FEN viewer
Now we have discovered the trick, it is time to think about how we can acquire the trick more efficient?
The trick starts with a plan. The plan must be triggered by salient cues. How can we learn these salient cues the easiest?
 |
| Diagram 1. White to move |
5rk1/pQrnqppp/4p3/2bn4/4N3/2B1PP2/PP2BP1P/3RK2R w K - 0 17
Everything revolves around f6. Who are the defenders and how to deflect them?
 |
| Diagram 2. White to move |
Once you KNOW that everything revolves around f6, it is easy to come up with a plan.
- 1.Qxc7 to distract one defender
- 1. ... Nxc7
- 2. Rxd7 eliminating the second defender
- 2. .. Qxd7 distracting the third defender
- 3. Nf6+ luring away the 4th defender by a royal fork
This leads to the following position:
 |
| Diagram 3. White to move |
3. ... gxf6
4.Rg1+ Kh8
5.Bxf6#
I'm thinking about making a FEN viewer that helps to find the salient cues. In this case, you need the mate pattern (diagram 3) and the fact that everything revolves around f6 (diagram 2). Notice how simple it is to finde the moves one the salient cues are seen.
I read down to the diagram and decided to attempt to understand the “requirements of the position” prior to looking at your description. My basis for doing this was to determine if I could perceive the salient surface level cues pointing toward the essence of what to do.
ReplyDeleteGiven my tendency for “safety first,” I gravitated toward figuring out if I could wreak havoc with the queen before ignominiously retreating it to safety. Nothing immediately sprang into my mind. There must be some other factor(s) pointing toward the solution.
Hmm, WBc3 aims into the heart of the Black king’s position along the long diagonal a1-h8. There is a half-open g-file, also allowing an attack on the Black king. A “combination” of both these facts could indicate a possible stock mate. (I couldn’t remember the name; I had to look it up: Morphy’s Mate.)
The problem is how to open the g-file, since the WBc3 already occupies the long diagonal. The PoP is the f6-square; the BPg7 must be induced to capture on f6. At first glance, the WNe4 could be sacrificed on f6 IFF there are no other Black defenders of the f6-square.
BUMMER! There are 4 defenders of f6: BNd5, BNd7, BQe7 and, of course, BPg7. That brings to mind the idea of capturing to divert the defenders.
Suddenly, a role for the WQb7 springs unbidden into mind: capture the BRc7, “forcing” the BNd5 to recapture—the first defender is diverted away from the f6-square! As soon as I saw that, the next step immediately became obvious: capture the BNd7 with the WRd1, attacking the BQe7 AND controlling e7 and d8 as retreat squares for the Black queen. The second defender is removed!
OH NO! What if the Black queen skedaddles to h4? After WRxc7, White has gained two knights and a rook for his queen, more than enough material compensation.
So, Black will most likely capture BQe7xWRd7—the third defender is removed from the f6-square! Now a sacrifice of the WNe4 on f6 FORCES Black to capture it with the g7 pawn. Otherwise, Black will lose the BQd7 to the royal fork, followed by a fork of the BBc5 and BRf8 after WNf6xd7. Black would lose too much material.
After g7xf6, the stage is set for Morphy’s Mate: WRg1+, followed by WBxf6#.
That final sequence feels almost anticlimactic.
QED
After writing all that verbiage, I looked at the rest of this post to see what I missed. Not much, apparently. I note in passing that the process described is a quintessential example of GM Tisdall's variation processing.
I'm very interested in your idea of a FEN viewer that would identify salient cues. That is a great example of the kind of training tools we need!
I was hoping you got an idea or two too.
DeleteI'm working on a set of 100 problems from Ehrwichs book.I'm looking for salient cues, and whether there is an hierarchy in saliency.
Often it is a chain from the sitting duck backwards. What rules are there?
