Double pruning
Have a look at the following position:
 |
| Diagram 1. White to move |
r1b2rk1/1pq1bppp/p3pn2/3n2N1/3R4/2N1B1P1/PPQ1PPBP/R5K1 w - - 0 15
#15. Ni H. vs. Adly, A., Shenzen 2011
Although you can't possibly know this yet, the decisive focal points of this position are:
- h7 [2:2}
- d8 [0:3]
The first moves are:
- 15.Nxd5 Qxc2
The landscape has totally shifted now: There is no acceptable alternative for black. He simply loses his knight, and after Nxf6 he would lose by Qxh7#.
 |
| Diagram 2. White to move |
White has given up his queen. And even the focal point h7 has ceased to exist. This means that white's logical reasoning is totally guided by the forcing moves alone now. The list of forcing moves has shortened to "CM". Since white has sacrificed his queen, only a check or a mate in 1 threat can save him now. This pruning of the list of possible forcing moves is even more important than the pruning of the branches of the tree of analysis as a consequence.
The list of forcing moves of black has become even shorter: "C". A counter attack (in case white would threaten a mate in 1) can only be based on a check.
We even have a list of candidate moves now for white:
- Nxf6+
- Nxe7+
The logic is guided by the double pruning of the list of forcing moves (CMCT is pruned to CM) followed by the pruning of the list of candidate moves. The A:D-ratio (#attackers:#defenders) is only judged after a move has taken place (in the mind).
 |
| Diagram 3. White to move |
Notice how the A:D-ratio of d8 has changed dramatically to [1:1} now after
- 16.Ne7+ Kh8
Rf8 has become overloaded now.
This means that I must overhaul my study plan. Instead of focusing on new mini-skills, I must rewrite my logic for the 1100 positions I have already absorbed.Which may sound dramatically but is actually a relief!
I believe there is one big mistake in the ratio od A-D
ReplyDeleteIn the original position - you write:
"Although you can't possibly know this yet, the decisive focal points of this position are:
h7 [2:2]
d8 [0:3]"
And it is actually....
h7 [2:2]
d8 [1:3]
Why white attacks d8 square once? Because the piece is standing on this LINE (LoA). It is especially valid as there are no pawns along the file.
Thats why "Notice how the A:D-ratio of d8 has changed dramatically to [1:1]..." as the d8 square was attacked once from the very start.
Now I am curious if that was a type (0 instead of 1) or you count the A-D only as the legal (immediate) moves.
By the way I solved the original position with the conceptual way, but my visualization made me fool as I thought there is no mate as Rxc8+ does not lead to mate due to the Ra1 controlling c8 square. But of course it is Rd8+ (after Nxf7+) and the back rank mate.
I highly recommend the CM Azel Chua YouTube chess channel. It is very much underrated one as he does not throw any BS into it, neither the explositions, ads, etc.
However there are at least 20 videos you must watch as they are the same type of deep concepts explained the easy way as you explore and experiment with my friend Tempo. I watched these and I will be rewatching a few times, to see how can I merge the concepts and practical application of the author (Azel Chua) and yours!
https://www.youtube.com/@AdviceCabinet/featured
PS. I am not sure if blogspot let me show my blog profile name, but if it does not - this is Tomasz (the same when we explored the salty mines 10 years ago: it was a hell of the challenge).
Hi Tomasz! It is not a typo.
DeleteBut don't get distracted by what is essentially an arbitrary choice.
The essence of the post is:
*the A:D ratio is not a good guide for re-engineering a tactic. You can only use it to judge the result of a move after you have made it in your mind.
* Apparently an improved version of CMCT is needed.
I will try to concoct such improved version.
I have a Chessable course of CM Azel Chua. I remember I was not impressed. I will dive into it again. I might be wrong. Thx!
Tomasz: I’m glad you’re part of this conversation!
DeleteI've found that periodically Google Blogspot will lose sync (in terms of sign in) with Google. The "clue" is presented by "Comment as:" being gray instead of orange, and your comment will be logged as from Anonymous rather than from you.
Simply log out of Google, wait 5-10 seconds (for it to propagate that info throughout the Google blogosphere) and then sign in to your Google account again. Then go to the Temposchlucker blog. You should see the "Comment as:" in orange.
On rare occasions, I've had to repeat the process. That is usually caused by having an unread Security message. Just read it, then repeat the process.
The videos contain good new ideas!
DeleteGreat to hear that Tempo! I knew you will enjoy these videos! And they are very short, consise and educational at the same time. I immediately spotted the same type of ideas and explanations what you provide here - at the YouTube videos of Axel Chua!
Delete@Robert
Yeah, you may be right. I use Brave browser and that maybe another reason I am not in the highest league of Google members ;)
Anyway it is not a big deal for me. The most important the ball is rolling. By the way - I have been reading the blog for the last 7-8 years regularly even though I vary seldom make comments
Tomasz
I just watched one of the videos:
DeleteCM Azel Chua - Mastering Chess Calculation with the Burger Technique
LINK: https://www.youtube.com/watch?v=kwkSBSmaYSo&t=158s
There are intriguing parallels between some of his suggestions and the evolution of this blog: different terminology, same concepts.
