Mig 
Greengard's ChessNinja.com

Computer Blunder

| Permalink | 27 comments

Staying in America Latina, the 8th Magistral de la Republica Argentina is underway. Each year a top computer program participates in the round-robin, typically crushing the category 8 field. This year Shredder is in the lead again despite a bizarre blunder that lost a piece and a game to Lafuente in the third round. As operator Roberto Alvarez reports, this was apparently a one-in-a-million hash table error. Shredder simply didn't recapture a bishop after spending three minutes on the trivial move. Hard to know if it was a RAM glitch or a program glitch.

In some ways, having machines play in regular tournaments like this is more interesting than the high-profile matches. A quantity of games against a variety of opponents shows how even a relatively weak player (i.e., not 2700, or even 2500) can occasionally draw routinely against a top programs with early exchanges. If the human vs machine battle is to continue in the shadow of the Hydra-Adams demolition, humans had better better book up on this stuff. But such games are monotonous. Watching GMs grovel for draws in dry positions isn't chess and isn't interesting.

27 Comments

Seems as if human blunders have already been incorporated in chess programs ! ;-)

Bummer~

Mig, you are also right about less strong players and chess-computer's strength : a friend of mine playing around 2000 (human ELO) has beaten 2-3 times Fritz 7 and draw with Fritz 8 (unleashed mode naturally). Reading a book on anti-computer strategies certainly helps.

Anyway, this bug with Shredder should be there ! Reminds of eg. Karpov playing with Kasparov for WC and hunging up a piece (with Queen sac and then rook check, do you remember ?!) - his mind just went somewhere else for that moment ...

More to come ...

Perhaps Shredder has developed a taste for "Alekohol"?? HHmm??

Either him or his creator ... 8-))
[just joking of course]

He he he ! Just noticed your name whiskeyrebel !
Right ! You are 101% correct man !

Please notice that in my previous message I actually meant "Anyway, this bug with Shredder should NOT be there !"
Moreover, "More to come ..." goes to AI miracles waiting for us in the future ..

sorry for so many posts ..

Now that is truly scary! Computers really are beginning to think like people! ;-)

Facing a machine with tournaments rules it is not possible for GMs to win or perform an interesting game with a strong computer.

I think it could be possible to have very good games human-computer facing the machine in a human confortable condition for playing chess.

Tournament rules (if you touch you must move, time control set to 2hx40 moves ecc) are done for interesting humans vs humans events.
They are not to let GMs show their absolute level but their level RELATIVE to the rules of the tournament.

Computer has free access to all the tools of modern tecnology.
Human can use only his brain.
Human is limited by rules in using eyes and hands for example...instead we can say that computers use not only the electronic correspondent of a human brain but also the electronic correspondent of eyes and hands.

I'd like to see good GM vs strong computers games but as computers can use all they need to give their best, in the same way humans must have what they need to give their best.
Can touch pieces.
Can ask for all the time they need (ok, computer will have the same...and the request has to be reasonable, if a human asks 2h for each move simply cannot play).
No zeitnots with blunders ask for more time again, must be an event between gentlemen) and, more important, everytime humans need a rest, a walk, eat something, even a shower, during the game they have right to have it: not "play play a move quick, time is ending" that is the blunder condition.

Do not say that in this way games are not equal:
computers and human have the same rights but simply their needs are different.
Not giving a human the rest he needs is like not giving a computer the RAM he needs for his hash tables.
Human does not need RAM but need rest, give him the rest, computer does not need rest but needs RAM, give him the RAM...it is an equal condition.

There is something that bugs me about this post a little when the comment is made:

"Watching GMs grovel for draws in dry positions isn't chess..."

I don't understand where or why the idea comes from that only super-tactical positions "are chess". Endgames "are chess", strategic positions "are chess", and tactical positions also "are chess", it seems to be a strange point of view and I believe the types of positions a lot of amateurs would call dry but that GMs play on in and win shows the difference in appreciation for the subtleties of chess. And yes, these subtleties "are chess".

