The Trick in action
Yesterday I participated in a simul. There were 14 participants and the simul was given by a Fide Master.
I didn't exactly measure it, but at average he needed 4 seconds per move. Which meant that I had to make a move every 56 seconds. I could keep up into the middlegame, until I was compelled to shift gears from System 1 to System 2. All of a sudden, 56 seconds was not enough to make a reasonable move anymore. So when he stood at my board, I was obliged to play a move which I knew wasn't good. It costed me a pawn and it caused weaknesses in my position.
His plan was simple:
- in the opening: develop and do nothing weird
- in the middlegame: trade everything that is dangerous
- wait for my mistake
- trade off every piece and convert the game to a win
When the first participants began to drop out, matters start to become worse. He needs lesser and lesser time before he stands in front of you.
A master is somebody who can play the one- and two-movers in every part of the game.
An expert can only do so in some parts of the game.
The difference between a master and a grandmaster, and between a grandmaster and a super-grandmaster is a matter of scale. I can't see that of course, because it is out of my league, but it has to be. What else can it be? I gathered enough circumstantial evidence to prove that beyond doubt. My doubt, that is.
Mastering one- and two-movers in all parts of the game is enough to become a master.
This means that you can study any game to help you absorb the one- and two-movers. As long as you know what you are after.
ReplyDeleteYour experience mirrors my own (from the viewpoint of the person giving a simul). I noted in a comment back in 2015:
ReplyDeleteI've done simultaneous exhibitions (nothing on the scale of Judit Polgar; that's not what I'm trying to express). I had more skill and knowledge and experience than my opponents. Consequently, I could "SEE" things they could not see, and so I used that advantage to win. I didn't try to experiment with the openings or take unnecessary chances in the middlegame. I generally allowed my opponents to beat themselves. I once played GM Arnold Denker in a simultaneous exhibition. He did nothing out of the ordinary; he just waited for the inevitable mistake/oversight (even though it was not an egregious error) and then just ground me down with technique.
You said:
Mastering one- and two-movers in all parts of the game is enough to become a master.
Corroboration of GM Botvinnik's statement:
When Mikhail Botvinnik lost on first board during the 1955 Soviet-American match, the world champion explained the result simply, "It shows I need to perfect my play of TWO-MOVE variations."
"It’s the little things that count, hundreds of them." — Cliff Shaw
While contemplating the implications of your recommendation to absorb the one-movers and two-movers, I realized that there is a direct connection to pattern recognition and mental simulation using action scripts (calculation).
ReplyDeleteI ASSUMED that pattern recognition of “chunks” refers to TOTAL patterns, not the individual components which comprise a total pattern.
I ASSUMED that the associated action(s) at each step had to be embedded into the overall “chunk” almost in a linear sequence.
For example, I assumed that the Greek Gift sacrifice pattern had to be absorbed in its entirety, including all “variations on the theme.” All of the ramifications (alternatives along the way to the end result) had to be included in the overall “chunk” in order to consider it firmly embedded in LTM. From start to finish, the entire process and intermediate structures had to be absorbed AS A SINGLE COMPOSITE UNIT.
In retrospect, that seems naive at best. Given the complexity of such a pattern, “obviously” there must be smaller units within the larger composite pattern. These smaller units are the one-movers and two-movers.
Klein’s Recognition-Primed Decision model is primarily composed of pattern recognition and mental simulation. Again, an erroneous assumption is that an overall pattern is recognized and then the mental simulation of associated actions ASSOCIATED WITH THAT SPECIFIC LARGE PATTERN takes place.
I think Tisdall’s variation processing is closer to reality. There is some surface “cue” that starts the investigative process. It could be the prerequisite(s) for a one-mover or two-mover which is “obvious” in the position. That smaller unit is simulated, advancing deeper into the essence of the position. This next position is recursively investigated using the same approach: identify another salient one-mover or two-mover, and advance again. Perhaps at some intermediate stage, a larger composite “chunk” is recognized, bringing with it more specific actions to be tried in sequence (checked for typicality and applicability).
For example, in the Greek Gift sacrifice, an obvious salient clue is the vulnerability of the h7 (h2) square to being captured with check by a Bishop. That is the precursor salient clue. Another salient clue may be no defensive knight (or a knight that can be driven away) on f3 (f6). It does not matter what order these clues are recognized, but it it is more likely that the Bishop sacrifice will definitely trigger “Greek Gift sacrifice” before anything else.
