Something Hikaru Said

[bob]

Too late to edit:

BTW, just because a GM resigns when a knight down does NOT mean his opponent can play that ending perfectly. And the closer to the first move the knight is lost, the greater the probability for error since the game is far more complex with all pieces but that missing one still on the board.

So that is NOT a proof that GMs won't lose a knight up.

[bob]

Someone should take a set of 1,000,000 GM games, and first locate positions where they are a piece ahead, and then analyze to see if they "played perfectly". I remain unconvinced after watching many of the old zonal and inter-zonal events were I watched GMs drop pieces and even queens on occasion. That's not perfect chess. I'm not drinking this "perfect play when a piece ahead" cool-aid without some sort of supporting evidence. "I believe" or "I think" is not evidence. If they can play perfectly when a piece ahead, they can play perfectly when material is even.

I'm fairly certain even you think a GM can win against a king and pawn with a full set of pieces, so somewhere there is a hard limit on what kind of odds a computer can give and still have a chance. Hikaru, and many of us, think that knight odds is beyond the limit. You apparently think it may be substantially more. Let's agree to disagree.

We are not talking about today, however. We are talking about way in the future. 20 years ago a GM could give pawn odds and win easily. 10 years ago it was give away nothing and come close to breaking even. Today we are nearly at the point of giving up a pawn is not enough to cause a program loss. What happens over the next 10 years, the next 100 years? Humans are getting NO stronger, but the hardware and software improvements continue to push computer ratings higher and higher.

I am not sure if humans will not get stronger in the future when they can learn a lot from chess engines.

I don't know that humans can learn tactics from a computer when they currently don't understand what its search is seeing. Much less years into the future.

[Uri Blass]

Someone should take a set of 1,000,000 GM games, and first locate positions where they are a piece ahead, and then analyze to see if they "played perfectly". I remain unconvinced after watching many of the old zonal and inter-zonal events were I watched GMs drop pieces and even queens on occasion. That's not perfect chess. I'm not drinking this "perfect play when a piece ahead" cool-aid without some sort of supporting evidence. "I believe" or "I think" is not evidence. If they can play perfectly when a piece ahead, they can play perfectly when material is even.

I'm fairly certain even you think a GM can win against a king and pawn with a full set of pieces, so somewhere there is a hard limit on what kind of odds a computer can give and still have a chance. Hikaru, and many of us, think that knight odds is beyond the limit. You apparently think it may be substantially more. Let's agree to disagree.

We are not talking about today, however. We are talking about way in the future. 20 years ago a GM could give pawn odds and win easily. 10 years ago it was give away nothing and come close to breaking even. Today we are nearly at the point of giving up a pawn is not enough to cause a program loss. What happens over the next 10 years, the next 100 years? Humans are getting NO stronger, but the hardware and software improvements continue to push computer ratings higher and higher.

I am not sure if humans will not get stronger in the future when they can learn a lot from chess engines.

I don't know that humans can learn tactics from a computer when they currently don't understand what its search is seeing. Much less years into the future.

GM's analyze games that they play with chess engines and they can learn
more about their mistakes today thanks to engines.

Today engines do not give verbal explanations but in the future people may develop also engines that give good explanation why the moves are wrong.

Even without explanation of the engine the GM can find by analysis of playing different lines against the computer to see why the move that the computer suggest is better.

[bob]

Someone should take a set of 1,000,000 GM games, and first locate positions where they are a piece ahead, and then analyze to see if they "played perfectly". I remain unconvinced after watching many of the old zonal and inter-zonal events were I watched GMs drop pieces and even queens on occasion. That's not perfect chess. I'm not drinking this "perfect play when a piece ahead" cool-aid without some sort of supporting evidence. "I believe" or "I think" is not evidence. If they can play perfectly when a piece ahead, they can play perfectly when material is even.

I'm fairly certain even you think a GM can win against a king and pawn with a full set of pieces, so somewhere there is a hard limit on what kind of odds a computer can give and still have a chance. Hikaru, and many of us, think that knight odds is beyond the limit. You apparently think it may be substantially more. Let's agree to disagree.

We are not talking about today, however. We are talking about way in the future. 20 years ago a GM could give pawn odds and win easily. 10 years ago it was give away nothing and come close to breaking even. Today we are nearly at the point of giving up a pawn is not enough to cause a program loss. What happens over the next 10 years, the next 100 years? Humans are getting NO stronger, but the hardware and software improvements continue to push computer ratings higher and higher.

I am not sure if humans will not get stronger in the future when they can learn a lot from chess engines.

I don't know that humans can learn tactics from a computer when they currently don't understand what its search is seeing. Much less years into the future.

GM's analyze games that they play with chess engines and they can learn
more about their mistakes today thanks to engines.

