Tuesday, November 18, 2008

How "knowing" can hurt doing

The Boylston Chess Club blog cited a recent interesting set of studies, the most recent entitled Why Good Thoughts Block Better Ones.

Sometimes expertise in a field can actually interfere with performance when new situations arise, situations in which your prodigious knowledge actually interferes with recognizing the situation as new and requiring a different or better solution. Imagine if someone switched the 'a' and the 'd' on your keyboard: if you were a really good typist (e.g., you type by hand, not by looking at the keys) this would screw you up for a really long time.

Investigators have studied this biasing influence of knowledge in chess players. What they did was provide a relatively easy smothered mate problem, and then another problem with two solutions. One solution, which takes more moves, is a smothered mate; the other faster solution is not. Those masters asked to find the optimal solution to the second problem performed worse when they first saw the smothered mate problem.

In the study cited above, they extended this research, monitoring eye movements (saccades) in the two groups of players. Interestingly, the masters that initially saw the smothered mate problem stared much longer at the squares and pieces that were relevant for that problem, and had trouble getting their eyes to the right places for the new problem. By measuring the time to find the solution, and comparing players with different ratings, they estimated that the bias effectively reduced the masters' ELO by 300 points.

These studies have many interesting implications. I'll focus on one.

The research highlights a danger in memorizing sets of tactical puzzles (which in general I think is perfectly fine, as I argued here). Namely, the problems act as a kind of cognitive gravity sink. You see something that looks similar to a memorized problem, but overlook the fact that there is a small but killer difference between the memorized position and the one in front of you (this is something I noticed while doing the Circles, and wrote about here with an example).

So, for those of us that like to solve tactical puzzles, what should we do? Most obviously, it seems better to pick problem sets with many little variations on a similar theme. For instance, including many back rank mates some of which take quite a bit of setup, as the opponent has defensive resources. This will stop you from making knee-jerk moves and force you to look at the position to see if the move you "recognize" is actually safe. Of course you should do this all the time, but as the studies show, when it is a motif we have seen time and time again, it can warp our minds to the point where it is hard to see the truth in the position.

17 Comments:

Blogger chesstiger said...

A perfect example why Rowson in Chess for Zebras says why we have to unlearn what we have learned.

Also i want to add that Dan Heisman's advise to look for a better move once you have found a good move still is valid and good advise but it's not that easy as it sounds since one is kinda blocked with the 'theme' of our first solution.

11/19/2008 01:00:00 AM  
Blogger Blue Devil Knight said...

CT: maybe you should change your name to chesszebra. :)

11/19/2008 11:29:00 AM  
Blogger Blue Devil Knight said...

CT: I think we should be careful of denigrating knowledge too much. It's better to know the back rank mate pattern than to not know it, after all.

On the other hand, another thing I had mentioned in my original post but decided to delete to shorten my post is that it suggests you need to learn a lot of patterns. If you only know two or three (to take one extreme), you will try to find those all the time, whether the position calls for it or not (like a scientist trying to fit recalcitrant data into his theory, rather than just change his theory, as in the cartoon in my post).

The more patterns you have stored, the less likely you are to read the position incorrectly.

11/19/2008 02:14:00 PM  
Blogger chesstiger said...

Now there is a thought, me the chesszebra. :-)

Rowson doesn't say that you dont have to learn those paterns and dont put all the knowlegde into your brain. He only says that one must not swear on them as if they are the holy book. Because sometimes all those rules are stopping us from playing the best move(s) in a given position.

So agree with you that one must know all those rules and knowlegde but the next step in improvement is to know when to sin against the rules and knowlegde.

11/19/2008 03:08:00 PM  
Blogger Temposchlucker said...

I think there is a difference in a bias caused on the level of saccades and the emotional bias that blurs the ability to discriminate well. Allthough both may interfere with your chess.

11/19/2008 04:19:00 PM  
Blogger Blue Devil Knight said...

CT: Yes, he discusses this nicely in Chapter 6, which I will someday get around to summarizing. :)

Also, there are issues of time scale. If you learned the back rank mate a year ago you are not necessarily going to have the same bias as if you looked at it two minutes ago. The more recently activated the pattern, the more likely you will be biased by it in a negative way.

I think the work in this research may help explain tactical fatigue in a position. Perhaps it would better be described as tactical saturation. That is, after ten minutes looking at a position, you usually don't see any new tactics. This may because your mind has already homed in on possibilities that don't work out, but thereafter you don't see the squares and pieces for what they are, but as part of failed tactical possibilities.

Tempo: Most likely the bias isn't caused by the saccades in this study, but seem to indicate the cognitive bias, which drives the saccades. I'm not sure what you mean by emotional bias. That could be many things.

11/19/2008 05:06:00 PM  
Blogger chesstiger said...

"I think the work in this research may help explain tactical fatigue in a position. Perhaps it would better be described as tactical saturation. That is, after ten minutes looking at a position, you usually don't see any new tactics. This may because your mind has already homed in on possibilities that don't work out, but thereafter you don't see the squares and pieces for what they are, but as part of failed tactical possibilities."

I know what you mean. I have experienced it many times myself. But i think it's not only the 'already homed in on possibilities' but also the time presure. You must find a move quickly and in a glance you see move x that is good and play it without much thinking. Kinda a bit hope chess.

11/19/2008 05:41:00 PM  
Blogger Blue Devil Knight said...

CT: Hope chess would be different: this is the phenomenon, that I discuss in the link above, in which after a long look at a position you eventually stop seeing new tactics. As Soltis says, if after ten minutes you haven't found it, you probably won't.

11/19/2008 06:55:00 PM  
Blogger Polly said...