Diagram 1 came from the following game (following 16.… Rc7)—and ended in a draw(!):
Delete[White "Egor Dranischnikov"]
[Black "Henrik Teske"]
[WhiteElo "2341"]
[BlackElo "2534"]
[Date "2012.6.8"]
[Round "4.10"]
[Result "1/2-1/2"]
[Event "4th Rheinland-Pfalz Open 2012"]
[Site "Altenkirchen GER"]
[Eco "D04"]
1.d4 Nf6 2.Nf3 d5 3.e3 Bg4 4.c4 e6 5.Nc3 Nbd7 6.Qb3 Bxf3 7.gxf3 dxc4 8.Bxc4 Nb6 9.Be2 c5 10.dxc5 Bxc5 11.Qb5+ Nfd7 12.Ne4 Rc8 13.Bd2 O-O 14.Rd1 Qe7 15.Bc3 Nd5 16.Qxb7 Rc7 17.Qxc7 Nxc7 18.Rxd7 Qh4 19.Rxc7 Bxe3 20.Bb4 Bxf2+ 21.Kf1 Bb6 22.Bxf8 Qh3+ 23.Ke1 Qh4+ 24.Kf1 Qh3+ 1/2-1/2
It is a prime example of the sports aphorism “It ain’t over till the fat lady sings.” It also demonstrates the resilience found in most disadvantageous positions, confirming the Theory of Infinite Resistance formulated by Australian chess player Bill Jordan. I put this in an earlier comment (April 27, 2025).
Link:The Ultimate Guide to Defending Lost Positions in Chess
Link:https://juntaikeda.substack.com/p/9-the-ultimate-guide-to-defending
GM Ian Rogers:
It is a theory designed to encourage players to fully utilize the defensive resources available in a bad, or even strategically lost position. The theory postulates that when a player makes a serious mistake or reaches a bad position, if he or she continues to try to find the best possible moves thereafter he or she can put up virtually infinite resistance and should not lose…Of course some positions are beyond even perfect defence but their number is far smaller than imagined.
I was far too hasty in concluding that Black would NOT play 18.… Qh4, retaining the Black queen in return for a material deficit. Perhaps the fact that Black was significantly higher rated may have dictated that choice of moves as well as the final result.
GM Stockfish opines that White has a significant advantage (+4.73) after 19….Bxe3. White’s 20. Bb4 throws away the win. It was GM Stockfish’s 27th choice; 20. Rg1 was the 1st choice.
One of the things I’ve realized about the past few examples is that there is a sequence of thoughts that are in essence a cascade of consequences rather than a linear step-by-step thinking process. I really hesitate to refer to that sequence as a “rule” or a heuristic because the component elements vary drastically from one specific problem to another. However, there is a regularity to the sequence that is not necessarily obvious after a casual (NOT causal) glance.
ReplyDeleteThe sequence begins with an “obvious to the most casual observer (including club-level amateurs)” fact about the position. Let’s note that starting fact in the last two examples you have given. I repeat the FEN to avoid having to go back to the example posts.
1. FEN: 2r1r1k1/ppp1bppp/8/n4b2/2PPqB2/P4N2/4BPPP/R1Q1R1K1 w - - 0 18
2. FEN: 5rk1/pQrnqppp/4p3/2bn4/4N3/2B1PP2/PP2BP1P/3RK2R w K - 0 17
There is no commonality in the pieces-on-squares in the two positions.
In the first position, there is an obvious (and salient) tactical clue: the LPDO BNa5. Because it is an undefended piece, it is a magnet for attention, a “weakness.” That should automatically “trigger” a “search” for a way to exploit it. As we all “know,” directly attacking ONE loose piece with no other goal usually is a waste of time and will often misplace the attacking piece. There is a “chunk” associated with that BNa5: it involves BNa5 (with a “threat” on b3 to fork WRa1 and WQc1 [I overlooked this threat completely]), BPb7 (a potential defender), WPc4 [B.A.D. 1:1], WQc1 and WBe2 (defending c4). There is no other target available within that chunk. Ergo, we are going to have to find a different chunk in order to add an additional target and connect those two targets with a line of attack.
In the second position, there is also an obvious (and salient) tactical clue: it is the chunk associated with the WBc3, the open long diagonal (a1-h8), the half open g-file (with a White rook available to occupy it in one move), and the Black king position (the king itself and the g7 and h7 pawns protecting it). It is the confluence of the lines of attack through g7 (to g8 and h8) that should trigger recognition of a stock mating pattern. It is irrelevant whether we know the formal name of this mating pattern (Morphy’s Mate) or not. The label is useful for communicating concisely with others about this pattern but serves no function in the processing sequence. Once the stock mating pattern skeleton is recognized as potentially applicable, then the natural subsequent steps of the sequence will be to identify and eliminate all defenses of the f6-square.