We don’t have to “SEE” everything at once in a large”chunk” at the beginning. We just have to keep recognizing and simulating the action steps until we reach a recognizable goal. If there are missing cues or lack of typicality at any stage, we reject that specific approach (one-mover or two-mover) if it cannot be modified successfully, back up to an alternative and drill in again. The mental simulation MAY reveal an alternative overall goal. If so, we switch to recognizing alternate patterns (more one-mover and two-mover patterns) that focus on that new goal. For example, if we start with an assumed goal of checkmate, at some point in the mental simulation we may realize that the opponent can “bail out” by giving up a certain amount of material to escape checkmate. In that case, perhaps we simulate a different alternative with a gain of material goal.
I expect we will end up with trial and error again. But now driven by System 1 and not System 2
DeleteAssumption is the mother of all f-ups.
DeleteWe will definitely end up in “trial and error” mode, assuming that “trial and error” means (Merriam-Webster Online):
Deletea finding out of the best way to reach a desired result or a correct solution by trying out one or more ways or means and by noting and eliminating errors or causes of failure [to reach a desired result or a correct solution]
also : the trying of one thing or another until something succeeds
What’s missing from this definition is a sense of randomness caused by a lack of goal directedness. Note the emphasized (by me) phrase: “the best way to reach a desired result or a correct solution”.
If there is a goal, then there must be some sense of directedness toward that goal. Investigating various options (informed by experience which has resulted in SKILL) does NOT require a “throw it up against the wall and see what happens” random process.
Here is an example.
FEN: 8/8/8/1PkN4/1p6/8/8/1K6 w - - 0 1
What is White’s reasonable expectation for a result in this position? He has a “winning” material advantage (the knight — +3) but IFF he can safely promote the WPb5. White has a reasonable expectation that he can win or draw, at a minimum, given that if he loses both his knight and pawn, he still has a good position in front of Black’s pawn.
GM Stockfish gives all 14 possible moves for White: 8 moves by the WNd5, 5 moves by the WKb1 and 1 move by the WPb5. All but one of these moves results in a draw; the correct move is a win.
The overall goal is to promote that pawn. In order to reach that long-term goal, an intermediate goal is to protect that pawn. In fact, the IMMEDIATE goal is to protect that pawn. We begin by eliminating all moves from consideration that do not achieve all three goals. All of the king moves are eliminated. All knight moves that do not protect the WPb5 are eliminated, leaving 2 knight moves: 1. Nc7 (most obvious, but it can be driven away from the defense of WPb5 by 1… Kb6) and 1. Nc3 (obvious, but the knight is eliminated immediately by capture, followed by the elimination of the WPb5).
The only other move is 1. b6. That seems to allow Black to capture the knight and then take care of the pawn. HOWEVER, there is an additional wrinkle: the square of the pawn. If Black captures the knight, then the WPb5 is free to advance to promotion undeterred. The Function of stopping the passed pawn prevents Black from capturing the knight. The knight protects the passed pawn from behind it.
All of that is within the two-mover event horizon.
What is the difference between trial and error guided by System 2 and trial and error guided by System 1?
DeleteSystem 2 needs a word salat for every decision, while System 1 prunes by logic from the first step. It starts with the pattern how a passer is safe when guided from behind by a knight.
Then you have your hands free to approach with you king and do whatever is needed.
There is no randomness in the process, because the absorbed patterns retrieve logic from memory when triggered.
I used to have a lot of discussions in the past with people advocating the iterative process from the Scrum method.
ReplyDeleteSome things you must plan from a holistic view beforehand. When you build a home, you cannot drive a pile iteratively. You can't say after you build the bathroom, actually this pile should be 20 cm further to the left.
The same is true for big chunks, I guess. You only can plan the structure of the whole combination beforehand when you see the big picture. If you just follow the short-movers you might well end up in a tunnel or a rabbit hole.
This whole blog wouldn't exist if I hadn't a vision of the Trick beforehand.
ReplyDeleteFEN: 3n4/3k1p1p/B2P2p1/8/2P2P2/P5PP/2P1r3/1R4K1 b - - 0 1
ReplyDeleteGiven that Black is down in material by 4 pawns, capturing on d6 seems like a good idea—or is it? What pattern is latent in this position?
1.… Kxd6
The pattern becomes actual. Black has created two weaknesses: his king can be attacked (check) while simultaneously the LoA WBa6→BRe2 is opened. Because a check is the highest level of forcing move, Black must move out of check, leaving the BRe2 to be captured.
