FREE - In Google Play
FREE - in Win Phone Store
Can someone please explain the point of advanced chess (chess while consulting houdini)? Like what is the point of this match? http://www.chessbase.com/Home/TabId/211/PostId/4009384/first-female-advanced-chess-match-drawn-050413.aspx
Isn't every game just going to be drawn?
Computers have excellent tactical ability but poor positional judgement. The point of Centaur Chess is to combine the positional judgement of a human with the calculating ability of a computer.
The games would all be drawn only if both human partners have the same positional judgement.
What's the difference between tactical ability and positional judgement. Wouldn't a computer compute the best postions to carry out a swift tactful win? I'm confused lol.
Yeah, it's boring, I don't think it will ever catch on.Once people get over the fact that computers play better it will be back to business as usual I assume. Tournaments, clubs, etc. It's still the same great game with the same great competition.I think it's just people's reaction trying to deal with a world where world-beater chess playing programs exist. However not only are "solutions" beyond memorization, but many times the moves computers suggest aren't practical for people to mimic in any case. Either because they're not correct, or too tactical. It's reassuring the highest rated player in the world plays such human chess so successfully. Indeed, no reaction is necessary at all. As I said, IMO it will be business as usual and this strange variant will be passed over without much consideration.
In simple terms, computers are very accurate (sometimes even perfectly accurate) given a small enough window. Say 5, 10, even 15 moves deep depending on whether forcing moves are present in the position. But if moves must be made with consideration beyond this window then the moves are artificial at best, and nonsensical at worst. e.g. closed positions, openings in general, and some endgame positions are good examples. This is where conceptualization and judgment beat out pure calculation.
I see in the article the game only lasted 39 moves. I'm not sure what happened, but a good strategy would be to guide the game to an endgame and then outplay your opponent there (if you're better than them). If you're the weaker player you should try for very tactical / open position or to otherwise stay in the middlegame as long as possible. If you do enter an endgame ideally it yields to calculation like a reduced pawn king and pawn endgame or knight endgame or queen endgame.
What's the difference between tactical ability and positional judgement.
Tactical ability is "I-go-here, he-goes-there".
Positional judgement is "that square is going to become weak, but my piece activity should prevent him from exploiting that advantage".
Between you two I think I got it understood. Thanks.
Its houdini. I'm pretty sure the strength of this engine is greater than Magnus Carlsen. Certainly stronger than the two idiots playing this match. My guess is that they will do whatever houdini suggests, leading to a bunch of boring draws.
Obviously there are positions where the computer's choice is sub-optimal or it wouldn't even be a discussion.
Tactical Ability refers to being able to create combinations, ie a series of moves that results in a win of material, mate or a distinctly advantageous position.
Positional Judgement refers to being able to assess the long-term advantages and disadvantages of one's own position as well as the opponents, and how to exploit your own advantages, minimize your opponents advantage and exploit his/her weakenesses - IMHO. For examples: being able to tell when there is a weakness on the white or black squares; when a knight is better than a bishop (usually in blocked pawn positions); knowing how to take advantage of the opponents backward pawn(s), knowing how to formulate long range plans and knowing which ones are most likely and least likely to succeed. And lots more!
Thing is, houdini plays endgames almost perfectly. It is hard (if not impossible) to outplay a computer with a rating of 3600 in the endgame. Sure they may make small inaccuracies, but with the endgame databases they have built in, computers will never lose a drawn endgame.
Well for one thing engines aren't tablebases... that's why tablebases are included (I assume they're not included for advanced chess, which would be fairly idiotic IMO).
Another point is computer ratings and FIDE ratings aren't a 1 to 1 comparison. Computer ratings (I've been told) tend to be inflated compared to what they would be if the ratings were obtained by playing in human tournaments.
And of course, yes, engines do outplay people in the long run. But not in every individual position... which is the point.
Could someone post an example of where a computer's choice is too "computery" or sub-optimal? I'd like to see a real-life example of the difference, but I'm not strong enough to know when I might see it myself.
Every rook endgame position
Find some technical rook endgame position from a book (or website if it's out there) and watch the computer fail to draw a drawn position or win a won position.
Or a closed position. Ivanov's game where he lost against a GM. The game was drawn but then Ivanov blundered in the endgame when it seems his time was too short to continue cheating by getting moves from houdini. He made a move on his own and lost. But the middlegame was full of pointless moves (the engine didn't know what to do with the closed position). They weren't losing moves, but they moves were mostly just shuffling pieces back and forth.
ivanov is a superman, he can checkmate with a lone bishop.
Maybe I should try a centaur correspondence game with a willing opponent someday. It's probably not fair of me to judge centaur chess without doing it myself -- although it would seem incredibly slow moving, I could still imagine there being a lot of satisfaction from a perfectionist perspective -- the feeling that you are really looking into the deep truth of a position, with all its beautiful complexities being discovered. That's something we just can't do in a normal game of chess.
Also, it would probably help you understand the strengths and drawbacks of engines better (for example, if you use all the engine suggestions and slowly but surely get beat by a better centaur). Really, what better way to learn about them than to struggle to win games with the engine itself? As waffllemaster said, although computers can beat humans in a practical game, it's not because every single move of theirs is perfect.