Thursday, June 11, 2009

Computers in chess

The computer is both useless and indispensable for postmortems.


Blogger Tommyg said...

Hmmm? I think the computer is indispensable but only when used correctly. If it is a crutch and no independent thought is given before AND after the computer run through, then yes, it is useless (although it is the person's fault not the computer, since the computer only does what it is told)

And now as I look at your post, if you are only talking about postmortems immediately after a game then yes it is probably useless there as well. The computer should never be the first run through. (my coach was adamant about that!)

However, once a person has really gone over the game themselves, the computer engine(s) can provide a lot of ideas. The trick is for the person to then spend time actually studying the suggestions from the computer AND coming up with his or her own opinions about those computer suggestions. Then and only then, can the computer become a goldmine.

6/12/2009 01:16:00 PM  
Anonymous Chunky Rook said...


(At least I think) I couldn't agree more. On the one hand, the computer gives you a quick objective opinion -- at least within its possibilities and limitations -- on the potential of a position and also helps you figuring out magnificent blunders and tactical wins (2+-) you might have missed. On the other hand, the computer yields lines without explaining the ideas behind them; all you get is raw moves. So unless you're already advanced enough to grasp the idea behind a silicon variation, you'll just end up mindlessly staring at the execution of moves that you'd never understand, let alone think of yourself.

Chessmaster used to have a feature that'd translate computer lines into a natural language assessment of the position, saying things like "Black's defence has been slightly weakened; White's pawn structure has slightly improved". Now that's about as helpful as being given raw moves because it explains an outcome, not an idea. But the idea as such is exactly what the next generation of chess computers (the ones that will then go on to build the Matrix) ought to be able to do: being capable of explaining the ideas behind a move order to a human being who's not as well-versed in chess ideas as a master.

I've been playing a tournament these past few months in the lowest-rated player category. All but one player refused to do a post-mortem on the grounds that "we patzers don't see enough, anyway, so I'll just head home and let the game go through the computer". I think they miss the point of chess improvement entirely. Chess improvement is about ideas, not moves. Is it surprising that I won against each of them except the one dude who was willing to do the post-mortem?

6/13/2009 03:25:00 AM  
Blogger Johan Verduyckt said...

First the player has to go over his game (with opponent if available and willing) before a computer is set to analyse the game and the so called improvements of the player his analysis to see if the analysis is correct.

With other words, if one only let the computer do the work then indeed the postmortems of the computer is useless and of no value. If it is to check the analysis of the player (and maybe opponent) upon flaws then it can be helpfull as long as one reminds that one has to check even this analysis oneself aswell.

To conclude, dont believe everything your computer (program) says about the game you played. Make sure you agree with it by examining it.

6/13/2009 04:35:00 AM  
Blogger Polly said...

Chunky Rook: "All but one player refused to do a post-mortem on the grounds that "we patzers don't see enough, anyway, so I'll just head home and let the game go through the computer". I think they miss the point of chess improvement entirely. Chess improvement is about ideas, not moves. Is it surprising that I won against each of them except the one dude who was willing to do the post-mortem?"

It doesn't matter how weak two players are there is always something to be found in the post-mortem. Besides if you're lucky some higher rated players will come over and look at the game and give some suggestions. It's funny when I've gotten an interesting position during a game that my opponent and I are looking over, and suddenly several masters jump in and totally pick apart the position, or they go off on some tangent and my opponent and I are like "Where did we leave off?"

Besides it's the face time with your opponent after the game that is the drawing card of OTB play vesrus playing on ICC.

6/13/2009 05:33:00 PM  
Anonymous Chunky Rook said...

Polly: My thoughts exactly. In my view, post-mortems aren't even all that much about "finding the correct moves", but about experiencing the thought process of another person and listening to how they articulate their plans, what words and patterns they use to assess a position, and so on. The only way a computer can articulate an idea is by showing you moves and adding a +- value.

6/14/2009 04:48:00 AM  
Blogger Blue Devil Knight said...

From a publisher of a chess book:
"We all have computers, but the position in the principal variation is extremely unbalanced and hard to assess. I definitely do not trust engines' evaluations in that particular case.

"Chess Stars books are not a digest of old books dressed with some computer analysis here and there. They attempt to mark new ways and steer the fashion into other directions. For that they count on human evaluations. Computers must associate every position with a number. On the contrary, a strong GM can sense a bad coordination and rule out lines that the engine claims to be good for him. Thus he can save weeks or months of futile analyses. So everyone can probe endless lines on his personal PC, but eventually one or two games at top level will tell the verdict about the principle variation. Or maybe they already have?!"

And a response from a reader:
"That was poetic and I could not have agreed more or have said it any better. Perusing this thread I was horrified on just how dependent players are on their engines to do their 'thinking' and evaluations for them. I get the feeling some players assign human attributes to engines...