When I was in vegas for the North American Open this last winter, I was playing a game 5th round I believe and I had messed up the opening a little bit and on move 13 or so I saw I could force a queen trade into what I viewed as an interesting endgame where I would only be slightly worse (as compared to possibly much worse) in other lines. As I force the queen trade my opponent mutters under his breath, but certainly loud enough I could hear "This is not chess". A wonderful, classy sign of good sportsmanship. Well, I completely outplayed my opponent in this "non-chess" position and had a winning position. During move 60 I saw my opponent turn over his scoresheet during my move and start numbering the back of his scoresheet so I got up and whispered "I'll get two scoresheets" as I started walking to the back of the room he said "just play the game" which I just ignored and went to get two scoresheets, when I came back a good 45 seconds later, my opponent was already cleaning up and had resigned. Ah, I love being the badboy of "fake chess".

Is this the same question as that of playing styles : eg. Fischer vs Petrosian (or Karpov maybe) ?!
Anyway, I agree, this is the beauty of chess : most quiet and most complicated positions need their own unique and magic handling.
Very difficult for a player to excel in both type of positions but in no other way can play the best games he can !
This is true chess I think.

Nice story, jegutman! I don't usually run into such behavior. I am a USCF master, and players of my rating tend to accept that different people have different styles, and that it is legitimate to try to win in whatever style you choose.

(Those who do *not* accept this tend not to become very strong players, since prejudicing yourself too much against a particular style tends to hamper your development as a player).

I confess, I sometimes have a perverse desire to Ulf-Andersson my opponents by exchanging queens on move 8 and then trying to squeeze the most of a +=/= endgame.

I know that I loathe playing Wojtkiewicz as black in particular because of his style.

Watching a GM grovel by desperately swapping off pieces to reach a drawish endgame against a computer is not the same as having good technique or a dry style. Anti-computer chess may be the only way for humans to survive, but if it means warping the game so far from what makes it interesting, the point has been proven and the experiment has lost interest, at least it has lost my interest.

If the human is playing this way in a clever attempt to win, that is completely different. You can't win without risking at some point. Trying to get technical draws is another matter, and simply reduces the game to an anti-computer puzzle that rarely succeeds and is as anti-chess as anti-computer when it does succeed.

I'd rather give the human access to an opening database or other ways of changing the encounter to make it more balanced, if this is to continue. That said, I'd like to see Anand or Kasparov take on Hydra before deciding it's time to move to the next phase.

It brought a smile to my face to see the computer blunder. Regulations should be tight on computers in tournaments and matches... any internal error, operator error or program malfunction should be left to the computer on its own to sort out... a human chess player can't exactly ask for a massage or other outside influence to aid him. Humans should certainly insist on this too, instead of being too gentlemanly- a point that hasn't been made much of is that Kasparov allowed an operator error to be corrected with no qualms in his last Deep Blue match.

I just saw an episode where a WSOP player got to have a massage for nearly 3 days straight at the table, why can't we? Better then any drug to help my game!

I'm sure this has been written before but I've never actually see it in print. Why do computers get endgame tables and opening books in matches with humans? What happens when a GM plays a computer without an opening book? It has always seemed crazy to me that the computers are pre-programmed for opening play in man vs machine matches....its unfair to the human by largely eliminating the chance that the computer will make any meaningful errors early in a game (the human, of course, has no such guarantee). I'm not an expert on computer chess, but I would guess the machines would make some inaccurate and planless moves without opening books that could give humans more room to breathe for the rest of the game and make the contests far more interesting. This would give a huge mental boost to human players....they must look at the games now like some horrible gauntlet like marathon where one misstep leads to a loss. Computer chess proponents make such a big deal of their calculating power...make the silicon beats stand on their own from move one. How much of an impact would this have on results?

I think because of the fact that early machines were playing a very low-quality chess, we were all just accustomed to any aid that would help machines appear less ridiculous : ready-learnt opening, ready-learnt endings and such things.

Now that machines are beating almost each human easily and they can understand a good deal of strategic elements, why give them such power in the opening and in the ending ?
This is artificial.

Anyway, having them on their own in opening would be unfair too, because human is studying variations all the time.

The best thought is to use machines only for preparation and throw them out of human events.

What about Fischer Random or 360 Chess for Human vs. Computer? No opening preparation, just strategy vs. calculation power. It'd be interesting I think.

Yes, I think if you just transform a normal-chess program to play Fischer Random it will lose promptly against good players : however, they will fix it ! They fix everything ! :-))

Computer opening books and endgame tablebases are just explicit rules for the computer algorithm during certain portions of the game, just like some early programs had special algorithmic cases coded for the endgame for mating with certain pieces. The opening/endgame databases are really just the equivalent of billions of special cases that have hardcoded algorithms available for those cases.

