﻿﻿ MatPlus.Net

Website founded by
Milan Velimirović
in 2006

0:27 UTC
 ISC 2024

Remember me

 CHESS SOLVINGTournamentsRating lists1-Apr-2024
 B P C F

MatPlus.Net Forum General Find the longest "golden sequence" (non-trivial non-endgame attack/defense)

### Find the longest "golden sequence" (non-trivial non-endgame attack/defense)

Initially, I was interested in the question "what are the longest sacrificial middlegame attacks/defences?" (Kasparov's Immortal is one of the most famous sacrificial middlegame attacks). But this question is pretty vague, so I decided to specify it.

What are the longest known "golden sequences" in chess? A golden sequence is defined as...

• A variation which allows the protagonist side (White or Black) to win or draw or gain an advantage. Depending on the protagonist's goal. Variation starts when the protagonist is at least down a piece / without a piece.

• While the variation happens, the protagonist has to stay down material (at least down 1 point); most of the variation the protagonist should be at least without a piece; at least 2 queens should remain on the board; at least 3 heavy pieces (queens or rooks) overall should remain on the board.

• The variation shouldn't be based on an obvious repeating idea OR all of the protagonist's moves can't be checks. (one of those conditions is probably excessive)

Position with the golden sequence doesn't have to be legal. Correctness of the sequence doesn't have to be 100% proven. Moreover, subtle mistakes in the sequence are allowed (because it's just that hard to find longer golden sequences). In principle, even violations of the conditions above are allowed if the sequence follows "the spirit" of the idea.

Here are some examples of golden sequences, mostly half-sound and not 100% proven:

• Kasparov's Immortal has a golden sequence of 14 moves: 25. Re7+ Kb6 26. Qxd4+ Kxa5 27. b4+ Ka4 28. Qc3!? Qxd5 29. Ra7 Bb7 30. Rxb7 Qc4 31. Qxf6 Kxa3? 32. Qxa6+ Kxb4 33. c3+ Kxc3 34. Qa1+ Kd2 35. Qb2+ Kd1 36. Bf1 Rd2 37. Rd7 Rxd7 38. Bxc4. Both sides make subtle mistakes, so the sequence is half-sound. https://www.chessgames.com/perl/chessgame?gid=1011478

• A position from Dubov vs R Svane, 2019 has a golden sequence of 21 moves, if Black makes a couple of subtle mistakes: 21. Rh7+ Kg8 22. Rxf7 Rxf7 23. Qxg6+ Kf8 24. Qh6+ Rg7 25. Bxd5 Ke8? 26. Qh5+ Kd7 27. Qh3+ Ke8 28. Qh5+ Kd7 29. Be6+ Kc6? 30. Qf3+ Kb5 31. Bxc4+ Ka5 32. Qd5+ Bc5 33. b4+ Ka4 34. Qg2 Bxb4 35. Qc6+ Kxa3 36. Bb3 Bf5+ 37. e4 Bxe4+ 38. Nxe4 b5 39. Qc1+ Kxb3 40. Qc2+ Ka3 41. Qa2#. (The actual game was slightly shorter.) It's half-sound. https://www.chessgames.com/perl/chessgame?gid=1978667

• Stoofvlees vs. Igel is one of the most sacrificial engine games I've seen, it has a golden sequence of 21 moves: 1. Nb5!! axb5 2. axb5 b6 3. Ra6 Kb7 4. Bd4 Nc8 5. c3 Ba5 6. c4 g4 7. Rxa5!! bxa5 8. b6 Qd6 9. c5 Qc6 10. Nd2 a4 11. Nb3!! axb3 12. Ra1 Nd6 13. Ra7+ Kb8 14. cxd6 Qc1+ 15. Bf1 Nh6 16. Qe2 Qc6 17. Qa6 b2 18. Qa2 b1=Q 19. Qxb1 Nf5 20. b7 Nxd6 21. Ra6 Nxb7 22. Rxc6. It's sound enough, given that it was played by two engines. (I do realize that engines make mistakes too.) Also, it's the most silent golden sequence I know, Black King received only a single check. https://www.youtube.com/watch?v=5UyuTl6mp58