Note that the initial identification of a standard “chunk” automatically triggers a standard response. In the first case it is to find or create a second weakness. In the second case it is to find potential ways to eliminate the identified defenders of the critical f6-square. As Nimzowitsch opined, we attack targets because any defender is weakened (constrained by function) by the mere fact of being a defender. The logic is inexorable.
The natural sequence moves from one recognized pattern to another, driven by the logical consequences of each observed pattern. That IS a “rule” that is always applicable. If a required pattern is “missing in action” within our repertoire or the required pattern is “known” but is occluded by tunnel vision focused on one aspect (or the wrong aspect) at any point in the logical sequence, then we will only find the solution sequence with great difficulty or through sheer dumb luck.
The logical narrative is a chain from beginning (the first move) to end (the sitting duck).
ReplyDeleteYou can start to pick up the chain at any point, and work your way forwards and backwards from there.
I'm looking for maximum pruning. If you start at the beginning, it is basically trial and error.
For to start in the middle, CCT is designed. Since every move must have a forcing element in it.
But there are situations where 80% of the forcing moves is irrelevant. So you introduce a lot of redundant checks (pun intended).
The best way is to start with the end. From K to g7 to f6 branching to e7, d7, d5 to c7. Notice how ALL branches are relevant.
So the question is: how do we know that g7 is relevant. What makes it stake out from other PoPs?
I'm 100% in your corner when it comes to maximum pruning! The fewer the potential moves/variations, the more likely we are to avoid mistakes and overlook something vitally important.
ReplyDeleteYou ask:
How do we KNOW that g7 is relevant?
In short, we don't KNOW at the beginning of our investigation. (I started to use "analysis" instead of "investigation," but that has connotations I would rather avoid.) As GM Tal once wrote [paraphrasing], "Something [some "chunk"] in the position is recognized as familiar and from that familiar chunk the chase of the bluebird of happiness [search for a combination] begins." That chase is NOT by trial and error nor necessarily by trying forcing moves until we either stumble on the proper sequence or we exhaust all options with no solution in sight. That initial familiar "something" guides the way forward, not just by trying forcing moves (although that is often the case), but by following a logical sequence such as "Here is a weakness. What resources are available in this specific position to take advantage of that weakness?"
All we do initially is recognize the outline of a KNOWN [to us] pattern that is potentially promising; familiarity breeds attempt.
Dr. Lasker contrasted the combination player with the position player by the direction of their thoughts: the combination player tries forcing moves, hoping to find an end point that is favorable. The position player has a favorable end goal in mind and works backward, hoping to find a viable sequence of moves that will lead to the end goal position. Neither type of player is right or wrong.
The play of the master follows a different path. Position play is completed by combination play. Combinations are available when one player has made moves not in keeping with the overall requirements of the SPECIFIC position; Dr Lasker referred to these as false values. Position play is based on the true values, which preclude the OPPONENT from wreaking havoc with a combination. That's why the great combination player Zukertort twice failed to defeat Steinitz for the world championship. Dr. Lasker categorized Steinitz as rather poor at combinations but magnificent at position play.
Perhaps that's just a PhD in philosophy musing in print, but I don't think that is the case.
Alright. But what do I tell my FEN viewer?
DeleteRandom thoughts:
DeleteYour FEN viewer can start by identifying surface-level weaknesses, then uncovering the logical consequences of those weaknesses. For example, I think LPDO is more basic than B.A.D. squares because LPDO involves a single unprotected piece, independent of everything else on the surface level. Elementary tactics (fork, pin, etc.) occur nearer to the surface level than diversion, interference, etc. It is partly a matter of more pieces involved, and more complicated relationships that unfold over time.
After identifying a surface-level weakness, there are logical consequences that follow from that designation. Eventually, there are functional interrelationships between individual chunks, giving a complete picture of the specific position.
How to program the sequence discovery process is a mystery to me.
One odd thing I noted about FEN that may have programming ramifications is that there is an undocumented "long form" for FEN. Instead of compacting the numerical value for consecutive squares into a single digit greater than 1, each empty square can be denoted by "1". This gives 64 square/piece values. Why do it this way? Because it MAY simplify using regular expressions in order to search for relationships on the board.
