Monday, July 30, 2007

More on classifying tactics, with an experiment

For those interested in the Circles, you might check out allegroknight's new discussion of how he might modify them. This is something a lot of you have a good deal of experience with :)

My previous post was a bit unfocused, as I wasn't exactly sure where I was going. Now that I'm there, I can be more clear.

The search for a practically useful and relatively exhaustive scheme for classifying tactics was conducted fairly extensively by King of the Spill and Temposchlucker . I wonder what people think of those posts now? Tempo, what do you think of your scheme? Here is a great line from Tempo's post: "King of the spill tried to develop a theory about tactics. If such theory will help to improve your chess is another matter, but I see developing theories as a means to clear up the mess that usually resides in the head. A sort of defragmentation of the hard disk so to speak."

When I wrote the post, I suggested that with longer combinations, things get too messy and any classification scheme will be so simplistic as to cloud the precise concrete calculations needed to find answers. However, in the comments loomis gave a nice analysis of the included position in terms of the simple categories. His analysis not only helped me understand the position, but makes me think the understanding gained would actually help me find similar combinations of tactics in the future.

Heisman says that complex combinations are words built up from the alphabet of basic tactics. If most combinations can succumb to a cogent analysis as loomis provided, that would be sweet, and suggest that when working through tactical problems, it will really pay off to take the time to struggle to find the solution not only in terms of variation crunching, but to take the time to consciously decompose it into the basic elements, and more generally to construct a narrative about the position that would help you see similar patterns in the future.

Indeed, the idea that constructing a personal explanation of particular positions can help you see them in the future has been directly confirmed by a simple experiment. The paper, by de Bruin, Rikersa and Schmidt, is called The effect of self-explanation and prediction on the development of principled understanding of chess in novices, and was published in 2006.

As described in the paper, researchers took a large group of people that had never played chess, and taught them how to mate with a rook and king versus king. They broke them up into three groups: group one simply observed the solution on a computer. The second predicted the moves of the computer before shown the move of the computer. The third group predicted, but also were required to generate their own explanation of each move the computer generated (e.g., 'This move cramps in the King'). It was found that when subjects were given rook mate problems of their own to solve, in future sessions, the third group (the one that generated the explanation) performed better than both other groups, while the first and second groups performed the same. If the result is general, that suggests simply guessing moves is not enough: you should also try to understand and explain the moves once you see them (something a more experienced player is likely to do anyway). Perhaps this suggests you construct your own explanation before reading the explanation of the annotator, or even use the J'adoube method of just using Rybka to analyze the position and then do the hard work of explaining its moves yourself.

On the other hand, their study was on a very simple chess problem, and with pure begineers, so we can't be sure that the results will generalize to more complicated problems in more seasoned players.

23 Comments:

Blogger funkyfantom said...

Yeah, the higher-level patterns you can use, the better.

Like, this combination wasn't just a discovery, double-check and somthered-mate it was an example of Philidor's Legacy.

We have the Philidor Rook Ending, various mates with names ( Epaulette Mate) and loads of nameless combinations that never get names.

It is often very arbitrary, to me at least, why some combination templates get names while others don't.

7/30/2007 01:15:00 PM  
Blogger Blue Devil Knight said...

Good point FF.

If we were philosophers, we would argue about whether there are tactical 'natural kinds', or whether particular classification schemes are simply useful conventions. Philosophers of biology like to argue about whether a species is a 'natural kind,' for instance. Frankly, I'm not sure what a natural kind is. Perhaps something that exists independently of human conceptual systems. This question becomes particularly vexing when it comes to chess.

Joking (i.e., philosophy) aside, I am really most interested in finding the most useful and comprehensive classification scheme for the simple tactics. I am starting to get convinced that most combinations will involve some mixture of such elements.

7/30/2007 01:40:00 PM  
Blogger J'adoube said...

J'adoube Redux?

Oui!

ROLFLMAO. . .

BTW, congrats on your recent triathlon finish. As a retired triathlete, I know how hard they are. . .

7/30/2007 01:48:00 PM  
Blogger Blue Devil Knight said...