• Torch played against Leela without a rook for 21 moves: 1. fxe6!! Nxd6 (...) 20. Rh4+ Kb5 21. Qxd5. https://youtu.be/H2qo9fHyXL4?si=qRuOahZXPd3Lsgk0&t=840

• This composition (Shinkman, William Anthony, American Chess Bulletin, Jan 1910) has a checkmate in 36 moves, but it's not a golden sequence because all moves are checks / the length is achieved through an obvious repeating idea: https://lichess.org/analysis/standard/6q1/1p6/1r1P1rpQ/4n3/1P1K1Pk1/8/bP1B1Ppn/6b1_w_-_-_0_1

• Grigory Serper's endless array of sacrifices was just 22 moves long: 17.Nd5 cxd5 (...) 40.Qd6+ Kf7 41.Qxe7+ (there was an unnecessary repetition) with a golden sequence of only 15 moves. https://www.chessgames.com/perl/chessgame?gid=1289099

You get the idea. Getting a lengthy golden sequence is extremelly hard, even if you allow some mistakes.

However, this grotesque, illegal position (by me) seems to have a completely sound golden sequence of 33 moves:
(= 14+20 )

https://lichess.org/analysis/fromPosition/6rb/qq1n1b2/pqp3Bk/pq2Pppq/p2pR2b/pQpP1rQ1/P1PN2P1/2BRK2Q_w_-_-_1_1

1. Qhxh4 Qxh4 2. Rxh4+ Kxg6 3. Qxg5+ Kxg5 4. Nxf3+ Kg6 5. Rh6+ Kg7 6. Rh7+ Kxh7 7. Qxf7+ Rg7 8. Ng5+ Kh6 9. Ne4+ Kh7 10. Qh5+ Kg8 11. Qe8+ Kh7 12. Kf2 Rxg2+ 13. Kxg2 Nf8 14. Rh1+ Kg8 15. Rxh8+ Kxh8 16. Qxf8+ Kh7 17. Ng5+ Kg6 18. Qf6+ Kh5 19. Ne6 Qg7+ 20. Nxg7+ Qxg7+ 21. Qxg7 f4 22. Qf7+ Kg5 23. Qxf4+ Kg6 24. Qh6+ Kf7 25. Qh7+ Ke6 26. Qg6+ Ke7 27. Bg5+ Kd7 28. e6+ Kc8 29. e7 Qd5+ 30. Kf2 Qc7 31. e8=Q+ Kb7 32. Qe7 and on the next move the sequence ends (not enough heavy pieces on the board).

A modification of the position may be a forced checkmate in 37 moves (seems to be confirmed by Stockfish):
(= 15+19 )

https://lichess.org/analysis/rn4rb/qq1n1b2/Nqp3Bk/1q2Pppq/3pR2b/RQpP1rQ1/p1PN2P1/2BRK2Q_w_-_-_5_1

1. Qhxh4 Qxh4 2. Rxh4+ Kxg6 3. Qxg5+ Kxg5 4. Nxf3+ Kg6 5. Rh6+ Kg7 6. Rh7+ Kxh7 7. Qxf7+ Rg7 8. Ng5+ Kh6 9. Ne4+ Kh7 10. Qh5+ Kg8 11. Qe8+ Kh7 12. Kf2 Rxg2+ 13. Kxg2 Nf8 14. Rh1+ Kg8 15. Rxh8+ Kxh8 16. Qxf8+ Kh7 17. Ng5+ Kg6 18. Qf6+ Kh5 19. Ne6 Qg7+ 20. Nxg7+ Qxg7+ 21. Qxg7 f4 22. Qf7+ Kg5 23. Qxf4+ Kg6 24. Qh6+ Kf7 25. Qh7+ Ke8 26. Qh5+ Kd8 27. Qh8+ Kd7 28. Qh3+ Ke7 29. Bg5+ Kf7 30. Qf5+ Kg7 31. Qf6+ Kg8 32. Qg6+ Kf8 33. Bh6+ Ke7 34. Qd6+ Kf7 35. Qf6+ Ke8 36. Qe6+ Kd8 37. Bg5#

or 21... Qd5+ 22. Kg3 Qxe5+ 23. Qxe5 Kg6 24. Qe6+ Kg7 25. Qh6+ Kg8 26. Qg6+ Kf8 27. Bh6+ Ke7 28. Bg5+ Kd7 29. Qxf5+ Kd6 30. Qf8+ Kd7 31. Rb3 a1=Q 32. Nc5+ Qxc5 33. Rb7+ Ke6 34. Qf6+ Kd5 35. Qf5+ Kd6 36. Be7#

How far beyond 37 moves golden sequence can we go?

