Risk compared to Backgammon

[Metsfanmax]

In contrast, the american verion of checkers has been strongly refuted. This means that top computers can never lose a game, only win or draw. Computers will never theoriticaly succeed in that (get data info in all positions that can arise through optimal play) and still need some years to practicaly succeed in that

I would say that for all practical purposes chess has been "solved" in the sense that I don't believe even the World Chess Champion would win one game out of a hundred against them today.

That's a strong claim; I wonder what the numbers are. Nevertheless, the top chess engines are now 500 FIDE points better than the top grandmasters. At that level of play, 500 point upsets are incredibly rare. It is possible for a 2300-level player to beat a 2800-level player, but even though that happens once in a (very long) while no one insists that the 2300 level player is comparable to the 2800 level player. I'm confident that computers are now better than humans for the same reason.

[premio53]

In contrast, the american verion of checkers has been strongly refuted. This means that top computers can never lose a game, only win or draw. Computers will never theoriticaly succeed in that (get data info in all positions that can arise through optimal play) and still need some years to practicaly succeed in that

I would say that for all practical purposes chess has been "solved" in the sense that I don't believe even the World Chess Champion would win one game out of a hundred against them today.

That's a strong claim; I wonder what the numbers are. Nevertheless, the top chess engines are now 500 FIDE points better than the top grandmasters. At that level of play, 500 point upsets are incredibly rare. It is possible for a 2300-level player to beat a 2800-level player, but even though that happens once in a (very long) while no one insists that the 2300 level player is comparable to the 2800 level player. I'm confident that computers are now better than humans for the same reason.

I may be a little off. A 500 point spread would give a human maybe a 4% chance of winning a game.

[Metsfanmax]

In contrast, the american verion of checkers has been strongly refuted. This means that top computers can never lose a game, only win or draw. Computers will never theoriticaly succeed in that (get data info in all positions that can arise through optimal play) and still need some years to practicaly succeed in that

I would say that for all practical purposes chess has been "solved" in the sense that I don't believe even the World Chess Champion would win one game out of a hundred against them today.

That's a strong claim; I wonder what the numbers are. Nevertheless, the top chess engines are now 500 FIDE points better than the top grandmasters. At that level of play, 500 point upsets are incredibly rare. It is possible for a 2300-level player to beat a 2800-level player, but even though that happens once in a (very long) while no one insists that the 2300 level player is comparable to the 2800 level player. I'm confident that computers are now better than humans for the same reason.

I may be a little off. A 500 point spread would give a human maybe a 4% chance of winning a game.

Do you know of a reference that has this analysis? I'd be curious to see how strongly this is a function of the actual ratings (after all, a 500 point spread between a 1400 and a 900 is not the same as a 500 point spread between a 2800 and a 2300, even if it should be in principle, because ratings are much less likely to be an accurate representation of a player's skill at the lower end).

[premio53]

Here is one link:

http://www.thechessworld.com/learn-ches ... ng-system-

[Kaskavel]

Yes, but computer's performance is stable, human's is not. A 2800 player occasionaly make performances at 2500 or 3100. The computer never makes a performance at 2900 or 3500.
Another parameter that complicates matters is that humans "cheat". They exploit the computer's small weaknesses. In the last match, Kramnik almost won a game by bringing up a position where he knew that computer would not understand that the best move was exchanging bishop for knight (Bxf3 in a catalan) because the horizon of knight's superiority was too big for a computer. They systematicaly try to bring up closed calm strategic positions and avoid great tactical calculations. The computer's ability to do that by taking orders to do the opposite is more complicated and difficult.

[premio53]

Yes, but computer's performance is stable, human's is not. A 2800 player occasionaly make performances at 2500 or 3100. The computer never makes a performance at 2900 or 3500.
Another parameter that complicates matters is that humans "cheat". They exploit the computer's small weaknesses. In the last match, Kramnik almost won a game by bringing up a position where he knew that computer would not understand that the best move was exchanging bishop for knight (Bxf3 in a catalan) because the horizon of knight's superiority was too big for a computer. They systematicaly try to bring up closed calm strategic positions and avoid great tactical calculations. The computer's ability to do that by taking orders to do the opposite is more complicated and difficult.

It's actually the computer that cheats with a built in book but they are they are so strong now that it is just about impossible for any human to find those weaknesses at tournament time controls which is 40 moves in 2 hours. The only way possible I would think a human can still compete is through "correspendence chess" where there are no time limits. Even then I have my doubts. Like I said earlier, I believe programming a Risk program would be more similar to Backgammon than Chess with the dice. The computer would never dominate like it does in chess.

