Tuesday, October 23, 2007

Punishment versus reward in the postmortem

Everyone in the chess improvement business agrees that it is crucial to go over your own games. And, it is always added, especially your losses. Indeed, my old coach would frown when I'd bring a win to the table for us to go over. The losses will highlight the weakest parts of your game, and it is necessary to eliminate weaknesses to get better at chess.

There are two reasons this approach to postmortem analysis is flawed. First, at my level, I make multiple errors in all my games whether they be wins, losses, or draws. The percentage of suboptimal moves is about the same in my wins and losses, so why forsake the opportunity to see how I could have played better in a win (e.g., my game posted yesterday was a win, but I made a few awful moves).

Second, it ignores the importance of positive reinforcement in learning. It is well known that animals learn faster if they are coaxed into a behavior via rewards rather than punishment. It can be quite useful to receive positive feedback on excellent moves that I made, moves whose power I didn't actually appreciate at the time, the happy accidents. Having such moves reinforced, studying them, making sure I see why they are good, will help maintain my skill level and make it more likely I'll make the move again in the future. Again, the frequency of such moves in my games is similar in wins, losses, and draws. Plenty of opportunity to learn.

This can be especially important in the opening: many novices play book openings without fully understanding why the moves are good. Pointing out the principles on which the book moves are based will reinforce them so that the student can be better equipped to apply said principles when they are brought out of book. This is better than merely quickly going through the opening until you find the first off-book move, and then simply looking up the right move. Of course, a coach doesn't want to waste the student's time. If your student is rated 2200, you don't need to tell her "That was a good move: it's called a fork and creates an unstoppable threat." Duh.

It is up to the teacher to know the level of his student and pitch his lessons appropriately. In solipsistic postmortems it is especially easy to know the level of the student, but harder to see without help the good and bad moves that you overlooked during the game. That's why Fritz can be so helpful: an objective teacher that will show you GM-level moves, but frustratingly leaves it up to you to explain why they are good (or why the one you played is bad). Fritz is a tiny super-GM that lives in your computer. Sweet. My new method of postmortem is to go over the game adding copious annotations on my own, and then go through one more time with the help of GM Fritz. When I was doing the Circles, I didn't have a lot of time for postmortem, so I would just quickly go through games with Fritz to get a sense for the good and bad moves.

The stuff above on positive reinforcement is teaching 101, something all teachers know. And what better people to learn from, about learning, than the people whose livelihood depends on its success? That said, they also know that different people have different styles. Some students learn better through punishment and guidance than through mere positive reinforcement. But it is easy to integrate both into your postmortems to find what mixture seems best for your personal learning style. It may feel like you are learning more when your self-esteem is shot, you are feeling the gut-wrenching agony of a defeat, having someone tell you how bad a move was and why. But almost a century of experiments in psychology suggest that you will learn just as much, if not more, through positive reinforcement. Don't you think you'll learn something when your teacher says "Damn, that was a great move! You've created a threat of a knight fork, you are opening up lines for your pieces giving yourself attacking chances, and at the same time dominating his bishop."

Hat tip to Squirrelchess, who brought this topic up in a recent post.

10 Comments:

Blogger Temposchlucker said...

Yeah, but what about the teacher? The whole day long he is humiliated by his boss, his customers, his wife and his cat. He has tootache since 4 o'clock and he didn't sleep well after going too late to bed because of playing blitz. What better is there for him than to restore his self esteem and powers by fulminating against a lower rated person on the other end of the globe in the anonimity of the internet? With an ocean in between to prevent that things become physical? Especially since you even sucks more at chess than him? You want not to deprive him from the only thing that keeps him mentally healthy, would you? By allowing him to only give compliments which he already does all day long and which he does not mean?

10/23/2007 07:37:00 PM  
Blogger Wahrheit said...

I agree that both one's wins and losses can be equally valuable to study--in my opinion it's best to look at all of your "serious" games, and the main thing is to focus in on a couple of things; good moves that your opponent made that you didn't appreciate in advance (or completely overlooked) and your really bad moves (what was I thinking about and what are the warning signs I'm about to do something like this next time?) I don't spend to much time analyzing or worrying about moves that weren't "best" but didn't degrade my position very much. Since I (and all of us) have only a certain limited time to study any given game, this is what's been working for me.

10/23/2007 07:44:00 PM  
Anonymous svensp said...

Sorry for this long entry. At the end there is a conclusion (still pretty long), for details there is the long text :) Now I even feel the additional need for some sort of foreword so as not to be misunderstood: Everyone has ones own method and there is nothing that works for everyone. Freedom of choice of methods rules. So I don't want to state this as an absoulte truth, it's just my opinion which may well be wrong.