By the way, I'm not implying that what I did above is a good composition. No, it's merely trying to achieve a record.

If you can define your own term, and claim a record for that term, virtually every position ever constructed can be some kind of record.
It means very little unless you have a buy-in that this term is meaningful.
I, for one, do not buy it.

I view an illegal position as (essentially) a fairy chess variant.
With that door opened, what you call a record might have a record for being the most vulnerable record ever constructed (hello infinite boards).
The only question is: would anyone really care to claim it?

At the very least, please understand, you're now forced to further define the constraints of your record.
The more you must constrain that record, the more you begin to appreciate that it's no more significant than a person claiming to have the world record for being themself.
note: if this term were significant, you would be asked to further define what kind of "mistakes" are permitted.

Show us a rendering which is not so grotesque.
What can you achieve from a legal position with no promoted force?
If you can demonstrate something nice from such a position, perhaps your term will earn a greater degree of acceptance.

Frankly, even then, I am likely to remain skeptical, but at least we can have a discussion about the significance of your term (and the constraints of your "record").

I think one problem is going to be soundness. The longest directmates we know use "obvious repeating ideas" where we can prove the problem sound by proving the repeating idea sound. When it comes to the longest which are not repetitive, soundness is harder to prove.

There are some long directmates at http://chessproblem.net/viewtopic.php?f=7&t=44, the longest being

Lelo, 1951
(= 4+4 )

#53

Solution:
1 Bf8!
(1 Ne2+? Kd2 2 Bb4+ (2 Be7 Be5! 3 Bg5+ Kc2 4 Bd7 Kb1 5.
Ng1 b2 6 Nf3 Ka1) 2... Kc2 3 Ba4 (3 Bd7 Bf6+ 4 Kc8 b2 5 Bf5+ Kb3 6 Bd2
Ka2 7 Bc3 Bg5 8 Be5 b1=Q) 3... Bf6+ 4 Kd7 b5 5 Bxb5 b2)

1... b2!
(1... Kc2 2 Nd3 Bf6+ (2... Bc3 3 Nf2! b2 4 Bd3+ Kc1 5 Bh6+ Bd2 6 Bxd2+ Kxd2 7 Bg6 Kc1 8 Nd3+ Kb1 9 Nb4+ Ka1 10 Nc2+ Ka2 11 Bf7+ Kb1 12 Na3+) 3 Kd7 b2
4 Nb4+ Kb1 5 Bd3+ Ka1 6 Nc2+ Ka2 7 Bc4+ Kb1 8 Na3+ Ka1 9 Bd3)
2 Na2+
(2 Ne2+? Kc2 3 Nd4+ Kc3)