2. c5+
What is the point of such a simple example? First, the two players are fairly evenly matched in rating; both are at least IM or GM. Second, the slightly higher rated player made a serious mistake in a simple one-mover position. It does not require the calculation skill of a Kasparov to “SEE” what would happen. Third, it is not a matter of randomly looking for “forcing” moves using CMCT (Check, Mate threat, Capture, other Threat). It is a matter of SEEing the interrelationships between the various pieces (the lines of attack) AND taking advantage of those interrelationships when they occur, even if it is accidental (not the result of brilliant play).
NOTE: White goes slightly “wrong” around move 15, but then Black proceeds to make several slight mistakes. The final mistake extinguishes all hope.
[White "Gennadi Zaichik"]
[Black "Lothar Vogt"]
[WhiteElo "2485"]
[BlackElo "2510"]
[Date "1988.??.??"]
[Round "9"]
[Result "1-0"]
[Event "It"]
[Site "Berlin GER"]
[Eco "A00"]
1.g3 g6 2.Bg2 Bg7 3.d4 Nf6 4.e4 d6 5.Ne2 O-O 6.h3 e5 7.Nbc3 Nc6 8.Be3 Bd7 9.O-O exd4 10.Nxd4 Re8 11.Re1 a6 12.Kh2 Ne5 13.b3 c5 14.Nde2 Bc6 15.Qd2 Qa5 16.Rad1 Rad8 17.Bg5 Qb6 18.Nf4 Rd7 19.Bxf6 Bxf6 20.Nfd5 Bxd5 21.Nxd5 Qd8 22.Nxf6+ Qxf6 23.f4 Nc6 24.Re3 Rdd8 25.Rd3 Nb4 26.e5 Qf5 27.Rxd6 Rxd6 28.exd6 Qd7 29.a3 Nc6 30.Qd5 Re2 31.Qxc5 Qe6 32.Qc4 Qxc4 33.bxc4 Kf8 34.Kg1 Ke8 35.Rb1 Nd8 36.Bxb7 Kd7 37.Bxa6 Kxd6 38.c5+ 1-0
FEN: r1bqk2r/pp1p1ppp/2n1pn2/8/1b1NP3/2N2B2/PPP2PPP/R1BQK2R b KQkq - 3 7
ReplyDeleteThe interesting thing is how simple one-movers and two-movers can determine the game outcome.
Source: chessgames.com
The same position, 42 years apart, World Champion E. Lasker in the first game, future World Champion B. Spassky in the second game. The result? Both LOST the game. It’s hard to believe that Spassky, even as an 11 year old, would lose so quickly in the opening. Lasker, on the other hand, managed to keep the game going for 47 moves before screwing up in the endgame (hard to believe, given his prowess in the endgame). The circumstances were different: Lasker was giving a simul; Lutton was a fairly strong player at that time. Spassky was playing a tournament game (I think); I’m pretty sure that he was unfamiliar with Lasker’s game.
[Event "Simul, 26b"]
[Site "Pittsburgh, PA USA"]
[Date "1906.10.17"]
[EventDate "?"]
[Round "?"]
[Result "0-1"]
[White "Emanuel Lasker"]
[Black "Benjamin Howard Lutton"]
[ECO "B45"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "94"]
1.e4 c5 2.Nf3 Nc6 3.Nc3 e6 4.d4 cxd4 5.Nxd4 Nf6 6.Be2 Bb4
7.Bf3 d5 8.exd5 Nxd5 9.Nxc6 bxc6 10.O-O Bxc3 11.bxc3 Ba6
12.Re1 O-O 13.Qd4 Qf6 14.Bd2 Qxd4 15.cxd4 Rab8 16.Rab1 Bc4
17.a3 g6 18.g3 Rb6 19.Rxb6 axb6 20.Be2 Bxe2 21.Rxe2 b5 22.Bc1
Ra8 23.Kf1 b4 24.axb4 Ra1 25.Re1 Nxb4 26.Bb2 Rxe1+ 27.Kxe1
Nxc2+ 28.Kd2 Nb4 29.Kc3 Nd5+ 30.Kc4 Nf6 31.f3 Nd7 32.Ba3 f6
33.Bd6 Kf7 34.f4 h5 35.Bc7 g5 36.fxg5 fxg5 37.Kd3 Nf6 38.Bd8
Kg6 39.Bxf6 Kxf6 40.Ke4 g4 41.Kf4 e5+ 42.dxe5+ Ke6 43.Ke4 c5
44.Kf4 c4 45.Ke4 c3 46.Kd3 Kxe5 47.Kxc3 Ke4 0-1
[Event "URS"]
[Site "URS"]
[Date "1948.??.??"]