Today engines do not give verbal explanations but in the future people may develop also engines that give good explanation why the moves are wrong.

Even without explanation of the engine the GM can find by analysis of playing different lines against the computer to see why the move that the computer suggest is better.

I don't disagree with that. But I don't think he is learning something that is generalizable. Which means it will be very limited in its application, relating only to positions close to the one being studied. As opposed to developing an understanding of things like weak pawns, or minority attacks, or king-side attacks, which apply in a large percentage of all positions.

[syzygy]

So you ASSUME he plays perfectly when healthy and at a reasonable time control? One can have a computer go over several of his games to verify that is not the case.

Yes, at knight odds. Do you not understand the difference between perfect play at knight odds (or in KRK for that matter) and perfect play from the opening position?

And what is your comment on this:

99% of the time the problem is the HUMAN will get the score down from 3 to 2, then from 2 to 1, and eventually to lost.

If that were true, also a GM would still have a chance to draw the game when behind a knight without compensation. In reality, even Carlsen resigns in such a position when playing a decent GM.

A knight up is generally speaking an easy win for any GM. At that level, a knight up is simply decisive and there is no practical chance of a mistake by the winning human GM (i.e. an error that changes the outcome of the game).

I simply understand the concept of "perfect play" period. And I don't believe any human is capable of that.

So you don't believe that a human is capable of perfect play in KRK period. Fine with me, but that is a big period that necessarily ends any reasonable discussion with you on this topic.

[bob]

So you ASSUME he plays perfectly when healthy and at a reasonable time control? One can have a computer go over several of his games to verify that is not the case.

Yes, at knight odds. Do you not understand the difference between perfect play at knight odds (or in KRK for that matter) and perfect play from the opening position?

And what is your comment on this:

99% of the time the problem is the HUMAN will get the score down from 3 to 2, then from 2 to 1, and eventually to lost.

If that were true, also a GM would still have a chance to draw the game when behind a knight without compensation. In reality, even Carlsen resigns in such a position when playing a decent GM.

A knight up is generally speaking an easy win for any GM. At that level, a knight up is simply decisive and there is no practical chance of a mistake by the winning human GM (i.e. an error that changes the outcome of the game).

I simply understand the concept of "perfect play" period. And I don't believe any human is capable of that.

So you don't believe that a human is capable of perfect play in KRK period. Fine with me, but that is a big period that necessarily ends any reasonable discussion with you on this topic.

Why the straw man switch? Nobody is talking about KNK or KRK. We are talking about starting from move one with ONE piece removed, Not the same thing at all. And for the record, I am not sure a GM can play KQKR perfectly. They should win obviously. But "perfectly" (i.e. always playing the move that produces the shortest distance to mate? Doubtful. And that is trivial compared to just removing a knight.

[duncan]

My problem with this is that is is (a) based on speculation; which is (b) based on what programs can do today or within the next few years. The original question concerned the impossible case of having 32 piece EGTBs available, something I don't believe will ever be possible. But if it were possible, I don't see anything that suggests that 2.0 or 3.0 is an absolute upper bound on this question. I'm not suggesting that knight odds WILL be surmountable for a computer. I've only said that I don't see anything that says a computer can't win with even larger handicaps, at some point in the future... IE at 100M nodes per second we know what they can do, and 1 giga-node per second is imaginable. But what about one tera-node per second? one peta-node? That's just a paltry 10^7 more than we can do today. But we are 10^8 nodes faster than my selective search program in 1970 that was doing one node per second...

I am interested to know your response to kai that max elo is 4877

http://www.talkchess.com/forum/viewtopi ... ew=threads


The gain from doubling the nodes I fitted with a/(b*x^c + 1), where x is the number of doublings, getting the correlation 0.99
The plot is here:


The 40/4', 40/40' and 40/120' CCRL and CEGT levels are shown, and the resulting gain from doubling in this extrapolation is ~70 points at 40/4', ~55 points at 40/40' and ~45 points at 40/120'. The limiting value I get by summing up to infinity over all doublings (infinite time control), and is 1707 points above the Houdini 3 40/40' CCRL level. So, I get 4877 Elo points on CCRL the rating of the perfect engine, similar to what I remember Don got some time ago.

The draw ratio I fitted with a shifted logistic, getting the correlation 0.999. In self play we can expect a very high percentage of draws going to very long time controls.
The plot is here:


The hardest to quantify to me was the win/loss ratio, which I somehow assumed to be constant at longer TC. It seems not to be the case, win/loss ratio seems to decrease with time control (or nodes). I fitted it with 1 + 1/(a*x + b), getting a correlation 0.96.
The plot is here:

[syzygy]

So you ASSUME he plays perfectly when healthy and at a reasonable time control? One can have a computer go over several of his games to verify that is not the case.