J'adoube: why am I not surprised that you have done triathlons? Do you ever do anything half-assed?

7/30/2007 02:02:00 PM  
Blogger Temposchlucker said...

It still stands out that you can shoot a single target by trapping it first, or by attacking two targets at the same time while only one can be saved.

Usually a target is a piece of wood or a square. But it is important to broaden this vision and to replace "target" with "any advantage". If it is advantageous to trade the bishop for a knight, any tactic that accomplishes this is good. This are not the tactics we exercise in problemsets. But they form 95% of the tactics in a game.

Alas, I have to wait after the tournament before I can adress this subject further.

7/31/2007 02:03:00 AM  
Blogger funkyfantom said...

As many of us are aware, "Art of Attack in Chess" is full of these meta-patterns :

Greek Gift, Classic Double-Bishop Sacrifice.

Many GMs cite this book as a must read.

7/31/2007 09:33:00 AM  
Blogger Blue Devil Knight said...

FF: I have started that book twice but am not ready for it. Note that Tisdall includes the main mating nets from Art of Attack, in a convenient appendix. Also, 'How to beat your dad at chess' and 'The art of checkmate' both also provide a whole slew of practical mating nets.

Indeed, it was reading 'How to beat your dad' that triggered a transformation in my approach to chess, making me much more aggressive, willing to seek out sometimes speculative attacks. It is one of my favorite books.

It is probably easiest to get a feel for combinatorial thinking when it comes to mating nets. Plus, moves are forced, so there is less chance of error.

7/31/2007 10:17:00 AM  
Blogger funkyfantom said...

The interesting thing about these meta-patterns is that a GM can usually sense correctly whether or not it pays to calculate out all the variations of a potential Greek gift situation, for example.

The novice will not even notice the pattern. The intermediate will either recklessly make the sac because he "feels aggressive" and his opponent will likely botch the defense, or he will just play a "safe" move so as to avoid the mass of variations needed to be calculated.

The expert will go through all the calculations and determine whether the sac is sound or not, but will lack the intuition about how the calculations were going to turn out that the GM has.

7/31/2007 11:36:00 AM  
Blogger Wahrheit said...

We might begin with the base concept upon which all "higher order" classification schemes rest--in most practical cases a successful tactic results in a position where there is an unanswerable double threat, that is a position in which at least two moves are threatened, both of which result in a significant improvement in the tactician's position (usually the win of material), and the opponent's various moves can only stop one of them.

I well realize that there are situations of forced "mate in x moves" where the double threat does not apply, but outside of gross blunders by the opponent this type of position can only come about when a big advantage has previously been gained by the use of double threats. I read over the King of the Spill and Tempo posts and Loomis' comment and in the cases they cite it all boils down to an unanswerable double threat.

The practical use of this simply that looking for a move that does two things at once is a great way to get your "candidates." For instance, if a moves stops the opponent's threat and improves our position in some way it's a strong candidate.

By the way, checking the king is a major "threat" in this scheme, it threatens to capture it next move and of course is involved in a large proportion of combinations as it is the most forcing move in chess. Any check that also threatens something else is a killer unless getting out of check solves the other problem as well.

7/31/2007 06:40:00 PM  
Blogger Loomis said...

This post seems to parallel what Tempo has written recently about making a conscious effort. It's better to put your mind to an active test than try to absorb knowledge passively.

7/31/2007 06:59:00 PM  
Blogger Blue Devil Knight said...

wahrheit: How do you think traps and skewers and pins fit in with the notion that all tactics involve two threats? They seem to involve one threat.

Especially traps, but in skewers and pins, I guess you could say it is an indirect double attack against both pieces along the line of the pin or skewer; that makes sense in the case of skewers, but with pins it seems a somewhat forced way to describe it: also, pins are not the most deadly tactic: it is more a forced decrease in the activity of a piece (a reduction in its freedom, even if it technically still has the same mobility, or number of possible squares to go to).

Now when a pinned piece is attacked, that is a tactic that involves two threats!

8/01/2007 12:04:00 AM  
Blogger Temposchlucker said...

With a trap there is a problem with space, so there is no escape. As said, a mate is just a trap of the king.