Thanks, Tomasz!
In the interest of brevity, I’ll not repeat my comment on 1 JUN 2026, when you presented this position with an extensive analysis. Re: PoPLoAFun REDUX, May 31, 2026
ReplyDeleteI assume that the first thing that System 1 throws up for attention is the interaction revolving around the h7-square. CMCT would identify the “chunk” around that area as the most forcing because it involves a check, with possible mate. System 2 checks and finds that it fails if implemented immediately because the A:D ratio on the h7-square is B.A.D. [2:2]—which is NOT fatal. There might be a glimmer of recognition of Function of the BNf6: it provides the A:D balance for Black on the h7-square. It’s not a weakness with respect to this “chunk” but it MAY be WRT some other “chunk;” need information about other “chunks.”
Searching for another potential “chunk” follows immediately after realizing that the first “chunk” offers nothing useful for the NEXT MOVE. Perhaps the attention is drawn to the d5-square [3:2]. Capturing on d5 may lead to material gain. WAIT! The BNf6 is constrained by a higher forcing level [CMCT] function: it must not move or White can checkmate on h7. The A:D ration shifts to [3:1]. Additionally, 1. Nxd5 is a fork (of BQc7 and BBe7) plus a discovered attack (WQc2 → BQc7). Duplo moves are ALWAYS “interesting”!
Perhaps at this point System 1 identifies the “vis-a-vis” between the two queens. If the WNc3 captures on d5 (trying to win material through multiple exchanges on a given square), Black can simply capture the queen on c2. That pin on the WNc3 doesn’t look appealing.
THIS MAY BE WHERE THE CLUB PLAYER GIVES UP! It’s just too complicated to figure out all those interrelationships, so play something “safe” instead.
Forging ahead, are there any possible CMCT forcing moves available after 1. Nxd5 Qxc2? Of course: 2. Nxe7+ (highest level of forcing move) Kh8.
It is likely that at this point (if not before), System 1 “SEES” that the path to d8 is now opened for the WRd4 with back rank implications. It may also note that the BRf8 just acquired an additional function: guard the back rank from the White rook incursion, in addition to preventing checkmate on f7 by the WNg5.
Intermezzo: Is there a saving grace? The BNf6 can interpose on g8 to block the WRd8, so that relieves the Black rook of the “protect the back rank” function. Nope, System 2 says that won’t work: the WRd8 can capture on g8 with mate because it is protected by the WNe7.
So the move sequence to the finale is clear after traveling along the logical sequence: 1. Nxd5 Qxc2 (almost forced) 2. Nxe7+ Kh8 (forced) 3. Nxf7+ Rxf7 (forced) 4. Rd8+ Rf8 (almost forced—“toss another log on the fire” to delay the inevitable) 5. Rxf8+ Ng8 (forced) 6. Rxg8#.
Summarizing: I don’t think there is just one path to the solution. The order of the System 1 pattern recognition of at least 4 “chunks” is unimportant; it is the overall logical sequence of moves that wins the day. As you noted, CMCT are excellent pruning shears!
I'll check out those videos - thanks for the suggestion!
The question is, does every considered move contain enough information to decide to go forward OR to prune the branch?
DeleteCheck: Qxh7+ does it work? No, Nf6 prevents it. PRUNE.
Can I get rid of Nf6? No. PRUNE
Attack the black Queen: Nxd5. Can Nf6 take it NO because of Qxh7# NEXT exd5. Qxc7 winning the black Queen STOP
Qxc2 Nxf6+ (scaling up to the king again) gxf6 NO MATE because the white queen has disappeared. PRUNE.
Nxe7+. Kh8 NEXT Nxf7+ Rxf7 Rd8+ winning.
It seems that the moves contain enough information to decide between pruning and the next forcing moves.
There is double pruning power: my forcing moves and the constraint of the moves of the opponent. And when you run out of forcing moves, there can't be no tactic.
So maybe it's just a matter of becoming proficient in it.
I am pretty curious if you explored connecting between the known positions as this could be a nice idea, right? For example if we know the final position we can solve in 10 seconds (3 moves tactics) how much it is beneficial if we can visualize it easily and efficiently from the A position (the starting one) to the B position (the one we know it is easily winning), before reaching the position C (the final one).
DeleteTomasz
I may be wrong (it wouldn't be the first or the last time THAT occurred – I am frequently WRONG!), but I think it has to do with something I learned from GM Valeri Beim. He referred to certain moves in the middle of a logical sequence as being "resulting moves." Not in the sense of producing a given result, but a cognitive result created by working through the earlier moves. They are not readily visible from the beginning but only become "obvious" after a move or two is made (either on the board or in the mind). I don't think (short of memorizing the pieces-on-squares AND the entire logical sequence) that the entire logical sequence of moves can be recognized by System 1 as a whole. AFAIK, System 1 relies upon smaller units of recognition, melding them together via abstraction to produce an overall recognition pattern. Otherwise, System 1 would be using “pieces-on-squares” and we “know” (or think we know) that “chunks” and/or “templates” don’t reside at the physical level.
DeleteHow would I know: I’m NOT a cognitive psychologist!