2... Kb3
(2... Kd2 3 Bh6+ Kc2 4 Ba4+ Kd3 5 Bd7)
3 Bd3 Kxa2
(3... b1=Q 4 Bxb1 Kb2 5 Nc3 Kxc3 (5... Kc1 6 Bh6+ Kb2 7 Bf5) 6 Bg7+)
4 Bc4+
{3K1B2/8/1p6/8/8/3B4/kp6/b7 w - - bm #44; 05:23 @ C2/R0/K6/P2/X6;
Ignacio Santos CPU: AMD64 3200+ with 512 MB hash for Chest 3.9 *** 3K1B2/8/
1p6/8/8/3B4/kp6/b7 w - - bm #44; 12:20 @ C2/R0/K6/P2/X6; ChestUCI Vers.4.0
AMD1,2/128}
4... Kb1 5 Bb3 b5 6 Ke7 b4 7 Kf6 Kc1 8 Bh6+ Kb1 9 Kg5 Kc1 10 Kf5+ Kb1
11 Kf4 Kc1 12 Ke4+ Kb1 13 Ke3 Kc1 14 Kd3+ Kb1 15 Bf7
(15 Bc4 {
ChestUCI} b3 16 Kc3 Ka2 17 Bxb3+ Kb1 18 Bf8 Kc1 19 Ba3 Kb1 20 Kd3 Kc1 21.
Ke2 Kb1 22 Bd6 Kc1 23 Bf4+ Kb1 24 Ke3 Kc1 25 Ke4+ Kb1 26 Bc4 Kc2 27 Bd3+
Kd1 28 Bb1 Ke2 29 Be3 Kd1 30 Kf3 Ke1 31 Bf4 Kd1 32 Bg5 Ke1 33 Ke3 Kd1 34.
Kd3 Ke1 35 Bh4+ Kf1 36 Ke3 Kg2 37 Be1 Kf1 38 Bf2 Kg2 39 Ke2 Kh2 40 Kf3
Kh3 41 Be1 Kh2 42 Bd2 Kh3 43 Bg5 Kh2 44 Kf2 Kh3 45 Bf5+ Kh2 46 Bf4+ Kh1
47 Be4#)
15... b3 16 Kc3 Ka2 17 Bxb3+ Kb1 18 Bf8 Kc1 19 Ba3 Kb1 20 Be6
Kc1 21 Bg4 Kb1 22 Kb3 Kc1 23 Bf5 Kd2 24 Bb1 Kc1 25 Ka2 Kd2 26 Bb4+ Kd1
27 Be7 Kd2 28 Bf6 Ke3 29 Kb3 Kd2 30 Bg5+ Ke2 31 Kc3 Kf3 32 Kd2 Kg4 33.
Bh6 Kh5 34 Be3 Kg4 35 Ke2 Kh5 36 Kf3 Kh4
37 Bc5 Kg5
(37... Kh3 38 Be7 Kh2 39 Kf2)
38 Be7+ Kh5 39 Kf4 Kh6 40 Bf6 Kh5 41 Kf5 Kh6 42 Bg5+ Kg7 43 Ke6
Kf8 44 Kd7 Kf7 45 Kd6 Kf8 46 Ke6 Ke8 47 Bg6+ Kf8 48 Bh6+ Kg8 49 Bb1 Kh8
50 Kf6 Kg8 51 Kg6 Kh8 52 Bg7+ Kg8 53 Ba2#

(Annotations by Ignacio Santos, who cites Andreas Möbus as the source.)

Another long directmate is:

Christopher Jeremy Morse, after G. Rinder
The Problemist, Jul 1980, C6385; P1197233
(= 2+9 )

#43

Solution: 1. Kd1 Kg7 6. a8=D Kb3 7. Qa1 f5 8. Qb1+ Kb5 15. Qa5 Kb8 16. Qa6 Kc7 17. Qb5 Kc8 23. Qe5 Kf8 24. Qe6 Kg7 25. Qf5 Kh8 26. Qxf3 Kg7 27. Qf5 Kh8! 28. Qg5 Kh7 29. Qe5 Kg8 30. Qf6 Kh7 31. Qf8 Kh6 34. Qe5 Kd3 35. Qb4 38. Qe4 Kh3 39. Qe6+ Kg3 40. Qf5 Kh4 41. Qg6 Kh3 42. Qf5 Kh2 43. Qh4#.

10. ... Ka4 11. Qb2 Ka5 12. Qd4 f3 13. Qc4 transposes to the mainline

Rosie Fay, thank you for those examples. But those compositions have little heavy pieces on the board or have "dummy" pieces which are completely immobilized/isolated from the rest of the board. What are the longest variations in compositions which involve two queens (+ some pieces) and don't have hopelessly immobilized/isolated pieces? I understand that such compositions may be even harder to verify.

I think even not 100% verified positions and variations are OK for now.

Kevin Begley,

QUOTE
If you can define your own term, and claim a record for that term, virtually every position ever constructed can be some kind of record.
It means very little unless you have a buy-in that this term is meaningful.

