consult with each other and I think the reason is that each players plans get frustrated!

Don

[Laskos]

That makes sense, but doesn't this argument cut both ways? What if the two programs, having different strengths, prevent each other from continuing down bad paths?

A possible (even tactical) bad shot is much more probable than a good one. Even with engines sharing 70% move similarity (don't look at Sim03 with chosen positions and very short time allotted), those 30% of "wrong" moves for other engine could much easier slip into a blunder rather than a brilliant move.

My guess is that for two equally in strength, different bots, the playing level would be a little weaker than each of the two. But not so much as compared to humans. In numbers, a highly speculative estimate, -50 points for bots, -300 for humans.

Kai

[Don]

That makes sense, but doesn't this argument cut both ways? What if the two programs, having different strengths, prevent each other from continuing down bad paths?

A possible (even tactical) bad shot is much more probable than a good one. Even with engines sharing 70% move similarity (don't look at Sim03 with chosen positions and very short time allotted), those 30% of "wrong" moves for other engine could much easier slip into a blunder rather than a brilliant move.

My guess is that for two equally in strength, different bots, the playing level would be a little weaker than each of the two. But not so much as compared to humans. In numbers, a highly speculative estimate, -50 points for bots, -300 for humans.

Kai

My guess would be similar to yours.

[Robert Pope]

Another thing to consider is that by the time the alternating engine gets a move, it is now 2 ply further down the tree. So, particularly for tactical shots, it's less likely that the second engine will blunder terribly.