Higher level of abstraction
The conclusion of our former discussion can be summarized as: in order to transfer knowledge learned from one position to another, system 1 (unconscious) must be fed by analogies with a higher level of abstraction. If you think deeply about it, it makes perfectly sense. But at first glance, is sounds like gibberish. How to make it more concrete? (which sounds as less abstract 😁 ). In other words, what the heck are we talking about??
 |
| White to move |
r5k1/pp3Bpp/2n3n1/6P1/3q4/B5Q1/P4PKP/4R3 b - - 0 0
[solution]
I have encountered this problem 9 times over the years. As you can clearly see, I have learned nothing from it. (tijd = solving time)
The black king is out in the open. The main question is: how can we prevent him from skedaddling into obscurity again?
This abstraction can be used in many, many positions.
For this particular position, the way to solution can be concretized as: how to prevent Kg8?
For me this is a thinking pattern example
ReplyDelete1. King safety? (.. Pawn promotion Material)
2. Find all checks
And i made Qb3+ even without recognizing its "checkmate", i just saw that bishop and rook did attack many squares around the black king
A more verbose explanation than Aox. . . with essentially the same result.
ReplyDeleteWhite is to move.
Black is ahead in material, which eliminates material gain as the lietmotif. There are no Pawn promotions available. Ergo, the position is likely about checkmate.
The Black King is vulnerable out in the open. Perhaps we could look at it as a "king in the box" problem? We KNOW that in order to checkmate, we have to control or block every “escape” square around the King. There is presently only one "escape" square - on g8. The White Bishop is on the wrong color AND has to control the f8-square, so eliminate it from consideration. The White Rook has to control the e6/e7/e8 squares, so eliminate it from consideration. The g5-Pawn covers the f6-square. By process of elimination, the White Queen will have to do the job. Limit our attention to White Queen moves. 1. Qc7+ does NOT cover the g8 square, so eliminate it from consideration. 1. Qd5+ is eliminated by the Black Queen just capturing the White Queen, so eliminate it from consideration. Where else can the White Queen move to control the g8-square? AHA! 1. Qb3+! An extra "bonus" is that the White Queen also checks the Black King, making it a forcing move. What does Black have in reply? Nothing useful. 1. Qb3+ Qd5 2. Qxd5#.
The "combination" of "king in the box" and the process of elimination gets us quickly(?) to the solution.
There are no right or wrong ways to do this, of course. Everybody has to create his own abstract reference model. An abstraction is always broader, that is how it connects the current position with other positions. That's the transfer, where we are looking for. At the same time, abstraction looses details. Which happens to be the price for going broad.
ReplyDeleteThe abstraction is the vehicle of transfer. During the feedback fase of solving a problem, you lay the connection between the current position and the more abstract idea. In the hope, that when you come across a lookalike position, system 1 will be triggered to feed system 2 with the right idea. System 2 will add depth again, by trying to falsify the suggestion of system 1. The method of system 2 is calculation.
I just renewed my Chess Tempo Gold membership. I thought I would browse the Forums. I selected Chess Tempo Forum -> Chess Tactics -> Chess Tactics Discussion to begin reading. I started on the last page (page 44), which is the oldest posts, and slowly worked my way back toward the present. I only selected posts that seemed relevant to our discussion here. I did see some very interesting posts by AoxomoxoA and munich regarding tactics training. Really good stuff, guys!
ReplyDeleteOne of the most interesting posts by munich was on page 14; Topic: sorting patterns/ patterns useful to know/ typical patterns. Here's how munich described the project:
With this thread I want to start a collection of puzzles that belong together (have the same/ similar pattern). I hope that people do not chat a lot inbetween, but mainly use it to post patterns. (=If you dont contribute a pattern, please dont post).
This seems to be a similar idea to using analogues (similar patterns), collected under an umbrella term (a "catchy" name for an abstraction) as the basis for studying tactics.
As a result of this discussion, I remembered one of my "shelfware" tactics books (that I haven't thoroughly read because it seemed too far "above" my current skill level): The Tactician's Handbook by Victor Charushin. The original book consisted of 7 volumes, each about 100 pages long. GM Karsten Müller combined all of the volumes into one book, checking everything with the computer, and adding instructive exercises. The book classifies 6 significant tactical patterns, giving example games and game fragments which have those patterns in them. He also shows how to set up the patterns and how the patterns can vary in practice. The original book relied on a lot of examples from Kasparian's puzzles; Müller tried not to use many examples from a single source.
The 6 patterns are named:
1. Alekhine's Block
2. Combination Cross
3. Domination
4. Lasker's Combination (the double Bishop sacrifice)
5. Mitrofanov's Deflection
6. The Steeplechase
7. Less Common Combinations (such as the smothered mate, castling to win, the staircase maneuver, and famous mating patterns like Anastasia's Mate, Reti's mate and Légal's Mate)
Collecting and classifying “standard” (common) patterns is very useful for providing a training framework. Giving a descriptive “name” to these patterns helps in tying the examples together IN MEMORY so that System 1 has a retrieval mechanism.
It’s like having a “coat rack” on which to “hang your hat”. Hmmm, where have I seen that metaphor used previously?!?
Blokh is giving 1000++ 5x5 puzzles to classify tactical patterns in ct-arts
ReplyDeleteI have CT-Art 4.0 and have worked through many of those puzzles. I noticed that some of the 5x5 "puzzles" are localized (5x5) variations of the actual training problem, while others are somewhat functionally related but not the exact same underlying idea. In any case, the idea of a localized version combined with the full blow problem is worthy of note with respect to "seeing" analogies.
ReplyDeleteI looked under the Peshka subdirectory for files associated with the problems, and found nothing obvious. I've done disassembly of several significant software programs in the past, but no longer have the ambition (nor the tools) to do that again.