Bitboards can be also used for discovery of relationships.
There are multiple paths to the discovery of functional relationships between pieces and squares. Since the FEN viewer is not intended to play chess within a reasonable time, there is no particular need to make the search process hyper efficient if the purpose is human training.
When I know what I want to know, I usually can find a way to program it. I was thinking in layers and buttons. A button that shows all targets. A button that shows all attackers. A button that shows all PoPs. A button that shows all defenders etc. Each button shows a list, the user chooses from the list to move the focus of the program.
DeleteI'm not intending to write something that can generate a sequence. The sequence happens in the mind of the user. Did you notice that you can visualize the variation that you discovered in this post without effort?
It is about guiding the mind towards salient points, not about solving a puzzle. we have Stockfish for that.
But we must first have a clear idea what is salient. I have not.
Less is more. If only square g7 was highlighted in diagram 1, I could reconstruct everything myself. But how can my program know that g7 is the most important PoP?
In the end, I don't expect to build the program, but to get better by designing it alone. Only when I'm asked to give lessons to other people, there is a reason to actually build it.
OTOH, who knows. With AI programming has become a piece of cake.
DeleteYou observed:
DeleteDid you notice that you can visualize the variation that you discovered in this post without effort?
I did and was somewhat surprised by that effect. It has also changed how I approach a position. It somehow feels easier to "SEE" salient points.
I asked Copilot this question:
DeleteAre there any known chess training tools (software) that identify points of pressure, lines of attack and functional relationships?
Here is the answer I got:
Short answer:
Yes — several modern chess‑training tools explicitly identify points of pressure, lines of attack, and functional relationships between pieces. The strongest evidence from your search results points to ChessWorld’s micro‑training tools, which directly analyze tactical pressure, open lines, and piece‑function interactions. Additional tools like Lotus Chess, ChessTempo, ChessFlare, and Chessable provide related (but less explicit) capabilities.
Long answer:
♟️ Tools that explicitly identify pressure, lines of attack, and functional relationships
1. ChessWorld Micro‑Training Tools
ChessWorld provides the clearest match to your query. Several of its tools directly analyze pressure points, attacking lines, and functional piece relationships:
Tools matching “points of pressure”
• Killer Squares — “discover tactical hotspots instantly,” i.e., squares under high tactical pressure.
• Safety Check — anticipates threats along invisible lines, revealing pressure vectors.
• Loose Piece Hunter — identifies unprotected or overloaded pieces (functional defender relationships).
Tools matching “lines of attack”
• Liberated Piece Trainer — identifies newly opened lines and discovered attacks.
• Safety Check — highlights threats along files, ranks, and diagonals.
Tools matching “functional relationships”
• Loose Piece Hunter — detects overloaded defenders (functional dependency).
• Safety Check — shows how pieces interact defensively and offensively.
• Killer Squares — highlights squares whose control depends on multiple pieces.
ChessWorld is the only tool in your search results that explicitly surfaces pressure geometry, line‑based threats, and defender‑graph‑like relationships.
[End Copilot]
I didn’t include everything from Copilot.
The tools are “trainers” that have predefined problems.
While browsing the ChessWorld.net site for an overview, I did not find a specific FEN viewer function nor the capability to upload FEN/PGN into any of the tools. That capability may be there, but I didn’t want to register in order to do a thorough search.
It may provide some functional ideas for a FEN viewer.
I entered these three following comments in error to a previous post. I tried deleting them from that location without success.
ReplyDeletePART I:
Another example of deciding what is salient and what is not in a given position.
FEN - 1rb2r1k/1p1n2q1/p2Q2p1/P2Npn1p/2P1N2P/6P1/1R3PB1/1R5K w - - 13 36
Full game:
LINK: Tigran Petrosian vs David Bronstein (1956) Petrosian's Mortal
LINK: https://www.chessgames.com/perl/chessgame?gid=1033916
Black's last move was 35... Nf5. The greatest defensive player in the history of chess as well as one of the greatest tacticians played a horrendous 36th move: 36. Ng5.
What salient feature did he NOT "SEE?"
