Back in 2017 I concocted the tree of scenarios. It was, as usual, a big effort of system 2. I never have been able to make it practical. The past days I stumbled again on the tree of scenarios by accident, but now spiralling in from a totally different angle.
The framework I was looking for the past days already has been concocted, waiting to be used. In 2017, I had no idea how to do that. But currently I have quite a lot of fresh ideas. It makes a big difference whether you are still in the process of inventing or that a whole tree lies there, ready to be used.
It is not rocket science:
Identification of the target.
- B.A.D. (Barely Adequate Defended)
- Not/insufficient defended piece
- Invasion square
Identification of the action
- eliminate the defender
- add attacker
Identification of the tempi
- double attack
- exchange
- additional punch
- step out of a threat
Identification of the FUNction
- defender
- overloaded
- pinned
Other
Neutralize counter attack
Trap
Promotion
And that is about it!
FEN: r2q1rk1/pp3nbp/3pb1p1/4np2/2PN1B2/2N4P/PP1Q1PP1/R3KB1R b KQ - 1 13
ReplyDeleteIn this position, Black played 13… Nxc4. What should White do?
Black had superiority [1:2] on the c4-square, so capturing on the c4-square is ‘safe’ [Heisman] AND the White Queen is under attack. Unfortunately for Black, BBe6 is hanging [LPDO] and White has the right to move first in a capture sequence. Even though Black actually captured first, White’s 14. Nxe6 captures more material value. It is as if White has captured first. Back has two potential captures [WQd2, WBf1] whereas White has three potential captures [BQd8, BNf7, BNf1].
Black loses a piece for a pawn.
It’s curious that a strong player player was unable to “SEE” the capture sequence and realize that he would lose material if he initiated it. I suspect severe time trouble is the likely cause.
lichess.org Puzzle #xXoD9
From game 3+0 • Blitz
Not-Not-Not-Yes-Not (2360)
NM FalzeHope (2372)
1. d4 g6
2. c4 Bg7
3. Nc3 Nh6
4. Nf3 O-O
5. e4 d6
6. h3 f5
7. e5 Nc6
8. Bf4 e6
9. Qd2 Nf7
10. exd6 cxd6
11. d5 Nce5
12. dxe6 Bxe6
13. Nd4 Nxc4
14. Nxe6 Nxd2
15. Nxd8
Black resigned
I added one extra target to look for in the post: invasion square.
ReplyDeleteFurthermore I tagged all problems in my database that could benefit from an optimized tree of scenarios. About 65% of the problems is suited.
That is the work that is laid out for me: optimizing the tree of scenarios with my new insights.
It is NOT sufficient to simply state that weaknesses must be attacked. While true [KNOW THAT], it provides zero information on how to go about doing it [KNOW HOW].
ReplyDeleteYour current investigation is focused on [KNOW HOW]. BRAVO!
Identify the cues for types of weaknesses and develop the techniques to take advantage of each type of weakness. The "attack" always follows the "line of least resistance." Line in this case could mean a train of thought, a Line of Attack (LoA) or a variation of moves. It's applicable to all three cases.
There is a similar phrase that summarizes all martial arts throwing techniques:
"Open a hole, fill the hole - with the help of Mr. Gravity!"
That is mere KNOW THAT, unless you are trained in KNOW HOW (how to apply it in many different situations using many different techniques).
The problem always has been that there is way too much variation in the problems. Now I have selected 26 problems that I deem as problems that can be solved by the tree of scenarios. This means that I did not chose the traps, promotions and counter attacks. May be I can fit those in later, but right now, I want to keep matters as simple as possible.
ReplyDeleteAnd I have good news. I made a round along all 26 problems, to find out what cues could start the logic. And much to my surprise I found that in all 26 cases I could simply start by looking at the targets. Both the targets in the opponents camp and the targets in my own camp. I had to stretch the definition of target a bit, by adding the invasion squares. But that is a minor quibble.
As I noted previously, Blogger is not updating the My Blog List links on your blog.
ReplyDeleteAs I reviewed some of the recent The Chess Improver articles, I found this excerpt from the 23 JAN 2024 article Chess and AI by Valer Eugen Demian.
[EXCERPT]
When I asked Perplexity.ai chatbot:
“I want to solve chess puzzles. How do I do that?”
Here is the answer I got:
“To solve chess puzzles, you can follow these general steps:
1. Quick Evaluation: Assess the position.
2. Determine Objective: Identify the likely goal, such as checkmate or winning material.
3. Consider Opponent’s Last Move: Understand your opponent’s recent actions.
4. Identify Possible Moves: Look for particular targets and motifs.
5. Examine Attacking Moves: Analyze aggressive options.
6. Select Your Move: Make your choice.
7. Compare with Solution: Check your answer against the correct solution.
8. Study the Solution: Understand the correct approach and learn from it.”
“Additionally, when solving puzzles, you can use the following three-step guide:
1. Look for all possible checks in the position.
2. Then, consider all possible captures.
3. Finally, search for undefended or vulnerable pieces.”
“It’s also beneficial to practice regularly to improve your tactical skills. Whether you use a real board or a digital platform like chess.com, the key is consistent practice and thoughtful analysis.”
[Emphasis added]
[END EXCERPT]
There seems to be lots of similarity to what you have discovered over the last 23 years. I wonder if your blog is one of the sources that the AI gleaned for its recommendations.
The first 6 steps provide a general ‘thinking process’ framework - for puzzles, at least. There is a lot of similarity to the approach that Aox recommended quite a long time ago.
The three-step guide seems to be the generic CCT [Checks, Captures, Threats] advice.
