In search for an algorithm
I'm in search for an algorithm that might be able to help me to decide between different middlegame moves. That proves to be not so easy.
The enthusiast stories of IM Toth point in the right direction. It shows the GM games of interest that should be studied. It gives straightforward lines that tell a consistent story.
But when diving deeper into the matter, devils start to popup from the details.
The problem is that the only tool that I have at my disposal, is GM Stockfish. And Stockfish is notoriously bad at explaining moves.
The moves in the stories of IM Toth are not always computer checked. And for good reason. You are not allowed to use a computer during an OTB game, and neither is your opponent. Hence you must play in a way that is humanly understandable, based on simple plans with positive statistics. Meaning that it is easy to play for you and difficult to play for your opponent.
But when questions arise about the moves of IM Toth, you are at a loss. Since Stockfish can't help you there. Meaning, that at the end of the day, you have to think for yourself again.
 |
| White to move. Blacks last move was 13. ... Bc5 |
r1bqr1k1/1p3ppp/2pp1n2/p1b1n3/2PNP3/2NB3P/PPQB1PP1/4RRK1 w - - 1 14
Spraggett, Kevin vs. Smyslov, Vassily V
The story is as follows:
Not all blacks pieces are developed. Yet the pieces that ARE developed, exert great power.
Whites pieces are SEEMINGLY more developed, but on the squares where they currently stand on, they exert little or no power.
White makes a tactical mistake here by playing 14. Be3
Positionally the bishop is better placed on e3, but tactically it is a blunder.
The story in itself is very good. But the discovery that white has a much better move here came as a surprise.
Meaning
The good news is, that my ambitious plan is flawed. I cannot trust the positional patterns. Hence I can drop the positional exercises and speed up both the tactical exercises and the endgame exercises by a factor 2.
The bad news that it is more difficult to concoct an algorithm. TINSTAAFL.
PART I:
ReplyDeleteIs there a single “algorithm that might be able to help me to decide between different middlegame moves”? Perhaps, but I think such an algorithm might be counter-productive in some critical ways. System 2 is notoriously unwilling to devote scarce resources and attention to anything that has become “routine”—and there is nothing more “routine” than a canned set of steps which can be followed without any thought. It is also highly unlikely that any algorithm will be sufficiently “simple” so as to be usable in any and all situations, yet not divert attention away from the concrete details of a specific position. In short, going through a series of predigested “steps” is very likely to work “most of the time”—and then fail spectacularly at critical moments requiring a response to non-typical scenarios. Given the complexity of chess and the virtually unlimited number of practical alternatives in any given position, an algorithm seems unlikely to be the nec plus ultra method for UTILIZING (note: N-O-T developing) tactical skills.
At the highest level, Aox’s suggestion of starting with a simple question gets the analytical process working by narrowing the focus to tactical possibilities:
Is the general objective in this specific position about checkmate, gaining material or pawn promotion?
Getting “real” in the given position from the Spraggett, Kevin vs. Smyslov, Vassily V game, starting with that simple question allows two of the three possibilities to be eliminated without further (deeper) analysis; that “answer” is lying on the surface. There is no possibility of checkmate or pawn promotion.
It is important to avoid the idea that general development (all the pieces off the back rank except for the rooks) is the sine qua non of the opening. A localized advantage in development may be much more important in any given concrete position.
PART II:
ReplyDeletePoPLoAFun comes into play.
Are there any undefended pieces (LPDO)?
Yes, the WNd4 is unprotected AND attacked after Black’s last move (13… Bc5). There do not appear to be any “desperado” moves that this knight can make, so that reduces the options to either move it or defend it. Spraggett opted to defend it by 14. Be3, without really considering what would be changed in the position after that move. I am certain that a GM of his caliber was perfectly capable of deciphering the consequences—but for some unknown reason, he did not do so. I doubt that he was in serious time trouble after 13 moves. So why did he make an automatic (algorithmic) response apparently based on “When attacked, defend the attacked piece”? Given the stage of the game (late opening/transition to early middlegame), perhaps his sense of danger was not activated in time.