Temposchlucker asks (in different comments):
I'm looking for salient cues, and whether there is a hierarchy in saliency.
Often it is a chain from the sitting duck backwards. What rules are there?
So the question is: how do we know that g7 is relevant. What makes it stake [stand] out from other PoPs?
Alright. But what do I tell my FEN viewer?
But we must first have a clear idea what is salient. I have not.
WHAT ARE THE “CLUES” AS TO WHAT IS SALIENT?
The “clues” are weaknesses or vulnerabilities.
In the interest of brevity, I’ll use the term “weakness” from here on out.
Weaknesses may be on the surface and totally obvious (as in the Petrosian example given above) or they may be hidden deeper and require drilling down into the interrelationships between the pieces and the squares.
In every case, the triggering condition for System 1’s pattern recognition as the beginning point MUST a weakness. The weakness can be localized to a single piece/square or it can encompass several pieces/squares or several “chunks”. It must EXIST in the opponent’s position. HOW that weakness came into existence (whether through oversight, as in Petrosian’s example above) or from inadequate SKILL in avoiding weaknesses, or as a forced sequence of choices based on avoiding “the lesser of two evils” is not relevant.
Any specific weakness has corresponding stereotypical responses to exploit it. A LPDO piece must be attacked directly. A B.A.D. piece is exploited by adding attackers or removing/diverting defenders or adding to their functions until overloaded.
Philosophically, Dr. Lasker provided the basis for attacking, defending and maintaining the status quo in a balanced position. A balanced position does NOT mean a drawn position; it is a position in which the correct moves by both players will lead to another balanced position. Neither player has an uncompensated advantage. Once an uncompensated advantage is obtained the player with the advantage MUST attack (using combinations) based on that advantage; otherwise, the defender MAY find a way to restore the balance. The defender must make the barest minimum concession that will minimize the uncompensated advantage.
If there is no weakness, there cannot be a sequence of moves to take advantage of what does NOT exist!
Let’s revisit the two positions in the previous example posts.
PART II:
ReplyDelete1. FEN: 2r1r1k1/ppp1bppp/8/n4b2/2PPqB2/P4N2/4BPPP/R1Q1R1K1 w - - 0 18
A quick glance at position 1 should trigger recognition that BNa5 is LPDO. That is the most obvious surface-level weakness. System 1 should recognize that pattern AND an associated process to take advantage of it: look for a way to capture it. WQc1 can play to c3, attacking BNa5. Consequences (using System 2): WQc3 un-defends the B.A.D. WBf4. Black can simply capture the bishop, maintaining the material balance. NO CIGAR!
System 1 pivots to the next associated process: look for another potential weakness that has a line of attack to the BNa5. The WBe2 is on the wrong color squares; the WNf3 cannot attack the BNa5 in the foreseeable future, and would also be vulnerable to capture from a knight, so by instantaneous elimination of alternatives, the only LoA from BNa5 is to the BBf5 along a rank. Only a rook or queen can attack along a rank. So, the next obvious factor is to get the WRe1 to e5.
Drilling down deeper, the BQe4 and WBe2 are the only two pieces preventing an immediate move to e5. The BQe4 has the function of “protecting” BBf5 AND is vulnerable because of limited mobility along the e-file. It requires little imagination to “SEE” that moving WBe2 forces the Black queen to either d3 or c2 in order to maintain contact with BBf5.
WBe2 can move to either f1 or d1. On f1, it does not prevent a fork on b3 nor prevent the Black queen from playing to c2. WBd1 is the only move covering both possibilities. Black plays the BQd3 (forced because it cannot retreat along the e-file). WRe3 attacks the Black queen again, forcing capture on c4.
The way is now open to fork BNa5 and BBf5. The Black queen cannot protect both pieces.
All of this reasoning is a direct result of the initial salient weakness of BNa5, followed by known patterns of attack which are disclosed at each step. It ends by attacking the initial weakness (BNa5) along with a second weakness (BBf5), winning a piece. Here we end with an assessment that White will have a winning advantage if he can execute it successfully; it is ASSUMED that he has the requisite SKILL to do so.
