In the begin position of a problem I see certain salient cues. I know that there are salient cues that are not readily be seen, but which reveal themselves when you apply some logic.
On the other hand, there are salient cues that are perfectly seeable already in the begin position, but which I do not see because I'm not looking for them. Those salient cues are the ones where I can make progress because they form my blind spots.
 |
| Black to move |
2r3k1/Q4n1p/p2Brpp1/1p1R4/4P3/2q2P1P/6P1/3R3K b - - 1 1
[solution]
What I did see:
- target: Bd6
- defenders: Rd5, Rd1
- overloaded Rd1 => Bd6 AND back rank
What I did NOT see:
- double attack Qe5 => h2 AND d6
- back rank defense by Qg1
- counter attack white Qd7 => Re6 AND Rc8
Scenarios
- Exchange on d6 until a LPDO is left
- Chase K to h2
- Prevent Qa7 from interfering
With hindsight, my blind spots are staggering and amazing. A training method must focus on these blind spots.
I can't see what I'm not looking for. Initially, logic should guide my seeing. But I feel that is just a kind of side wheels for this position. Almost everything is salient in the diagram. Maybe only the difference between Qc1+ and Qa1+ should be found by reasoning. And the difference between 1. ... Nxd6 and 1. ... Rxd6. But then again, Qg1 as blockader is already perfectly visibible. When you look for it.
PART I:
ReplyDeleteThis is another ‘word salad’; sorry about that.
Temposchlucker asked a very important question (in a comment – 9 MAR 2024):
“We seem to have named all necessary elements now. Can we concoct a method from this which helps to acquire skill and prove it?
YES!
Cognitive research has shown that (regardless of what is considered to be a "chunk") that it takes at least 6-8 seconds of PERCEPTION, followed by up to one minute of ABSTRACTION [CONCEPTION], in order to establish a "chunk" in long-term memory (LTM). Any time shorter than that will NOT result in storage of a "chunk" in LTM. That was the basis on which I rejected MdlM’s Seven Circles of Hell approach, which was limited to tactical training.
Faster does NOT equate to better, but better DOES equate to faster. Sounds contradictory, but the relationship between faster and better is asymmetrical.
Let’s ASSUME that we are looking for a training method and a playing method that are composed of common elements. Remember: Train like you play and you will play like you train.
Think about the metaphor used by GM Botvinnik (Computers, chess and long-range planning) to suggest an approach to both training and playing. (This is repeated from an earlier comment, so bear with me; I “SEE” something additional to be learned from it.)
"A simile may help clarify this notion of the horizon and its determinants. Let us suppose that a parachute-jumper has come down in a bog [swamp] and wants to get to solid ground. The bog is wide; its edges are hundreds of yards away. How does our hero proceed, if he cannot find a clear path from where he is to where he wants to be? He cannot take in the whole plot at one coup and pick out the entire path . . . he must act soon . . . darkness is falling!"
"In all probability, he will inspect the bog in some given direction for the first five to ten yards, choose a path from hummock to hummock, as safe a path as he can find, and TAKE THE FIRST STEP. He will make the next step after a similar preparation. Our hero again inspects the five-to-ten yard horizon. It will already have changed because of his action. He accepts the solution to the second-step problem, and so on until he clears the bog."
"We could prove that ALL THE DEVICES USED IN CHESS WAR - attack (from both sides), blockade (by either side), and (mutual) retreat - CAN ALL BE ACCOUNTED FOR IN AN INITIAL MATHEMATICAL MAP CONSTRUCTED ON AN ATTACK CONTAINED WITHIN FOUR HALF-MOVES."
Given the context of the book, I ASSUMED this metaphor is about long-range planning. As usual, my limiting conception may have been too limited.
Botvinnik’s metaphor is mirrored in GM Tisdall’s suggested variation processing.
Think about standing in a swamp (a chess game) and trying to figure out what to do next in the IMMEDIATE FUTURE. Given the 64 squares and 32 (or fewer) pieces, as well as the interactions of those pieces, we very quickly “lose the forest for the trees” and we CANNOT keep all of it in mind simultaneously. Working memory limitations PREVENT us from “SEEing” the entire board at once. So, we MUST limit what we try to “SEE” at one time to something manageable. We already know that limit—from George Miller’s “7 plus or minus 2” rule.
[For clarification, when I use “rule” I mean “general principle,” not a hard and fast unbreakable law.]
PART II:
ReplyDeleteSuppose that Botvinnik’s metaphor and Tisdall’s variation processing are applied to both training and playing, and (perhaps) NOT limited to long-range planning. What would that look like?