Since there IS a LPDO piece, the next step would be to consider ALL alternatives, not just playing on “autopilot” by defending a well-placed knight in the center AND simultaneously developing the last minor piece in support of the center.
Are there B.A.D. squares?
Yes, the f3-square is B.A.D. [1:1]. An additional “warning signal” is that a knight placed on it would be a simple fork of the WKg1 (with check) and the WRe1, with potential gain of the exchange. This consideration means that the WPg2 is overloaded [Function]; the previous advance of the WP to the h3-square provided a target for the undeveloped BBc8. This is a simple chain of logic related to uncovering ALL of the PoPLoAFun.
Algorithmically, if only one weakness is detected, do not attempt to exploit it immediately. Instead, look for a way to create a second weakness that can be connected to the first weakness. WHATEVER YOU DO, DO N-O-T VOLUNTARILY CREATE THE SECOND WEAKNESS! I am certain the GM Smyslov had NO IDEA that Spraggett would create the second weakness. The combination of two weaknesses is sufficient to trigger a focused search for the means to take advantage of them. In this case, the LPDO WNd4, combined with the weak f-square was sufficient for GM Smyslov to immediately take advantage of it. This was no deep strategical plan or fantastic tactical combination; it was merely taking concrete advantage of what is immediately in front of your eyes using “simple” tactics!
Spraggett knew he was busted as soon as Smyslov played 14… Bxh3— and he resigned. White can refrain from taking the BBh3 (recommended by GM Stockfish), but is then just down a pawn without compensation because Black can simply move the BBh3 to safety. After 15. gxh3 Bxd4 16. Bxd4 Nf3+ 17. Kh2 (or elsewhere) Nxe1, Black is a pawn and the exchange ahead in material (which was the focus of our starting assessment based on Aox’s suggested first step).
Rather than an algorithm, I think it is much more useful to keep tight rein on how we think about a position. At each turn (whether our own move or the opponent’s move), we MUST be aware of what has changed move-to-move by the last move. It is always being aware that “To err is human, especially in chess!”
One of the key things that I learned early on was to apply focused attention to every single move; there was no algorithm involved at all. Perhaps I am mistaken, but I attribute that attitude (more than any knowledge or “thinking process”) to climbing the rating scale as far as I did.
1. stage from the opening to the conquering of the center
ReplyDelete2. positional algorithm
3. only used in the study room
4. what is the effect on the center (d5, e5 for white)
5. is the piece standing safe (no pin, no fork, no tempo
6. does the piece something on its new square
7. does it pin, fork or threat (wins a tempo or potentially a future tempo)
An attack starts with an invasion of some sort. I look for an algorithm that describes the best development move.
a. which piece to move
b. with the most effect
c. preventing loss of a tempo
d. gaining a tempo
In the end, the algorithm will become a logical narrative. And from the narrative, the patterns will become absorbed.
For now I neglect the later stages.
Stage 2: creation of an invasion square
Stage 3: attack
I focus only on the first stage: conquering the center. I want every developing move to be optimized.
It is a matter of chance. In the position above, white has MORE pieces developed, but they do less.
ReplyDeleteBlack has LESS pieces developed, but they do MORE.
Hence development is measured in what pieces DO. No matter where they stand. By measuring the activity of the pieces TIMES the amount of pieces that do something I get the grade of development. Better development results in better CHANCES.
In the above position, that translated into a tactical mistake. The chances that white makes such mistake is greater than that black makes a mistake.
How do I measure development in the first stage of the game (conquering the center)?
At a certain moment, blacks pieces were clogged up. Whites pawn on d4 hindered a black knight on d7 which in turn prevented Bc8 from developing which in turn hindered the development of Ra8.
ReplyDeleteSo black took measures to trade his pawn on e6 for whites d4 pawn.
The algorithm I am looking for is about tracing the clogged pieces on both sides.
The procedure I use is to:-
ReplyDelete1) Identify the candidate moves;
2) Count the number of things that a move does: defend this, attack that, develop a piece that is doing nothing, ...
3) Select the move that does the most
This is mainly for postions where there are no immediate tactics to calculate. I've used it many times.
That's certainly one route to go.
Delete