PART III:
ReplyDelete2. FEN: 5rk1/pQrnqppp/4p3/2bn4/4N3/2B1PP2/PP2BP1P/3RK2R w K - 0 17
Here the process is different. System 1 “SEES” two lines of attack converging on the Black king position: WBc3-f6-g7-h8 and WRg1-g7-g8 although the White rook is not yet on g1. This is the rudiments of a stock mating pattern (Morphy’s Mate; although System 1 may not regurgitate the label).
What associated process comes along with that pattern recognition? The g-file must be opened by force and with tempo while maintaining the White bishop on the long diagonal. System 2 says that is currently not possible: White bishop capturing on g7 eliminates the WBc3 and the mate pattern disappears. Moving the WRg1 should be done with tempo. Logic dictates that with two White pieces aimed at the f6-square which has 4 Black defenders, the only way to conquer g7 is to eliminate its defenders.
The first defender removed (by diversion) is capturing WQb7xBRc7. To maintain material balance, Black is forced to recapture BNd5xWQc7. One defender removed!
The second defender removed (by capture) is WRd1xBNd7, attacking the BQe7.
Here there is a split between the originally envisioned Morphy’s Mate and gain of material, depending on Black’s choice. If BQh4, the mating pattern is removed from consideration. White captures WRd7xBNc7 with advantage. If BQe7xWRd7, the third defender is removed from f6 and the original mating sequence is now available with additional benefits adding force to the sequence. WNf6+ also forks the BQd7. Capturing g7xf6 opens the g-file and prevents capture of the Qd7, which would in turn create another knight fork of BBc5 and BRf8. White finally reaches the envisioned Morphy’s Mate pattern with WRg1+, followed by WBc3xf6#.
Notice in all three example given that there is an initial recognition of a salient weakness, a corresponding salient process to take advantage of that weakness, and so forth until reaching the final goal.
If there are multiple surface-level weaknesses, it will not matter which one is recognized first. Eventually, all of the salient weaknesses will either be incorporated into the sequence (the principle of two or more weaknesses) or those that are critical for success will be identified and the remainder will be ignored.
One sitting duck is always there: the king. All squares around the King are legitimate focal points. When there are no weak pawns and promotion is far away, the next best pivot point is an invasion square. Or inducing a weakness.
ReplyDeleteI wonder if I can develop a sense for the salient cues now I know what I am looking for.
One of the things I dislike in some puzzle books is the absence of references to the source games. It helps to have the context for a puzzle. The reason for looking at the game is to get a 'feel' for the mistakes that created the weaknesses that made the puzzle solution possible. In particular, it is important to distinguish between unforced errors (IE, oversights, blunders, whatever) and weaknesses created without adequate thought as to possible consequences, usually by drifting along without serious thought in critical positions, making stereotyped moves according to the 'rules'. It also helps to identify critical moments in the game where things can go horribly wrong for either player. It certainly seems to me that I've improved my skill for "SEEing" the clues.
DeleteI forgot: one of the things that is missing in puzzle books is the rationale for filtering that the author(s) used to select the puzzle set. As my Mom would say, "The road to hell is paved with good intentions." The author(s) filter by their own criteria for what is interesting and good - for them.
DeleteSince WYSIATI (Kahneman: What You SEE Is All There Is) in the puzzles, you have no idea what biases may have driven the selection. Consequently, our training may be based on availability bias. We think we are getting a well-rounded set of problems that will (eventually) address all of our training needs. That is an assumption that should be viewed with suspicion.
You are probably right.. I will try not to forget add the game data when I have it.
DeleteI noticed that there are a lot of problems that authors borrow from each other.
The matter is, I often prefer context-less problems. Learning to play the board. But since I don't like studies (for training purposes), I assume I agree with you.
The main criterion is the frequency of occurrence.
Gobet talks about 300,000 chunks, while my interpretation of MDLM and CT-art 3.0 suggests that it could be closer to 3,000 too. With might lead to the thought that there is a lot of overlap within problems.
I don't know. Only since 3 years I have begun to look at my own games seriously. The latest months, with more knowledge about how to close the Vukovic gap, I try make my whole game a logical narrative from begin to end. So my preferences are likely to shift in the near future.
What I need to see is the salient knight fork on f6 against Kg8 and Qd7. From there you can find the rest by logical reasoning.
ReplyDeleteWhen I cannot find this salient tactic, I will not find the tactic during a game when this position should arise.