I agree. That's why I tried to give many examples of golden sequences in human and computer games, to demonstrate how hard it is to get such a sequence. I did give legal examples with human and computer games. Perhaps I didn't provide enough?

Here's a forced checkmate in 13 moves, from the game Dubov vs R Svane (so, 100% legal). It doesn't simplify into an endgame + both queens remain on the board + it's not based on an obvious repetition + not all of White's moves are checks. So, the sequence has a lot of "golden" qualities, even though technically it's not a golden sequence because White ends up with material advantage before 14th move.

(= 10+11 )

30. Qf3+ Kb5 31. Bxc4+ Ka5 32. Qd5+ Bc5 33. b4+ Ka4 34. Qg2 Bf5+ 35. Ka2 Rd5 36. Qxd5 Qd7 37. Qb7 a5 38. Ne4 Bxd4 39. exd4 Bxe4 40. Qxe4 Qf5 41. Qc6+ b5 42. Bb3#

How far beyond 13 moves can we get? Engine games suggest that about 21 moves is possible.

QUOTE
At the very least, please understand, you're now forced to further define the constraints of your record.
The more you must constrain that record, the more you begin to appreciate that it's no more significant than a person claiming to have the world record for being themself.
note: if this term were significant, you would be asked to further define what kind of "mistakes" are permitted.

Ideally, no mistakes are allowed. My own sequence supposedly has no mistakes. But I tried to show that even subtle mistakes may be of little help in constructing a long sequence. I don't think making the constraints hyper-precize matters at the moment. Creating long variations (which aren't based on repetitions) with many heavy pieces on the board seems really hard. For most reasonable constraints one can think of.

...

If you're a chess fan, it's logical to be interested in the longest forced variations. But does it mean that the curiosity of a chess fan should be satisfied with tablebase records and purely endgame studies and compositions with repetitions? Of course not! Because such length is either achieved by having little amount of pieces on the board or is "trivial" in some sense (because it has obvious repetitions).

So, it seems we need to create some restriction which rules out tablebase records and purely endgame studies and compositions with repetitions. I tried to give an example of such restriction. But I believe that whatever restriction we come up with, creating a long forced variation is gonna be really hard.

I think it's a bit strange that chess fans don't seem too interested in defining different types of length records. Should be a pretty important topic. Not all ways to achieve length are the same. My position with 37 moves shows a pretty non-trivial way to do so - so I do think it's something valuable.

@Tom,

If you're going to talk of achieving a "record", I merely suggest you should start with a legal position (at the bare minimum -- no promotions would be nicer).
What you have here is essentially a fairy chess position. Once you open that door, there's no end to how far this might go.

As I said, the only question is: will anybody care enough to upend your record?

It doesn't make much sense to speak of "records" in fairy chess.
For example, you might try to achieve a fairy proofgame with the most en passant captures.

Christian Poisson, Pascal Wassong, and Thierry Le Gleuher showed seven in Messigny 1998, using Madrasi (PROBID=P1000948).
Michel Caillaud showed eight in Phénix 1994, using Madrasi (PROBID=P0005964).
I made a joint with Kostas Prentos which showed ten in Mat Plus 2008, using Circe Parrain (PROBID=P1095546).
Kostas went on to show thirteen en passant captures in RIFACE 2015, using Anti take&make (PROBID=P1380673).
Apologies if I've left somebody out.

Is the record itself significant? Not per se.
These conditions are apples and oranges.
Nobody involved in this pursuit expects their record will endure.
Tomorrow, a new fairy condition might allow a trivial rendering with 14 en passant captures.

What does endure?
These problems are notable for the artistry with which they achieve what you might consider a record.
Forget the record. You don't need to count en passant captures to appreciate them.

Furthermore, I don't think any of these problems is fully computer tested (though the first one mentioned is likely verifiable), so I reject the notion that you need computer verification for a record (there's no computer verification for the longest orthodox proofgame, and for decades there was no verification for the longest helpmate -- and these are perhaps the best known records in composition).