With a "duplo" attack against two targets there is a problem with time. There are two moves needed to bring both pieces into safety.

These are the only two ways to force an advantage.

The difference between a skewer and a pin is only the value of the pieces.

A tactic is a forced advantageous trade. Usually advantegeous means: the value of the pieces invested is lower than the value of the pieces gained. Since the technique is the same, I propose to use the word tactic for any advantegeous trade, not only for material gains.

The example I used earlier: If I think it is advantegeous to trade my bishop for a knight, I can only force this by 1. trapping it or by 2. a duplo attack.
That's how it happened in my game: I pinned his knight against his king.

8/01/2007 01:48:00 AM  
Blogger Wahrheit said...

Yes, a pin is usually only a really serious threat if the pinned piece is unguarded, or if it can be hit a second time--then it's an unanswerable double threat...the typical Bb5-Bg5-type of pin in the opening is often more of a positional threat, limiting mobility and threatening doubled pawns, but I'd still call it a double threat, to the pinned piece and the one behind it. Only if there are other factors is it really deadly.

The example from your previous post is more subtle, but you'll note that every move checks and threatens to queen the pawn, or in the case of Be6 it threatens to check with the rook and queen the pawn the following move--still what I'd call an unanswerable double threat, just not both on the same move.

Interferences and deflections often operate this way, threatening something now and something in the future, but I believe it can all be classified as an unanswerable double threat (not just double attack).

I don't know if this is exactly what you were driving at or whether this is the most useful way of looking at the problem of finding tactics--whatever works for the individual in terms of sticking something in the long-term memory is the way to go!

8/01/2007 01:57:00 AM  
Blogger Burmese Rat said...

I suggest you stop turning chess into one of your rat neurotic experiment debaucheries and just play the damn game.

8/01/2007 08:09:00 AM  
Blogger funkyfantom said...

BDK- I'm getting the impression that someone around here is suffering from multiple-personality disorder.

(;-)

8/01/2007 09:22:00 AM  
Blogger Blue Devil Knight said...

Tempo: I will have to think about this some more. It feels forced and unnatural to focus so much on multiple attacks. But maybe it's because it's not how I usually think about them. I have never really understood King of the Spill's classification, so I'll check it out, and then think about how it relates to yours.

FF: It seems I have an infestation of rats!

8/01/2007 10:34:00 AM  
Blogger The Rise and Shine Good Knight said...

I never thought I'd say this, but I'm gonna have to agree with the rats. It seems like you spend all your time studying how to study chess rather than actually studying it.

8/02/2007 09:04:00 AM  
Blogger Blue Devil Knight said...

Rise and Shine: I write about thoughts I have as I do the Circles. If it isn't useful to you, that's unfortunate, but it is very useful to me, precisely because it is grounded in practical struggles, attempts to learn these problems faster (some of them are very difficult involving up to 14 moves).

At any rate, conceptual undertanding and practical study go hand in hand, which was the point of my comments on the previous post.

8/02/2007 10:32:00 AM  
Blogger transformation said...

TRSGK:
it must be the easiest thing in the world to go around and tell everyone how what they do is inadequate. so i ask, for you, what IS adequate?

what do you wish to create, not what do you wish to downgrade.

i am a classically trained architect, kind of like being a modern era engineer who can use a minicomputor or work station, but whom had to take ardous exams being able to quickly use a slide rule. not all bad. you learn the real core of your craft, and how to do manual alogrythms. you can use all the advanced stuff, but you know the real foundation.

the thing that i started to notice, once i got into professional practice, was that all painters and architects did was criticise how bad some xyz famous architects work was.

and the more and more i saw this, the more did i really start to appreciate simple, humble buildings. the clean, honest, unpretending ones down the street from each of us. its not GA Magazine, or some slick swiss gloss, or PA or Architectural Record or even Fine Home Building.

so lets just give credit to ANYONE at blogger who studies, studies hard, studies correctly.

when someone goes around and says how others are wasting their time, are they not wasting their own time doing so?

as thoreau said, "if we could be inside the average persons head, would we not see specticles such as in the most amazing tale of the arabian knights?" so BDK is his own Venus or Neptune, and we each are planets, moving at the speed of light, and it is a miracle.

so, tell us if you can TRSGN what it is that you wish to create, not how dubious is someone elses creation or effort or way. maybe you are talking about yourself here, and not BDK?

ne saying is the easiest thing in the world, and we are really talking about pleasure, learning, and improving, and, most of all, civil sharing among like minded persons in constructive exchange.