We start with KNOWLEDGE. If you have no knowledge, you have no basis to proceed except by trial and error. Not just a nodding familiarity sufficient to categorize by labeling (for example, by putting a name to a feature of what’s on the board, like a potential fork, pin, discovered attack, etc.), but detailed knowledge of what to PERCEIVE in addition to abstraction/categorization sufficient to CONCEIVE (what we “SEE” as the basis for future action). We ZOOM IN (perception) based on a limited (but NOT arbitrarily limited) number of pieces (a local area of tension) and ZOOM OUT (conceive) of the interactions between those pieces without attempting to “SEE” the entire board (the “big picture”). We repeat this ZOOM IN, ZOOM OUT process as many times as needed to eventually conceive of the entire “big picture” – we KNOW THAT and we KNOW HOW. The “big picture” is NOT seeing the entire physical board but “SEEing” the concepts that are salient and concrete.
We acquire SKILL by approaching each and every position the same way, whether training or playing. This is not synonymous with a formal step-by-step logical “thinking process” as has been proposed by a lot of very smart people. Such “thinking processes” make perfect logical sense, but – they don’t work because we don’t have the capability to keep all those steps in mind while simultaneously keeping everything happening on the board in mind. That just ain’t possible!
If you carefully examine any general subject you will find a similar process that limits focused attention to one small, concrete area at a time. Over time, you add other similarly restricted views, broadening your perspective (and conception) until you have comprehended the entire thing – but even then, you do NOT “SEE” the entire thing at once except at a very high level of abstraction.
For example, consider the design of a Very Large Database system. (I’ve done this, so this is based on personal experience.) You start by gathering requirements, perhaps by talking to potential end users regarding the INFORMATION content that is required to accomplish their tasks. You get each user’s viewpoint or a small group of users’ viewpoints sequentially – because you cannot talk to all users at once. You ZOOM IN to the lowest level. After gathering sufficient data, you start abstracting it into a formal definition (ZOOM OUT). After defining a specific local area, you move on to a different (but related) area, and repeat the process until you have all the requirements. You may start with a top-down formal set of requirements, driven by the stakeholder(s) who are paying for the project, but you will still have to get down to the lowest level to verify that the system meets the requirements of all stakeholders. (BTW, this is what GM Rowson and others mean when they say that we have to “talk to the pieces.”)
PART III:
ReplyDeleteThe same ZOOM IN, ZOOM OUT process is required for development of any non-trivial computer program. I’ve written many programs that were tens of thousands of lines of code. Trust me when I tell you I didn’t start out with a clear “big picture” in mind except in the vaguest, most abstract category imaginable. “I need a VLDB database to keep track of all accounts receivable and accounts payable.” Sounds like (just part of) an accounting system to me, and if you’ve ever worked on one of those large systems, you will agree that design and implementation is a ZOOM IN, ZOOM OUT process repeated hundreds, if not thousands, of times across many organizations and people.
A single person can do it, but it takes considerable time. It took me nearly two years to design the US government’s Excise Tax VLDB database system working by myself with end users and stakeholders. As I completed the specifications of requirements for specific features, a rather large team of designers and programmers worked to implement those features. Thank goodness I didn’t have to implement the system myself!
How do you train to do something that complex? BY LEARNING HOW TO DO IT AND THEN DOING IT ONE SMALLER PIECE AT A TIME! Eventually you acquire the SKILL to design and implement extremely large systems that are far beyond the complexity that can be held in a single mind at one time (other than by using a very abstract label like “accounting system”).
The past few blog posts pointed me toward this conclusion. In each case, the initial focus is on one specific concrete aspect of the position. After clearly “SEEing” one area (ZOOM IN), we added another area (ZOOM OUT), look at it until we clearly “SEE” it, and then looked for interactions between those areas. We repeated that process until we had determined what to do (the “big picture”).
The same approach can be used during training and during playing.
PART I:
ReplyDeleteThis position is a perfect illustration of how to go about the ZOOM IN, ZOOM OUT process. This is going to be tedious to write out, but it takes very little time to go through the sequence in the mind’s eye.
As you note, there are some salient cues to start the process.
Black has the right to move first. That makes Black the potential attacker.
Salient surface clues:
WBd6 is B.A.D. [2:2].
White’s King may be somewhat vulnerable to checks on the back rank.
All of the pieces attacking/defending WBd6 constitute one local area of tension. Since that is the only surface-level local area of tension, consider it in isolation from everything else for the moment. Ignore any pieces that do not directly impact this local area as if they do not exist. One or two exchanges on d6 are possible, but apparently losing for Black if he initiates the exchange sequence. HOLD THAT THOUGHT! Whether this is true or not may depend on the further sequence of moves.
Black can capture first using either the Knight or the Rook. Try ONE of these alternatives, not necessarily in any order.
(1) Let’s try 1… Nxd6 first (just because it usually benefits us to capture with the lowest valued piece first).