Very interesting post. It's not only bias towards certain tactics. In can crop up in late opening or early middle game stages. That may have been part of my problem in the game in my most recent post. My opponent played Nh5 which is typically followed up with f5. I responded to what I expected would be played (f5), and ended out overlooking the real danger of his follow up move.

Tiger: There is a difference between rules and patterns. We learn early on certain rules for developing our pieces such as knights before bishops, don't bring the queen out too soon, try to castle early, don't move the same piece twice before completing your development. etc. There are times we break these rules in response to what is going on in the position.

I look at patterns more like structures within the position that we need to recognize in order to avoid falling into a tactic, or to find a tactic. A simple example is the king and queen are on the same color squares. Being on the same color square increases to possibility of a knight fork, or a bishop skewer. If I'm in an ending with my opponent having a knight or bishop I have to be aware of thse possibilities to avoid losing a piece to these tactics.

I think familiarity with many patterns gives us more possibilities to consider, but we can drawn in the wrong direction based on something we saw recently. Maybe we won our last game using deflection. Next game we may have a similiar position so we're looking for a deflecting move, but that's not in the position. So we can keep looking for the deflection that's not there and miss the fork that is there.

11/19/2008 11:44:00 PM  
Blogger Blue Devil Knight said...

Polly: that's an interesting take on patterns.

11/20/2008 12:46:00 AM  
Blogger wang said...

Wish I had more time to read those two posts.

But you have hit on something very important. This reasoning is why I don't care much for a lot of the tactical puzzle books out there. They tend to isolate by theme. Mate in two, or skewer, or back rank mate.

CT ART does this to a smaller degree. You will get two or three double attack problems in a row when working on problems in the 10 difficulty range, but overall it's less pervasive than in most puzzle books.

I think this is why reviewing your own games religiously is of utmost importance.

1. You get familiar with patterns of thought, or where your thinking went wrong. I have noticed a pattern with my flawed thinking, and the kind of positions I tend to foul up.

2. If your narrow your repertoire, you begin to get similar positions over and over.

Again I think it goes back to using your knowledge as a jumping off point, and recognizing when you need to start thinking for yourself.

I don't remember the game, but Karpov lost early. The author of the book I was reading said it was probably because he mistook his position for another similiar one that occurs in a certain QGD line, and failed to realize the subtle differences. So it happens to GM's just far less than it happens to us.

It goes back to Rowson's point that playing chess is a skill rather than a bunch of "knowledge" to be acquired.

Similarly, I know a lot about football. I'd probably make a good coach, but I can't play the game well, because I don't have the physical skills to be successful at it. Sure I have chess "knowledge" but I have not yet (I hope) developed the appropriate skills to do it well.

That's my take on it.

11/20/2008 02:28:00 AM  
Blogger chesstiger said...

So may we conclude that after some pondering of a chess position our brain puts itself in a box to narrow itself to one thing and one thing only? Or is that to simplified to much?

11/20/2008 02:56:00 AM  
Blogger Blue Devil Knight said...

Wang: I think it depends how bad you are at tactics. Everyone should probably learn the basic mates (back rank, smothered, etc) really really well and train themselves to see the simplest one-move tactics. After that, though, I tend to agree with you. Many puzzle problems, while they are from real games, are about as useful as a composition for me. The thing to do is compile a bunch of tactics from games in your opening (or preferably from your actual games!).

11/20/2008 12:14:00 PM  
Blogger Blue Devil Knight said...

CT: it depends how challenging the position is. If it is easy, you won't be biased. If it hard for you, the bias effect will be very strong. I didn't mention this, but it was part of the study.

11/20/2008 12:16:00 PM  
Blogger Temposchlucker said...

For a cognitive bias I can imagine some Darwinistic cause. I mean, when you have seen a bear, you must be careful, there might be more around the corner. A bias with emotions is based on a preference for how the truth should look like. In the end it leads to some kind of fundamentalism.

If you think something bad will happen when Saturn has a square aspect with your ascendant, you will find abundant proof everywhere. And you will feel pity for those who think you are nuts.

11/20/2008 01:32:00 PM  
Blogger Polly said...

BDK: When I'm in an ending that's kind of drawish, but the opponent is playing on because of a time edge I try to be much more aware of simple tactics that can lose material. Going back to the knight fork. We know a knight will move from one color square and attack squares on the opposite color.

If we chase a knight that has given us a check from a dark square, when it moves it's going to land on light, but it will be attacking dark. So let's suppose we have a King sitting on e6 (light square and our queen is sitting on h8 (dark). No knight korks can occur there. Let''s put a knight on f4 (dark) giving check. If we attack the knight by advancing to e5 now both pieces are on dark squares and their close enogh that when the knight moves away to g6+ he's hitting the queen and king. If the king had approached the knight by coming to f5 the knight can only attack the queen with the g6 move.

If we're more aware of these situations then it's easier to kind the right king move. I'm painfully aware of this having done such a thing, that cost me a bishop in an ending with pawns on both sides of the board and minor pieces knight & bishop versus the bishop pair. Several moves later being totally rattled by the first fork I walked into the second fork. Within 6 moves I had tossed my bishop pair by walking into those forks.

Being aware of those are types of patterns help us. We will see the patterns when looking at fork problems. It's a matter of processing it in such a way that it can be used both offensively and defensively.

11/21/2008 02:53:00 AM  
Blogger Blue Devil Knight said...

Polly: I have definitely lost too many simple endgames by not looking for simple aligned pieces, pieces forkable by knights, etc..

When against a knight first and foremost to reduce thought just put my King and pieces on opposite colors (and try to get a single diagonal square away from the Knight, which means he has lots of moves to reach me).

11/21/2008 09:19:00 AM  

Post a Comment

<< Home