The problem I have with your "record" is that it is based upon a highly contrived term (one you invented yourself).
I can appreciate that this problem has value to you (and perhaps to others), but I'd encourage you to avoid any care about achieving "records" (except in cases where they are clearly defined, and they drive you toward some artistry that has nothing to do with the record itself).

If I may offer a friendly piece of advice...
It would seem a shame to squander your composing talents chasing a record length for some contrived term you invented yourself.

Find some record that has artistic value, and push yourself to better it (just remember that the goal is not to obtain 14 en passant captures in a proofgame; your goal should be to show something artistsic).

Never take credit for a record (unless you expect it to endure for decades against tenacious attempts to extend it).

That is why it makes no sense to speak of records in fairy chess; no record has any real hope to endure when you are free to invent new fairy elements.
The only records which endure are the ones so contrived that nobody cares enough to upend them.

I ask you honestly: can you really claim this would endure (if we all considered it an important record)?
If you can honestly say yes, please clearly define the constraints allowed, so there will be no dispute when somebody cares to extend this grotesquery beyond your imagination.

I pity the composer who engages in that pursuit.
I can tell you right now, with certainty, I will make no effort to extend this (nor even to study it), because it's clear to me this will be a waste of time (I see no hope for something artistic to emerge from this process).
Not if you added a prize of a thousand dollars would I even consider it.

Hint: the brilliancy of a sacrifice is completely lost in a position teaming with material.
I will prefer Shirov's famous endgame sacrifice over Kasparov's hyper-extended brilliancy, any day.

QUOTE

Christian Poisson, Pascal Wassong, and Thierry Le Gleuher showed seven in Messigny 1998, using Madrasi (PROBID=P1000948).

C+ by Jacobi in 18 seconds. It's Anti-PWC, not Madrasi.

@Tom

I certainly break your 2 queens-rule, but this study I consider a golden sequence and I believe it is in the spirit of Kasparov-Topalov. White is more than a piece down in the first 26 moves.

https://eg.org.ua/chessgame/nielsens-4447-33g1h7/

@Kevin Begley,

I don't agree that my term is contrived. Or rather I don't agree that the "purpose" of my term is contrived. The purpose of my term is this:

• Define what a "middlegame" is.

• Exclude trivial middlegame wins. (For example, just the starting position but Black doesn't have a knight, so White trivially wins after a long middlegame.)

• Define how to count the length of the variation.

• Exclude variations which use obvious "tricks" to extend length. Which try to violate the "spirit" of the condition (e.g. by adding pieces which can't move).

I think it's pretty natural. Anybody can invent their own term which fulfills the same purpose. But the result is gonna be the same. Getting long middlegame variations is really really really hard (which is obvious if you look at human and computer games). And my position with 37 moves still counts as "a long non-endgame variation which doesn't use repetitions to get its length".

QUOTE
Is the record itself significant? Not per se. (...) I'd encourage you to avoid any care about achieving "records" (except in cases where they are clearly defined, and they drive you toward some artistry that has nothing to do with the record itself).

I understand what you mean. Artistry is more important than records 99% of the time. But when we talk about hyper-extended combinations (like Kasparov's), then I think length matters in of itself, without artistry. Because such hyper-extended combinations show chess' potential for complexity, middlegame's potential for complexity. "How complicated can a position get in principle?" - the point is to answer this question from different perspectives. Not only from endgame perspective, but also from middlegame perspective.

QUOTE
Never take credit for a record (unless you expect it to endure for decades against tenacious attempts to extend it).

What "credit" do you mean, what exactly do you mean by "taking credit"? I take only very temporal credit, because getting 37 moves was hard (obviously 37 can be extended at least by a couple of moves; I can even do it myself). I don't expect to get into any history books or composition books or journals. I'd be happy if people destroyed the "record" on the very first day.

@Steffen Slumstrup Nielsen,

For some reason I can't open the link directly, but I saw the position.
(= 8+9 )

I can't see the intended variation. I get about 19 moves before black rook gets captured. Then it's just a single heavy piece on the board, it's too little, too far from a "middlegame".