Now mentally flip the right to move first and “LOOK” at this new situation from White’s point of view. Since White started out with barely adequate defenders for the d6-square, White cannot lose material from any exchange there, so “automatic recapturing” on the d6-square is the default action. ANYTHING WE DO ON “AUTOMATIC” SHOULD BE A TRIGGER TO SLOW DOWN AND “LOOK” FOR ALTERNATIVES!
Salient surface clues (in this new position after 1… Nxd6):
BNd6 is LPDO.
BRe6 is LPDO and if attacked, cannot move horizontally so as to continue protection of the BNd6.
The second clue is most important because it should TRIGGER a ZOOM OUT to “SEE” what other White pieces can be used to take advantage of BOTH clues. The WBd6 already went into the box, so the WQa7 is the only piece to look at.
AHA! ZOOM IN, adding the White Queen to the first local area of tension. The WQa7 can move to the d7-square, attacking both LPDO pieces.
Switch perspectives back to Black. ZOOM OUT. What can Black do to protect his two pieces? Nothing. Moving ether Rook into a mutual protection position still leaves the BNd6 en prise. White will regain his material by capturing on the d6-square with no additional vulnerabilities such as the back rank.
Evaluate this position as quiescent and at least equal for White. Not a satisfactory solution for Black.
(2) ZOOM OUT to the original position and try 1… Rxd6.
ZOOM OUT in this new position and switch perspective to White.
The potential vulnerability of the BRe6 is erased. [If we had chosen 1… Rxd6 to analyze first, there would have been no necessity to evaluate the BRe6 as LPDO. Lucky flip of the “first move” coin.]
White is virtually forced to recapture with 2. Rxd6 on the d6-square; otherwise, he is down a piece for a pawn and nothing to attack to compensate. Black is virtually forced (for the same reason) to recapture with 2… Nxd6. White is then forced to recapture with 2. Rxd6 on the d6-square. White is then a pawn up.
HOWEVER, the White back rank has now opened up for possible checks by the Black Queen. WE ARE N-O-T IN A QUIESCENT POSITION! We CANNOT stop here!!
This is a stepping stone position.
PART II:
ReplyDeleteMentally erase the exchanged pieces (Wbd6, WRd5, BNf7, BRe6) and LOOK with fresh eyes. ZOOM OUT and “SEE” what Black can do now.
Salient surface clues:
WRd6 is LPDO.
White King is vulnerable on the back rank to checks on either c1 or a1.
Black can check on either c1-square or a1-square.
(1) A check on the c1-square by the BQc3 allows White to interpose the WQa7 on the g1-square and Black’s “attack” has no other targets. Evaluate this as good for White and back up and look at the Black alternative 2… Qa1+.
(2) A check on the a1-square still allows the White Queen to interpose BUT now the BRc8 can drop onto the c1-square, pinning the White Queen to the White King. SERENDIPITY! Evaluate this as winning for Black. Already this variation seems good for Black which is an encouraging sign to KEEP LOOKING!
In response to 2… Qa1+, the White King must move to the h2-square 3. Kh2 to avoid losing the White Queen for a Rook. Forcing moves are always a good sign that the opponent is marching to our drummer!
Switch to Black’s perspective. ZOOM OUT. Are there any vulnerabilities in the White position? YES! The Black Queen can fork the White King and White Rook with 3… Qe5+. Evaluate the position after 4. Kg1 or 4. g3 4… Qxd6 as quiescent and winning for Black.
We are now at the end to the “calculation.” Black is winning by force.
There are a couple of things that we have to do for this process to work. Crucially, we have to switch perspective after each player’s move(s) and ZOOM OUT to “SEE” that new, changed position in the mind’s eye. Limiting the new perspective to a specific local area enables us to “SEE” what is going on at that moment. We’re NOT trying to “SEE” everything on the entire board (pieces, square and interconnections) at one time, and we’re not trying to SEE” multiple variations at once. We’re not trying to create a list of candidate moves, or a list of imbalances, or follow a fixed sequence of logical steps in a formal “thought process.” We’re just LOOKing at what seems most important from each player’s perspective at each moment and gathering information as we progress logically through the potentialities of the original position.
The hard thing to do is to mentally discipline our mind to “SEE” as little as possible at each step, but to consider everything that is possible for both players.
I tried to figure out what caused the blind spot. There are a few issues: Laziness to see the targets of the opponent. Laziness to extend my vision to double attacks and discovered attacks.
ReplyDeleteI don't like the word laziness very much in this situation. Since I am usually not so lazy. I certainly spend a few minutes on it. It is more a lack of good habits.
The laziness manifests itself as looking in trial and error mode. Had I in stead guided my vision, I certainly would have found the missing elements.
The goal is clear: fix the blind spots.
The method is: discipline to follow my list of salient cues until it becomes automatic.