Hm. If you look at the mistakes in your losses then you can realize sooner which mistakes are important and which are not. If you win, you might be tempted not to take them seriously as obviously the outcome of the game was in your favour even with the mistake being made. So why even care for that mistake (one might ask)? A difference that makes no difference (even if it could have) is no difference. Whereas mistakes in losing games made a huge difference: you lost. Another point: The goal to get rid of mistakes ("subtract negatives" like Heisman says) is by itself a "negative task", so this is at the core of its being. If one uses a "positive method" one will avoid "regions of negative feelings" when using it and thereby avoid the core of the task. In other words: To avoid mistakes, one first has to look at them as clearly as possible and this will always be a negative experience (but there is also a positive feeling in seeing later: here I could have made this mistake again, but didnt).

Another positive aspect of looking especially at mistakes in losing games: There are always tons of mistakes, some are serious, some are "not so serious", of what class any given mistake is, depends on the player. If one has 1200 than "1600-mistakes" are not such a problem; if one has 1600 than "2000-mistakes" are not such a problem and so on. In losses there is always a decisive mistake (as mentioned before) (or sometimes a bunch of non-decisive mistakes but I would guess that this doesn't happen in an important amount until 1950 or so) and when looking at a bunch of decisive mistakes one can see which mistakes regularly cause losses at this point in ones' chess development. imho it's a great measurement for progress to look at the "losing mistakes" and for seeing which are the most important mistakes.

To come to a conclusion:

1) in winning games mistakes didnt determine the outcome, this makes for an intrinsic tendency not to take them seriously (don't know if the grammar of this sentence is correct at all :) ).
In losing games mistakes (especially the mistakes that lost the game) are for the same reason in general taken more seriously.

2) Recognizing mistakes is in itself a "negative task". Using a positive method may for that reason not work so well. While recognizing mistakes is not that much fun, avoiding them and recognizing the avoidance later on is.

3) Seeing which mistakes lose your games gives you a strong hint on which mistakes to focus your training.

4) All of this doesn't take away the benefits of positive reinforcement for good moves (so i dont see the ignorance of this method that you mention; as it doesn't exclude additional positive reinforcement methods), but there is no alternative to look at the mistakes without any "pain killers". If it is really about the single mistake it shouldnt make a difference if it is in a losing or a winning game. Of course there is one psychologically. But the conclusion would be (to me at least) to rather look at the games where the mistake was important than at the games where it was just an accident with no consequences.

In addition to this: A lot of good moves are only possible because the opponent allowed them and so there could be a completely undeserved lack of "positive reinforcement" when playing against players that don't allow the same amount of "good moves", even though ones' play is no worse than before, maybe even better.

So that's it, congratulations to anyone who read till the end.

kind regards,
svensp

10/24/2007 05:12:00 AM  
Blogger Glenn Wilson said...

Hmm...yes winds and losses both should be examined. Especially if one is looking for trends and tendencies, habits, etc.

I am 2 posts into an eleven (or more) post series to look at my 11 most recent (at the time) ICC blitz games. Some are losses, some are wins. Some of the lossses I should have won; some of the wins I should have lost.

I usually only review my blitz games for opening review and preparation. Here I want to get a overall sense of my blitz play and how to improve it.

10/24/2007 06:08:00 AM  
Blogger takchess said...

Have you stopped using your chess coach ?

10/24/2007 07:09:00 AM  
Blogger Blue Devil Knight said...

Sven: I don't quite get it, and must get to work so must think more about it. On a first read I'm not convinced. When I spot a missed killer tactic in one of my wins, a simple tactic an 800-rated player should have seen, I feel as badly as if I had missed it in a loss, as it makes me realize I should have lost. I can learn just as much from it as such a blunder in one of my losses.

Glenn: it will be interesting to see the games.

Tak: I don't see him any more, by mutual agreement. He thought I really needed to improve on tactics before he could help me much, and that the Circles were a good idea, and that I should also just play a lot.

10/24/2007 08:53:00 AM  
Blogger Blue Devil Knight said...

Sven: new post coming this is too long for me to handle in comments....

10/24/2007 10:04:00 AM  
Anonymous svensp said...

BDK,

Thanks for your answer.

"When I spot a missed killer tactic in one of my wins, a simple tactic an 800-rated player should have seen, I feel as badly as if I had missed it in a loss, as it makes me realize I should have lost. I can learn just as much from it as such a blunder in one of my losses."

If it is really the same for you, you are right of course. I would guess for most people it's a difference if a mistake causes them to lose or not, also in terms of motivation to get rid of it. On the other hand, its true: A fork is a fork is a fork, win or lose.

A new post is coming? I'm curious already.

kind regards,
svensp

10/24/2007 10:16:00 AM  
Blogger Temposchlucker said...

It's just a matter of not founding your ego on your chess ability. You can't be punished if your ego is not connected to it.

10/24/2007 10:20:00 AM  
Blogger Dean said...

hmm, interesting comments. I think I'm definitely one of those who reacts better to the negative refinforcement :). I think with some players it's also an ego thing, they don't want the rest of the world and themselves to see too many losses.

10/24/2007 10:35:00 AM  

Post a Comment

<< Home