1. Rd2 Qc7 2. Rh2+ Kg6 3. Ne2 Qxa7 4. Nf4+ Kf5 5. Rh5+ Ke4 6. Re5+ Kf3 7. Bd4 Kxg3 8. Ng2 Kh3 9. Rh5+ Kg3 10. Rf5 b5 11. Bxa7 Kh3 12. Bf2 g3 13. Rg5 gxf2+ 14. Kxf2 Bd7 15. Rg3+ Kh2 16. Nf4 Rc8 17. Rg2+ Kh1 18. Ne2 Rc2 19. Kg3 - and the rook has to be given up.

But even 19 moves is already cool! I completely agree that it follows the spirit of trying to find the longest variations with the most pieces playing. I haven't seen a 100% legal and 100% proven such variation longer than 19 moves.

In studies, what are the longestt variations with "at least 2 heavy pieces on the board + many minor pieces"?

[Event "4.p Prusikin-45"]
[Site "?"]
[Date "2023"]
[Round "?"]
[White "Nielsen=S"]
[Black "(+4447.33g1h7)"]
[Result "1-0"]
[SetUp "1"]
[FEN "2q1br1n/Q4p1k/np6/8/1P4p1/P1NR2P1/1B6/6K1 w - - 0 1"]
[PlyCount "67"]
[EventDate "2023"]