[EventDate "?"]
[Round "?"]
[Result "0-1"]
[White "Boris Spassky"]
[Black "Rodgaisky (or Podgaisky)"]
[ECO "B45"]
[WhiteElo "?"]
[BlackElo "?"]
[PlyCount "16"]
1. e4 c5 2. Nf3 Nc6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 e6 6. Be2
Bb4 7. Bf3 Qa5 8. Ndb5 Qxb5 0-1
Mr. Stockfish is not convinced.
DeleteThe 12 games on Chess Tempo were split evenly: 4 White wins, 4 draws, and 4 Black wins. GM Stockfish "suggests" 5 moves that were better for Black than either 7... Qa5 or 7... d5, but still within the "plus-equals" evaluation. The difference between the two games is that Spassky blundered a one-mover immediately following 7... Qa5 with one-mover 8. Ndb5. Since the WNc3 was absolutely pinned, its defensive function on b5 was non-existent: simply 8… Qxb5 wins a knight for free.
DeleteLasker appeared to be following the usual simul strategy of batting the ball back across the net, waiting for the usual gross error by the amateur; things didn't work out that way in the end. He kept a slight advantage up until 20. Be2, allowing Black to exchange the light-square bishop and thereby consolidate his knight on the d5 (white square) outpost against the Black dark-square bishop (a two-mover sequence). 22. Bc1 gave Black more of an advantage (+1.38) because the bishop is really doing nothing there other than protecting the WPa3 AND it created a potential back-rank defensive problem AND it made defense of the WPc3 awkward. Black immediately activated his rook against the a3 pawn, with the idea of getting his rook to the back rank or to c4. Again, these were one-mover and two-mover ideas. I suspect that Lasker underestimated or overlooked the force of the one-mover 23… b4. The next few moves were forcing, so not very complicated for either player to find; Black ended up a pawn ahead. All of Black’s pieces ended up on white squares, so the bishop had no targets. White got back to a more or less even position after activating his bishop but then blew it by capturing the knight with the bishop, giving Black a winning king-and-pawn endgame. Lutton apparently was sufficiently skilled in endgame play to bring home the full point. The outside passed pawn wins, creating a “Fox in the Chicken Coop” situation on the kingside!
I'm not ready to think on such large scale. I'm thinking of the previous post. The first move 1. ... Qb1+ is logical, even without a grand plan behind it. Black escapes a skewer with tempo. I wonder whether that is a regularity? Is a grand plan always identifiable by a one- or two-mover? CCT seems to point in that direction.
DeleteAfter examining a few dozen examples it might be a rule indeed. Mastering the two-movers leads you to the main combination. Maybe there are situations where it will not work. But for now, we shouldn't bother, I guess.
DeleteChess Life, the USCF official magazine, has started a new monthly series titled Tactical TESTS. There are 12 puzzles, all taken from recent games. The puzzles generally increase in difficulty, with a range of tasks including mates, winning material, and defending. The source games are identified, but I haven’t been able to find the game scores.
ReplyDeleteHere are two examples from the June 2026 issue, pg 9:
TACTIC 9.
FEN: r1bR1NK1/1p3pp1/2p2n1p/p7/Pq2PP2/2N4P/1PP3BK/1R6 b - - 0 1
TACTIC 11.
FEN: 4r1k1/QP1nqpb1/3p1np1/1BpP3p/7P/2N2P1R/1P1B1KP1/8 b - - 0 1
The combinations are longer than one move, but if you can identify the one-mover to start the solution, it makes it much easier to follow the logical pattern. The one-mover highlights the major weakness in the enemy camp.
The FEN of tactic 9 doesn't seem to be correct
DeleteCorrected FEN: r1bR1Nk1/1p3pp1/2p2n1p/p7/Pq2PP2/2N4P/1PP3BK/1R6 b - - 0 1
DeleteI copied it into Chess Tempo and corrected it but failed to copy it back into the comment text. My apology.
Tactic 9. My first idea was to look at the discovered attack with bishop and rook. I couldn't make it work. Then I focused on the poorly placed defender of Nf8.
ReplyDeleteAnd I remembered more poorly placed defenders form my problem set.