Be alarmed when only looking in trial and error mode.
No one who was 'lazy' would expend so much dedicated time and energy over 25 years pursuing the will-o'-the-wisp of adult chess improvement.
ReplyDeleteI think GM Aagaard's observations about calculation may be applicable:
Calculation = The process of finding what you do not "SEE" automatically. How is this done? SLOW DOWN the flow of variations running through your head and LOOK for alternatives. The purpose of calculation is to aid decision making, NOT to SEE everything.
Salient cues ('clues') cannot exist across the entire board. In most positions, the number of salient cues to be SEEn is severely limited. I guess-timate that the total number will normally be 3-7, well within STM capacity.
Your 'blind spots' are not a result of missing knowledge; you already have all the technical knowledge required.
Try GM Aagaard's advise and SLOW DOWN. When you start looking at one specific feature (such as the B.A.D. WBd6), allow System 1 to explore the ramifications and connections to that cue. In short, don't shut off the input from System 1. Intentionally follow the LoA departing from and arriving at that specific location so that you can grasp ALL the associations that can be derived from focusing on that one local area.
GM Beim refers to the discovery of additional cues as "resulting moves." You CANNOT SEE resulting moves until you start fanning out your attention. Think of the process like the ripples created by a rock thrown into a pond. The primary waves are very strong, gradually weakening as the ripples spread out.
Or as a different (familiar) comparison: Think of it as a programming problem. You identify a central starting point (the key feature) as the salient point. You then proceed to iteratively elaborate more and more points (with a wider view) which are NOT "SEEn" from the initial central viewpoint.
You cannot be a successful programmer unless you have the mental discipline to follow this type of process.
In one of Gerald Wienberg's books (The Psychology of Programming), I read that the difference between amateur and professional programmers is laziness. The amateur designs and implements a program to solve a particular requirement. He then moves on to the next challenge—ALWAYS STARTING OVER FROM SCRATCH EVERY TIME. The professional, on the other hand (being 'lazy'), extracts additional knowledge and skill from every program with the idea of using that knowledge and skill in future programs WITHOUT RE-INVENTING THE WHEEL FROM SCRATCH EVERY TIME.
Do you have a 'toolbox' of algorithms, small programs and code fragments that you use to create new programs without starting from scratch? When I worked as a programmer, I had a toolbox running into several thousand lines of code. Whenever I had to design and implement a new program, I always checked to see if there were modules that I could adapt and glue together to solve the problem. If the programming language changed, I adapted the toolbox algorithms and rewrote them in the new language. [I eventually ended my career with around 35 distinct languages that I was fluent in.] Only if the problem was completely outside of my experience (my toolbox) would I start investigating alternatives from scratch. In every case, I would ADD that new knowledge and skill to my toolbox. I was much more efficient in developing programs quickly and with fewer problems than a lot of my co-workers for that very reason!
Most of what we do when playing chess can be acquired and used the same way.
BE 'LAZY' - in the best possible way!
Leverage and apply what you know; that is SKILL.
I'm fluent in 28 languages. Over the years I noticed a shift from programming to talking. People hardly need more programs. Everything is already programmed by someone out there. I just connect it. But usually they need to understand their problem. My speciality is to connect the programmers with the sales department. Since sales people speak the simplest language of all: what does it cost and when is it ready.
ReplyDeleteNowadays I let ChatGPT do the programming and the talking.
My favorite tool is the loop. You can do anything with a loop.
ReplyDeleteOne of my favorite "tools" was the awk programming language. It is an interpreter with a loop as its base—it reads one record of text, processes it using one or more pattern matches (based on regular expressions), then repeats the loop. I designed a program writer to read a definition of a programming language and a series of date-related textual regular expressions, which then created a new awk program that would search for potential Year 2000 problems. The program writer was about 20,000 lines of awk code, which would produce about 25,000 lines of awk code. (The typical awk script is usually very tiny, only about 5-10 lines of code.) AFAIK, it was the only working Year 2000 search program that could handle any and all programming languages without any modification.
ReplyDeleteSince we both like loops, I suggest using a Read-Eval-Print-Loop (the basis of languages like Lisp, Scheme, Racket)) to “SEE” through the blind spots.
RECOGNIZE one or more salient cues in the current position. (The “current position” may be the actual position or a mental projection into the future; that matters not.)
EXAMINE/EVALUATE the ramifications (lines of attack, interconnections, additional “resulting” salient cues, etc.) associated with the specific local area of tension defined by the salient cue.
PROJECT the ideas that you SEE outward from the local area of tension along a principal variation.
LOOP: LATHER, RINSE, REPEAT until there are NO more salient cues AND the investigation reaches quiescence in ALL variations.
REPL should be fairly easy to remember as a mnemonic and as a process.
KNOW HOW is as important as KNOW THAT!