[Kaskavel]

Both are cheating in opening preparation, just in a different way. Do not underestimate the abilty of strong players to learn and recall thousand of moves in the opening. I do it, and I am not 2800. And do not underestimate his ability to scan through them through pure instict plus choosing the one that leads to a more "human" position. Computer's basic cheating is in the endgame, containing all positions with 6 or less pieces on board. As far as a position is left with 6 or less pieces, computer make no mistake. Ever. This is more important that opening databases because in opening, human can always understand and judge, in the endgame, computer can play the strongest moves that sometimes make no sence to the human mind.

[Kaskavel]

And yes, of course. Computer will never dominate risk or backgammon. Or any other at least medium-related dice game. Playing the best move in those games does not lead necessary to a win, you can lose, even from an opponent who makes mistakes. Chess is a different category.

[Metsfanmax]

And yes, of course. Computer will never dominate risk or backgammon. Or any other at least medium-related dice game. Playing the best move in those games does not lead necessary to a win, you can lose, even from an opponent who makes mistakes. Chess is a different category.

I disagree; you just need to have a different definition of "dominate." A computer AI may never win 75% of the time, but it will significantly more often than humans, which means it will dominate on average.

[Chariot of Fire]

Both are cheating in opening preparation, just in a different way. Do not underestimate the abilty of strong players to learn and recall thousand of moves in the opening. I do it, and I am not 2800. And do not underestimate his ability to scan through them through pure instict plus choosing the one that leads to a more "human" position. Computer's basic cheating is in the endgame, containing all positions with 6 or less pieces on board. As far as a position is left with 6 or less pieces, computer make no mistake. Ever. This is more important that opening databases because in opening, human can always understand and judge, in the endgame, computer can play the strongest moves that sometimes make no sence to the human mind.

Very good point

[premio53]

Both are cheating in opening preparation, just in a different way. Do not underestimate the abilty of strong players to learn and recall thousand of moves in the opening. I do it, and I am not 2800. And do not underestimate his ability to scan through them through pure instict plus choosing the one that leads to a more "human" position. Computer's basic cheating is in the endgame, containing all positions with 6 or less pieces on board. As far as a position is left with 6 or less pieces, computer make no mistake. Ever. This is more important that opening databases because in opening, human can always understand and judge, in the endgame, computer can play the strongest moves that sometimes make no sence to the human mind.

I don't really disagree with what you are saying but the knowledge of opening moves is what makes a Grandmaster a Grandmaster. If the computer had to analyze from move one without a human entering a database (opening or ending) a human could probably steer it to a position that would be fatal for the computer much more often. Chess being solved was one of the fears Bobby Fischer had. That is one reason he came up with Fischer Random Chess but even in that variation I don't believe the results would be much different.

To see how big a difference dice can make: A computer will win about 97 out of every 100 games against the World Chess Champion. On the other hand "Snowie" the Backgammon computer king is rated only about 50 elo points above the strongest human players. Huge difference! In the next couple of years Chess programs will be 100% unbeatable by any human.

[nippersean]

It's just a matter of resources, they spent more money on chess.
The backgammon resources have been far less. Computer bods took the decision to take on the most difficult game (chess), because to analyse a position is hard for a computer. And they still kicked ass, in the most competitive game in the world.

OK re cof comment, Kasparov was clearly cheated, but if you read his books, he describes "the machine" as "all
seeing" and "bloodthirsty"

And to Kaska re computers inability to think, i wish you were right, but check Kasimdzhavonv v Kasparov Linares 2005 - Kasparov got given the search engine, thought he'd found a hole (playing white) , then realized the computer was right (playing black) and switched sides.

[premio53]

It's just a matter of resources, they spent more money on chess.
The backgammon resources have been far less. Computer bods took the decision to take on the most difficult game (chess), because to analyse a position is hard for a computer. And they still kicked ass, in the most competitive game in the world.

OK re cof comment, Kasparov was clearly cheated, but if you read his books, he describes "the machine" as "all
seeing" and "bloodthirsty"

And to Kaska re computers inability to think, i wish you were right, but check Kasimdzhavonv v Kasparov Linares 2005 - Kasparov got given the search engine, thought he'd found a hole (playing white) , then realized the computer was right (playing black) and switched sides.

So you believe Backgammon programs will eventually dominate the way they have in chess?

[nippersean]

yes, easy

[premio53]

yes, easy

I believe the dice will not allow the same level of domination. That gives hope to Conquer Club players when someone tries to become the next Conquerer with some computer program.