It will be interesting to “SEE” the evolving integration of your insights regarding tactics and the tree of scenarios. I expect that it will be one of the most practical and useful frameworks for training overall chess SKILL, not limited to just tactics.
First action: strip the list from everything that can be missed. Then only point 4 remains, look for particular targets and motifs. That's where I am now. Then rebuild everything from there.
ReplyDeleteFEN: 4rqk1/ppp4p/4b1p1/4R1Q1/5rP1/8/PPP4P/4R1K1 b - - 2 23
ReplyDeleteWhile awaiting elucidation of the tree of scenarios in conjunction with LOOKING for particular targets and motifs, here’s an application of that ‘thinking process” suggested by the AI.
Lichess.org Puzzle #VaxxU
General steps:
1. Quick Evaluation: Assess the position.
Black is ahead in material by a Bishop. Black has the right to move first. Ipso facto, Black has the advantage.
2. Determine Objective: Identify the likely goal, such as checkmate or winning material.
An immediate checkmate does not seem to be in the cards, so winning more material seems to be the obvious objective.
General principles [courtesy of chess.com]:
(a) Exchange pieces when it simplifies the game: If the position is complex and you're having trouble finding a good move, consider exchanging pieces to simplify the game.
(b) Exchange pieces when it weakens your opponent's position: If you can exchange a strong piece for a weaker one and weaken your opponent's position in the process, go for it.
3. Consider Opponent’s Last Move: Understand your opponent’s recent actions.
White’s last move was 23. Rde5, pinning the BBe6.
4.Identify Possible Moves: Look for particular targets and motifs.
The g4-square is a particularly ‘juicy’ square. It captures a pawn AND forks the WQg5 and WKg1 WITH CHECK, with forced win of the White Queen for the Black Rook! SERENDIPITY!
5. Examine Attacking Moves: Analyze aggressive options.
Nothing else seems to make sense: the “general principles” and the tactics all say “GO FOR IT!”
6. Select Your Move: Make your choice.
23… Rxg4+ is da BOMB!
7. Compare with Solution: Check your answer against the correct solution.
Yep, it WAS “da BOMB” – and it blew up in Black’s face. According to GM Stockfish, 23… Rxg4+ is the 27th worst move of all possible “candidate moves”!!
8. Study the Solution: Understand the correct approach and learn from it.
WHAT IN HELL WENT SO WRONG?!?
First of all, jettison the idea of choosing a move on “general principles.” This is a tactical position, which can be verified by looking at things like B.A.D. squares, under-protected pieces, Lines of Attack [looking all the way to the edge of the board], in short, all of those issues that have been under discussion for quite some time on this blog.
As Temposchlucker noted, let’s focus on 4. Look for particular targets and motifs.
White’s last move attacks the BBe6 [2:1] A-N-D (relatively) “pins” it to the WRe8 via the battery [WRe1+WRe5]. If the BBe6 moves, White has the threat of capturing on the e8-square, (absolutely) “pinning” the BQf8 while capturing a Rook. On a good day, that means White will end up a Rook ahead. 23… Rxg4+ virtually forces 24. Qxg4 Bxg4and Black is w-a-y ahead in material – momentarily. Now, it’s White who has the right to move first. 25. Rxe8 Qxe8 26. Rxe8+ and White ends up with a Rook versus Bishop and Pawn.
All of this is foreSEEable – but Black got carried away with following general principles and did not look at ALL the tactical things that White’s last move set up.
There might be a “lesson” in there somewhere:
You can’t calculate what you can’t SEE.
From game 5+0 • Blitz
Diggest2020 (1736)
Lincoln_Osiris (1747)
1. e4 e5
2. d4 Nc6
3. Nf3 exd4
4. Nxd4 Be7
5. Be2 Nf6
6. Nc3 O-O
7. O-O Bb4
8. Nxc6 dxc6
9. Bd2 Bxc3
10. Bxc3 Nxe4
11. Be5 Be6
12. Bd3 Nf6
13. Qf3 Nd5
14. Qg3 g6
15. Rad1 Qc8
16. Qg5 f6
17. Bxf6 Rxf6
18. Bc4 Qf8
19. Rfe1 Re8
20. g4 Rxf2
21. Bxd5 cxd5
22. Rxd5 Rf4
23. Rde5 [-5.9] Rxg4+ [-3.0]
24. Qxg4 Bxg4
25. Rxe8 Bd7
26. Rxf8+
Black resigned
WHAT IN HELL WENT SO WRONG?!? That is simple. Look for particular targets and motifs.. Look for the targets of your enemy too!!
ReplyDeleteThis AI-proposed "formal" thinking process is logical (step-by-step) and sufficiently generalized to work successfully in all possible scenarios. BUT, as you point out, misinterpretation of even one of its steps can lead to a false (partial) sense of what actually should be done.
ReplyDeleteStep 4 implies an orientation from the perspective of the player trying to solve the problem, which can easily lead to ASSUMING that only moves by the player need be considered. It's easy to misunderstand and misapply even the best of generic guidelines in specific situations.
"The exception that proves the rule" - Figure of speech
"It ain't what you don't know that gets you into trouble. It's what you know for sure that just ain't so." - Mark Twain
I'm fairly certain that if the AI was requested to elaborate on each of its steps, it would have done so. After all, it's simply a matter of combing through millions (billions? trillions?) of documents on the Internet to find "suggestions," collate and weight them and regurgitate the summarized information as KNOW THAT.
It's rather "funny" that the same essential process of searching through gazillions of alternatives with a fine-toothed comb (an evaluation function measured in centi-pawns) works so well for chess programs to be extraordinarily strong grandmasters.
The problem is that useful ideas drown in the trivialities and well-meant advice. Although AI doesn't score well in the "well-meant" department.
ReplyDelete