I assume that is the way to go in the study room. By gathering a set of problems that are logically connected.
I think you are correct: the way to go is to collect a set of problems that are logically connected. That leads to a question (perhaps at a higher level of abstraction): What logically connects the problems?
ReplyDeleteIMHO, the logical connection is NOT the tactical devices/themes (fork, pin, skewer, etc.) that are present to carry out the combination. Unfortunately, that is the usual demarcation criteria for problem sets, simply because that’s what has always been used as the supposed “trigger” for pattern recognition. It’s also easy to “SEE” those tactical devices/themes as salient features lying on the surface.
Instead, I contend that the motive(s) are the triggering pattern recognition. The overarching categorization of motive is the WHY a combination exists—and the reason in every case is a weakness of some kind, perhaps leading to a secondary weakness, eventually leading to a double attack (GM Averbakh’s general categorization) against two (or more) weaknesses simultaneously which cannot both be adequately defended. As Lasker rightly noted, the “lightning” of an attack flows along the path of least resistance, IE, toward the weakness(es). In essence, this is your definition of a “duplo” move: two goals achieved (a double attack against weaknesses) with one move.
In Tactic 9, the weakness in White’s position is clearly in the juxtaposition of the WRd8 (undefended, with limited mobility to maintain its defensive Function) and WNf8 (capturable [1:2], with no safe discovered check which can be used to extract it from being captured). At the very least, Black can simply capture 1.… Qxf8 2. Rxf8+ Kxf8, giving a more-or-less equal material situation. In the game, that’s what Black did, but he ended up losing the game. 1.… Qe7 attacks the undefended WRd8 without changing the attacking ratio on f8, a duplo attack. White can balance that attack with a defense by 2. WRbd1. At this point, Black can interpose 2.… Ne8, interfering with the line of defense of WNf8.
If we used the traditional categorization of patterns by identifying the tactical devices/themes involved, we would have a fork, followed by line closing BUT that classification would provide no generalization of pattern recognition to other similar positions—similar in weaknesses but not necessarily similar in tactical devices/themes used. The attention would be focused incorrectly on the mechanisms used to take advantage of the weaknesses, not on FIRST recognizing the weaknesses.
In Tactic 11, the primary weakness is the B.A.D. square e1 [2:2], defended by the WBd2 and WKf2. There is a pawn-king pattern that involves a mate if Black could get his queen to e1. Since Black has more than one attacker aimed at the B.A.D. square, it is actually under-defended because the king cannot capture on that square except as the last defender in a capture sequence. What could be more obvious than to recognize the encircling motif (immobility combined with superior force) as the potential solution? The tactical response to that pattern is to add attackers or reduce defenders (by capture or diversion). That leads to the (NOT “obvious”) one-mover 1.… Bh6. If White captures 2. Bxh6, 2.… Qe1#. The rest of the solution is how to get rid of the WBd2 defender of the black squares so that the Black queen and rook can attack the White king. The obvious solution is to capture WBd2 or force it to capture, allowing it to be captured. White’s queen and rook are both awkwardly placed for defense of the White king.
I’ve been wrestling with the issue of identifying pre-conditions for combinations. There seems to be no consistent way to identify potential combinations based on recognizing tactical devices/themes. The myriad ways that specific pieces can be “combined” into tactical sequences seems practically infinite. That does not lend itself to systematization of patterns to be recognized because so much is dependent on specific pieces-on-squares.
ReplyDeleteHowever, categorization of weaknesses is a much smaller set of possibilities. The weaknesses may be geometrical, but there is no specific geometrical pieces-on-squares pattern that must be recognized by System 1 in order to start a combination. In short, the weakness(es) signal the potential presence of combinations, no matter how deeply hidden that combination might be.
Here’s another Tactical TEST problem.
Tactic 7.
FEN: r5k1/N4p2/1p2p1pp/3p1n2/P2P1P1q/1QP5/1P3P1P/R3R1K1 b - - 0 1
White’s kingside has been weakened (open g-file) and there are two Black pieces in the near vicinity. The WQb3 is LPDO and cannot return easily to the defense of the king. White does have material superiority but Black can restore the material balance by capturing 1.… Rxa7 but that allows White to capture another pawn, with an outside passed pawn. The awkward position of the BNf5 precludes the typical queen+knight mating attack on h2 or f2. However, if the knight is on h4 (instead of the queen), then Black can threaten mate on g2.