My guess is that Fischer Random would benefit the computer not the humans. Humans are obviously much stronger inside of their comfort zone, whereas to the computer this concept is totally foreign. I for one do not think the question of opening books is a real problem... I mean a player like Kasparov or Anand certainly has a pretty well programmed opening book(they also have extensive endgame tables)!! In the old days(meaning 2 years ago), the opening books were even really bad for the comp as it was just using current theory without being allowed to check what it is about to play(something Kasparov or Anand certainly do). Now they prepare more carefully. Let's face it there is no really decent solution to the fact that humans do not play as well as computers. The only thing could think of is giving humans access to a version of the computer's opening book so that the human can make a good preparation. Even so, I for one suscribe to the runner vs. car school: let's just forget about it.

re: tablebases,

I'm not a programmer, but my understanding of tablebases was that they provided a 'road-map' of all possibilities given X pieces/pawns toward what we would call the end-game.

I also thought I understood that computers play poor endgames without them, or at least have a great deal of trouble reaching advantageous endgames when considering moves during the middlegame.

Computers clearly know how chess-pieces move, and they understand relative value... why is it fair for them to have this extra information? Why not let them deal with evertything from the opening onward on their own, a la humans?

Am I misunderstanding the idea of tablebases?

Artie.

The issues brought up on this board are somewhat strange. While tablebases are a clear advantage for the computer (not the 3-4 piece bases, but teh 5-6 piece bases), they rarely come up, and even if they matter, the game is no longer theoretical, the evaluation is known. Tablebases are made by computers, not by humans, it's essentially reverse engineering and endgame, what's wrong with a computer using available information. Certainly there are probably some people with amazing memory out there, but probably lacking chess skill that could memorize a tablebase (I mean there are people who have entire phonebooks memorized). I could clearly make the same argument for an opening book, let's imagine someone completed a 32 piece tablebase :-) (i.e. solved the game of chess including fischer random) This would be a database of incredible proportions, some people think such a thing would ruin chess, but it would be impossible to read through this tablebase in a person's lifetime (just to a rough estimation of the number of possible chess positions, then remember there's a difference based on whose turn it is to move in the same position, whether en passant is legal, castling is legal, etc.) Assuming chess is a draw, there would be tons of moves that led to a draw, and still tons of options, all too much to memorize, and tons of possible deviations. There is a ton of opening theory out there now, and you'll still see masters drop pieces by move 10 every once in a while when they're careless. Computer's don't have any particularly strange advantage against humans other than doing something a human would never try to do: apply a very basic positional understanding to a position that can only hold its own because it is backed by deep calculation even in relatively quiet positions. Also, don't forget, humans still have this advantage over computers that very few people realize: We can calculate DEEPER!!!!!. Computers have a problem in that they calculate every move they deem reasonable at a certain depth, well, most humans don't have to calculate these moves, I'm sure a player like Shirov who is known for incredibly deep yet tactical, non-forcing sacrifices would really smash a computer from the right position.

Humans have a tablebase in the sense that you can't be a world class player without knowing a large number of endgames almost by heart. Anand does not figure out the Lucena position everytime it comes up in his tree of variations... Obviously the computers is more exact.... but to leave a computer without a tablebase would really be unfair the other way...

DP - excellent point, but does an Anand really 'know' these positions by heart, or does he understand principles that allow him to take advantage of the position. I'm not sure it is apples and oranges at that point - the computer KNOWS the roadmap, where I suspect Anand 'senses' the right path and applies his skill and knowledge.

Tough for me to say - I'm neither expert in chess or tablebases :(

With such little material he has to have alot of very specific knowledge especially as to the theoretical results of certain positions, which is really important in deciding which positions to aim. Once he reaches them, his opponents either resign or he probably has practiced them enough that he can carry them out almost like a machine.

I visited this page first time to get info on people search and found it Very Good Job of acknowledgment and a marvelous source of info......... Thanks Admin! http://www.reverse-phone-look-up.net

Twitter Updates

    Follow me on Twitter

     

    Archives

    About this Entry

    This page contains a single entry by Mig published on July 20, 2005 11:10 PM.

    Ajedrez y Negocios was the previous entry in this blog.

    Lennon or Lenin is the next entry in this blog.

    Find recent content on the main index or look in the archives to find all content.