Yes, at knight odds. Do you not understand the difference between perfect play at knight odds (or in KRK for that matter) and perfect play from the opening position?

And what is your comment on this:

99% of the time the problem is the HUMAN will get the score down from 3 to 2, then from 2 to 1, and eventually to lost.

If that were true, also a GM would still have a chance to draw the game when behind a knight without compensation. In reality, even Carlsen resigns in such a position when playing a decent GM.

A knight up is generally speaking an easy win for any GM. At that level, a knight up is simply decisive and there is no practical chance of a mistake by the winning human GM (i.e. an error that changes the outcome of the game).

I simply understand the concept of "perfect play" period. And I don't believe any human is capable of that.

So you don't believe that a human is capable of perfect play in KRK period. Fine with me, but that is a big period that necessarily ends any reasonable discussion with you on this topic.

Why the straw man switch? Nobody is talking about KNK or KRK.

I did mention KRK and you made a general statement that covered KRK.

We are talking about starting from move one with ONE piece removed,

Yes, and that makes it MUCH EASIER to play perfectly for the winning side. (Perfect play by the losing side is anyway trivial as random play is already perfect.)

Not the same thing at all.

So stop making the general statement.

And for the record, I am not sure a GM can play KQKR perfectly. They should win obviously. But "perfectly" (i.e. always playing the move that produces the shortest distance to mate?

No, perfect is simply winning positions that are won (and drawing those that are drawn). It is nonsense to insist on "only shortest mate is perfect".

[bob]

So you ASSUME he plays perfectly when healthy and at a reasonable time control? One can have a computer go over several of his games to verify that is not the case.

Yes, at knight odds. Do you not understand the difference between perfect play at knight odds (or in KRK for that matter) and perfect play from the opening position?

And what is your comment on this:

99% of the time the problem is the HUMAN will get the score down from 3 to 2, then from 2 to 1, and eventually to lost.

If that were true, also a GM would still have a chance to draw the game when behind a knight without compensation. In reality, even Carlsen resigns in such a position when playing a decent GM.

A knight up is generally speaking an easy win for any GM. At that level, a knight up is simply decisive and there is no practical chance of a mistake by the winning human GM (i.e. an error that changes the outcome of the game).

I simply understand the concept of "perfect play" period. And I don't believe any human is capable of that.

So you don't believe that a human is capable of perfect play in KRK period. Fine with me, but that is a big period that necessarily ends any reasonable discussion with you on this topic.

Why the straw man switch? Nobody is talking about KNK or KRK.

I did mention KRK and you made a general statement that covered KRK.

We are talking about starting from move one with ONE piece removed,

Yes, and that makes it MUCH EASIER to play perfectly for the winning side. (Perfect play by the losing side is anyway trivial as random play is already perfect.)

Not the same thing at all.

So stop making the general statement.

And for the record, I am not sure a GM can play KQKR perfectly. They should win obviously. But "perfectly" (i.e. always playing the move that produces the shortest distance to mate?

No, perfect is simply winning positions that are won (and drawing those that are drawn). It is nonsense to insist on "only shortest mate is perfect".

YOU need to stop making general statements with ZERO evidence to support them. Such as a knight odds game will see the GM play it perfectly.

We are talking specifically about a game starting with all pieces on the board in the initial starting position, removing ONE knight. Not a simple endgame, but the starting position. Feel free to cite ANYTHING that supports the conclusion that a computer program 10 years in the future, 100 years in the future or 1000 years in the future will not be able to give up a knight and not be able to beat any human. I have not made ANY general statement that addresses KRK. Not a one. It has ALWAYS been about a knight-odds game from the initial position.

Anecdotal evidence about pawn handicap today means absolutely nothing about the basis for the question being discussed. 40 years ago a computer couldn't win a game with a queen advantage. 20 years ago they could not beat GMs in even material games. Today they can beat GMs with a pawn handicap. WHAT do you see that suggests that a knight is some magic asymptote to this curve that gives an absolute upper bound on material handicaps for the computer (or advantage for the GM)???

So stop speculating and posting nonsense. We don't know if there is an upper bound. I'm not saying there is or there is not. I am saying we don't know. You seem to somehow have an ability to get some sort of divine communication that provides such a bound. I suspect that a knight will not be enough in another 50 years. But that is an opinion only. But at least it is an opinion based on extrapolation from 50 years of computer chess data and advances. It is not wild speculation based on zero actual data.

[syzygy]

We don't know if there is an upper bound. I'm not saying there is or there is not. I am saying we don't know. You seem to somehow have an ability to get some sort of divine communication that provides such a bound.

Obviously there is an upper bound. If with all the explanations that you have now been given you still cannot understand why that is so, then I can't help you any further.

No need for divine communication. Sufficient is a good brain.