The question is: How can Black take advantage of the weaknesses?
The obvious thing to try is to maneuver into a checkmate position starting with 1.… Qg5+. That reduces White’s possible moves to two: 2. Kh1 and 2. Kf1. Generally, heading into a corner with no possible moves is not likely to be a good idea. 2. Kh1 Nh4 threatens mate on g2. The only possible defense is 3. Rg1, which fails to 3.… Qf6+ 4. Rg2 Qxg2#. That is straightforward.
So how does a second weakness enter the picture?
Switching to the alternative, 2. Kf1 tries to skedaddle from the kill zone. 2.… Qh3+ forces White to decide whether to retreat back to g1 (3. Kg1) which is met by the same mating pattern as before) or to try to escape toward the center/queenside (3. Ke2). BUT, the BQh3 now has a relationship with the LPDO WQb3! After 3. Ke2 Black takes advantage of the “pin” on WPc3 by using a knight fork 3.… Nxd4+, and regardless of White’s response, the White queen is coming off the board.
Currently, this is my take: the logical patterns cover a whole host of positions. They are not infinite in numbers at all.
DeleteThe related one-movers are finite too, but they can combine in a myriad ways. You must learn to solve the on-movers a tempo whenever they occur. (mate on g2, pin against Qb3).
I assume you are struggling with the logical patterns. I'm gathering a few examples for my next post to give you a better idea what to look for. But it takes a bit time to collect the patterns.
You wrote:
ReplyDeleteI assume you are struggling with the logical patterns.
That's a correct assumption.
I’ve already gained some more insight into the idea of using logical patterns. I’m eagerly awaiting your examples.
In the interim:
In reviewing Klein’s Recognition-Primed Decision model, I’ve come to understand it better. As previously noted, the RPD model is a DESCRIPTION of what experienced decision makers do, not a logical step-by-step PRESCRIPTION (IE, a logical process) of how to go about reaching a good (but not necessarily optimum) decision.
One of the key aspects of this model is that it describes what EXPERIENCED decision makers do in general. “Experienced” is synonymous with SKILLFUL expertise in the specific domain, and encompasses a wide range of possible skills. Novices do not have the experience (IE, skills) required to make good decisions using intuition (pattern recognition). Consequently, they must resort to “trail-and-error” or to a rational choice strategy based on comparative evaluation of two or more alternatives, applying a “King of the Hill” approach. This logical approach is embodied in various suggested ways to approach finding the best move, such as Kotov and Heisman.
Note that THERE IS NOTHING WRONG WITH SUGGESTING THIS LOGICAL STEP-BY-STEP PROCESS FOR NOVICES! An interesting aspect of this approach is that if more experienced players find themselves in a situation in which they have no skill (no System 1 recognition of what to do or how to do it), then the fallback approach is to this type of process.
Kahneman described the mental simulation of Klein’s model as being run by System 2. I don’t think that’s the case. Let’s review the RPD model.
The decision maker experiences a situation. Whether this is a puzzle or a position while playing a game is irrelevant.
There are four by-products of recognizing the situation as typical (IE, known to the decision maker) by prototype or by analogy.
First, there are reasonable expectations associated with the recognition.
Second, certain types of goals are plausible, and set the priorities.
Third, certain recognized surface-level cues trigger retrieval of salient features.
Fourth, one or more “actions” [sequences of moves] are retrieved in association with the expectations, goals, and cues.
I had assumed that each “action” was a single move or variation but that is NOT the case. An action is always a sequence of one or more discrete components (such as individual moves, tactical devices/themes, etc.). Think of an action as a single variation. Klein surmises that action sequences are often no more than 6 steps (moves?) long, and usually there are 3 or fewer relevant action sequences (variations) that make sense in the recognized situation pattern. Those numbers are not hard and fast limits.
I went astray assuming that pattern recognition was confined to the salient cues. I now think pattern recognition refers to the expectations, goals, cues and actions which have all become associated into a “chunk” or template.
So what is mental simulation? It is imagining and playing out (using System 2) one action sequence at a time, step-by-step, checking to see if typicality is maintained to the end of the sequence (when the goal or a recognizable stopping point has been reached—material gain or checkmate). If not, then any anomaly or unexpected occurrence is clarified and taken into account by subsequent steps in the action sequence. If a given action sequence does not produce a definitive answer, it is abandoned and another alternate action sequence is mentally simulated. Lather, rinse, repeat until a satisfactory action sequence is found.