"In complicated positions with unbalanced material, especially where king safety is paramount, a chess engine unless heavily guided by a strong player is virtually useless."

On the other hand, if we put a computer against the world's strongest GM, then what will be the result?

On the third hand, if a GM plays a line against a club player, while a computer recommends a different line, I'm probably going to listen to the GM. The computer may well be suggesting a line where you have to make the right move for twenty moves in a row, while the GM is playing a move where you have your choice of moves, while the opponent is the one with all the problems to solve. Computers don't yet deal well with things like flexibility, sharpness, and such. This is the lesson J'adoube had trouble with in his diatribes that caused such a cry.

For instance, given a choice between a very difficult mate in 20 that involves sacrificing a queen and requires you to find the right move each step, versus a line that it evaluates to 12 pawns in your favor but involves exchanging queens to a very simple (for humans) endgame, which do you think the chess engines will recommend? What do you think the GM would recommend? Which solution would likely involve principles that you would be able to apply in tons of other endgames?

6/14/2009 07:40:00 AM  
Blogger Blue Devil Knight said...

The above quotes are from this thread at Chesspub.

6/14/2009 07:42:00 AM  
Blogger Unknown said...

"Computers don't yet deal well with things like flexibility, sharpness, and such."

Actually, it's chess engines that deal with this, but to the point, they do now that the engines are being written by IM's and such.

The reason chess engines play like humans is because humans code them to play like humans. . .oops!

6/14/2009 02:13:00 PM  
Blogger Blue Devil Knight said...

J'adoube: A computer will always pick mate in ten over an easier in practice mate that takes longer. That's just one instance where computers aren't as good as people for giving advice.

I like how I made my general point over at your post when I asked the rhetorical question "What would you do if you had to choose between using a top-notch chess engine or a top-notch human player to help with post-game analysis?"

Good teachers do more than spit out variation trees, and that's all you get from an engine.

6/14/2009 02:41:00 PM  
Blogger Tommyg said...

What is funny about this debate as it stands right now is that it seems to be an either/or situation.

I don't think there is any question that having a very strong player who also knows how to communicate ideas go over your games with you is the ideal! But it also seems obvious that when used correctly computers CAN indeed help us learn, as long as we use the information that the computer gives us as an avenue to our own further exploration. We should never accept as chess gospel the computer or the human viewpoint!

I think it is important to remember that neither the hunman thought process or the chessic knowledge of computers is infallible! But when used together they can create a pedagogical buffet of epic proportions! (I have no idea where that bit of melodramatic writing came from---I sound like I have been reading Dan Brown novels?!?)

6/14/2009 03:25:00 PM  
Blogger Temposchlucker said...

First you have to formulate what you intend to reach with the post mortem. Personally I'm not interested in moves or variations but in new idea's. Since moves and variations never will reoccur. Computer analysis provides you with moves and variations, not with ideas. So I rarely use a computer.

Since the ideas have to be new it is useless to think too long about a position. You have to take your time to identify the characteristics of the position, of course, but every effort after that is close to useless since you can only come up with old ideas, by definition. For new ideas, you need another person. It simply takes too much time to come up with something new by yourself.

So interviewing your opponent is a good way to obtain new ideas. Further a coach who is willing to go over your games will show you new ideas.

I'm not sure if your own game is the best corpse for a post mortem since it's strange course is based on your old ideas anyhow. What is worse, it limits your possibilities to opponents and coaches as persons to get new ideas from.

If you are willing to accept corpses of other people you extend your sources of new ideas to all chess authors who play better than you.

6/14/2009 04:10:00 PM  
Blogger Blue Devil Knight said...

Tempo excellent points, though I think going over your own games is more useful because you will be more likely to remember the lessons learned.

Tommyg: that's why the original post didn't say they are useless, but also indispensable. Clearly it isn't either/or, but a matter of emphasis.

6/14/2009 10:19:00 PM  
Blogger Jeff said...

I've always thought the idea's where for humans to propose and the computers refute them with variations.

6/15/2009 03:59:00 PM  
Blogger Blue Devil Knight said...

Jeff: ideally humans also refute them, especially in tournaments. :) Also, computers will propose good ones too, though typically where there are complications and tactics involved.

6/16/2009 11:24:00 AM  
Blogger Unknown said...

I live alone and my dog doesn't want to learn chess. I play email chess and love it. I can annotate a game for review by other players but usually it is blunder this and blunder that.
So I have several computer games that I can insert the game and play. Sometimes, I will play against a chess engine, sometimes, I will play two chess engines against one another. For me, I have learned a lot this way.

6/22/2009 06:53:00 AM  
Blogger Blue Devil Knight said...

Icreatefun: good point, reinforcing my point that computers are indispensable for postmortems. :)

6/30/2009 09:19:00 AM  
Anonymous cheap computers said...

I think the computer is indispensable but only when used correctly.

8/07/2009 03:11:00 AM  

Post a Comment

<< Home