8/02/2007 03:34:00 PM  
Blogger Blue Devil Knight said...

DK--the way I look at it there are three types of comments on blog posts. First, relevant comments that flow with the topic. These can sometimes get quite acrimonious but it can be very productive. Second, comments that are off topic but a kind of meta commentary (e.g., "You think too much, just play"). The third is just gutter sweat like rat fucker or whatever his name posts.

The first type is what every blogger wants: it makes for a fun and interactive blog, people learn and all runs smoothly. Even if discussion gets heated, it is on topic and usually it is a positive. The second type is inconsistent. Typically I find them unhelpful and annoying, but ultimately it is always good to reflect on one's study program (ironically, RSGN's post triggered even more reflection on how I train, precisely the thing he is admonishing me to stop doing: what other effect could it have, though?). A little bit of this type of post is OK, especially if it seems to come from a good place (not just a potshot or thoughtless mean comment). The third type is just noise, and if they get too voluminous comment moderation must be enabled :)

Enough meta blogging. I'm getting my ass kicked by CTB level 5.

8/03/2007 02:25:00 AM  
Blogger The Rise and Shine Good Knight said...

Since this is a community of people dedicated to chess improvement, I would not consider replying to a fellow blogger's post a waste of time. I'm simply debating the usefulness of spending an inordinate amount of time thinking about something your mind does naturally. There's no need to classify tactics unless you're writing a book on them. All you have to do is become familiar with them through study and your brain will classify and recognize them for you. This is a group of people dedicated to chess improvement. This is not a group of people dedicated to pontificating about a bunch of pseudo-academic gobbeldygook. If that kind of stuff floats your boat, then have fun with it. It will not make you a better player.

8/03/2007 09:30:00 AM  
Blogger Blue Devil Knight said...

RSGN, your anti-intellectual tone surprises me, partly because as I've now said three times, these posts were born as I started to hit harder problems that didn't fit into the simple beginner schemes. I was wondering if there was a more general scheme that experts use, that they found just as helpful to them when the started to improve, as the basic 'forks' etc were to beginners.

So the whole point of the posts was that I was asking how much detail a classification scheme needs to be useful in practice. It is a practical question, but that doesn't mean the answer is simple, or that the answer won't involve some sustained thought. And I wasn't called the 'PBS of chess blogging' for nothing. If you get bored or impatient that is one thing, but your opinion about tactical classifications is another.

8/03/2007 12:31:00 PM  
Blogger Blue Devil Knight said...

Taken at face value, RSGN is saying that we shouldn't worry at all about classifying tactical patterns. This seems obviously wrong, for a couple of reasons.

First, the experiment I discussed shows that people do better in the future when they construct explanations of positions. This is not pseudo-academic. It is data. You can argue about its interpretation, but there it is, data about chess improvement that could have practical implications. Simply playing is not as good as playing with an eye to explaining what is happening in the position. This is not new: instructors do this all the time, Tisdall in his book (Improve your chess now) says it is extremely helpful, in practice, to construct internal narratives at a general level about the position

Second, at its extreme you wouldn't have us teach beginners what a fork is, or teach them that, when you have a piece pinned, attack it. No need to learn all that, as it will happen automatically if you just play enough games. You might be right (and there is something to be said for simply playing a lot). But there is a middle ground. If it helps someone to learn tactics faster to have the basic conceptual toolbox, then they should learn it.

The question is, how far, if at all, do we need to go beyond the basic classifications in practice? I concluded that we don't need to go far at all, as all combinations seem to be built up from such elements.

So RSGN essentially agrees with my conclusion, I would assume, but somehow attacked me in the process. I'm confused.

8/03/2007 12:33:00 PM  

Post a Comment

<< Home