1. Rd2 (1. Nd5 \$4 Qc2 \$19) (1. Rd6 \$2 Qc7 2. Qxb6 Qxb6+ 3. Rxb6 Nc7 \$11) (1. Ne4 \$2 Qc2 2. Qe7 Qxb2 \$11) (1. Qe7 \$2 Qe6 2. Qh4+ (2. Qxf8 Qe1+) 2... Kg8 3. Ne4 f6 4. Nxf6+ Rxf6 5. Qxf6 Qe1+ \$11) 1... Qc7 (1... Nc7 2. Ne4 f6 3. Qxb6 Ne6 4. Rf2 Qd7 5. Nxf6+ Rxf6 6. Rxf6 \$18) 2. Rh2+ (2. Qxa6 \$19 Qxg3+) (2. Nd5 \$4 Qxg3+ \$19) 2... Kg6 3. Ne2 \$3 (3. Nd5 \$2 Qxg3+ \$11) 3... Qxa7 (3... Bb5 4. Nf4+ Kf5 5. Rh5+ Ke4 6. Re5+ Kf3 7. Ng2) 4. Nf4+ Kf5 (4... Kg5 \$2 5. Rh5#) 5. Rh5+ Ke4 6. Re5+ Kf3 7. Bd4 (7. Ng2 \$4 b5+ \$19) 7... Kxg3 \$1 (7... Qe7 \$2 8. Rxe7 Kxg3 9. Ng2 Kh3 10. Nf4+ Kg3 11. Ne2+ Kh4 12. Bf6+ Kh5 13. Nf4+ Kh6 14. Re5 ( 14. Re2 g3 15. Re5) 14... Kh7 15. Rh5+ Kg8 16. Rxh8#) 8. Ng2 (8. Nd5 Kh3 9. Rh5+ Kg3 10. Rf5 b5 11. Bxa7 Kh3 12. Nf4+ Kh4 13. Ng2+ Kh3 14. Bf2) (8. Nh5+ Kh3 9. Nf4+ Kg3 10. Ng2) 8... Kh3 (8... Qe7 9. Rxe7 Kh3 10. Nf4+ \$18) 9. Rh5+ Kg3 10. Rf5 b5 \$1 (10... Kh3 11. Nf4+ Kg3 (11... Kh4 12. Bf6+ Kg3 13. Ne2+ Kh3 14. Rh5#) 12. Ne2+ Kh4 13. Bf6+ Kh3 14. Rh5#) 11. Bxa7 Kh3 12. Bf2 \$1 (12. Nf4+ Kh4 (12... Kg3) 13. Ng2+ (13. Rh5+ Kg3 14. Rf5 Kh4 15. Ng2+ Kh3 16. Bf2) 13... Kh3 14. Bf2) 12... g3 13. Rg5 \$1 gxf2+ 14. Kxf2 Nxb4 \$3 (14... Bd7 \$2 15. Rg3+ Kh2 16. Nf4 (16. Ne3 Bh3) 16... Rc8 (16... Rg8 17. Rxg8 Ng6 18. Nxg6 fxg6 19. Rxg6 Nb8 20. Rb6 Nc6 21. Rxb5 \$18) 17. Rg2+ Kh1 18. Ne2 Rc2 19. Kg3 Rc3+ (19... Rxe2 20. Rxe2 \$18) 20. Nxc3 \$18) 15. Rg3+ \$3 (15. axb4 \$2 Bd7 16. Rg3+ Kh2 17. Nf4 (17. Ne3 Bh3 \$11) 17... Ra8 \$1 18. Rg2+ Kh1 19. Ne2 Ra2 (19... Ra3 20. Ng3+ Rxg3 21. Kxg3 \$18) 20. Kg3 \$2 (20. Rg7 \$11) (20. Rg5 \$11) 20... Ra3+ \$19) 15... Kh2 16. Nf4 Nd3+ (16... Bd7 17. Rg2+ Kh1 18. Rg5 \$18 (18. Rg7 \$18)) 17. Rxd3 ( 17. Nxd3 \$4 Ng6 \$19) 17... Bd7 18. Rg3 (18. Rxd7 \$2 Ng6 \$11) 18... Ra8 (18... Rg8 19. Rxg8 \$18) (18... Rc8 19. Rg2+ Kh1 20. Ne2 Rc2 21. Kg3 \$18) 19. Rg2+ Kh1 20. Ne2 (20. Nh5 Rxa3 21. Ng3+) 20... Rxa3 21. Ng3+ Rxg3 22. Kxg3 f5 (22... Bh3 23. Kxh3 Ng6 24. Kg3 f5 25. Ra2 f4+ 26. Kh3 Kg1 27. Rg2+ Kf1 28. Rxg6) 23. Kf2 \$1 (23. Rd2 \$5 f4+ 24. Kxf4 Bh3 \$11) 23... f4 (23... Be8 24. Rg7 \$1 (24. Rg8 Ng6 \$11) 24... Bg6 25. Rg8 Nf7 26. Rxg6 Kh2 27. Rg7 \$18) 24. Rg8 (24. Rg7 \$2 Bh3 (24... Bf5 25. Rg5 Bh3 \$11) 25. Rh7 Kh2 26. Rxh8 b4 27. Rh4 b3 28. Rxf4 b2 \$11) 24... Nf7 (24... Ng6 25. Rxg6 Kh2 26. Rh6+ Bh3 27. Rh4) 25. Rg7 Nh6 26. Rh7 Bh3 (26... Kh2 27. Rxh6+ Bh3 \$18) 27. Rxh6 Kh2 28. Rh4 (28. Rb6 Bd7 29. Rh6+ \$18) 28... f3 (28... b4 29. Rxf4 b3 30. Rb4) 29. Kxf3 (29. Rb4 Be6 30. Rh4+ \$18) 29... b4 30. Kf2 \$1 (30. Rxb4 \$5 Bd7 \$1) 30... b3 31. Rb4 Be6 32. Rb6 \$1 b2 (32... Bf5 33. Rxb3 Bc2 34. Rb2 Bf5 35. Rb5 Bg6 36. Rb6 Bf5 37. Rh6+ Bh3 38. Rh7 Kh1 39. Rxh3#) 33. Rxe6 b1=Q 34. Rh6# 1-0

@Joost,

Thank you for the correction.
I only later realized that this problem was published after the "record" was set with 8 ep captures (not before).
Apologies for those mistakes, but I am glad I included this, because it illustrates why fairy "records" will always be vulnerable.

@Tom,

That's why it makes little sense for an author to claim a "record" in an unorthodox problem setting.

If some promoter wants to generate interest in a problem, by pretending it a meaningful and enduring "record", let that trespass weigh upon their conscience (not your own).

Composers go down this false path -- where they consider length a benefit.
Early on, I went down the same path myself.
I made several lengthy problems where you must sacrifice your queen, just so you can promote another queen with check, so you could sacrifice the second queen.
They were awful.
I was misguided -- rather than seeking chess problem artistry, I was composing problems as a substitute for the middlegame tactical brilliancies I wanted at the board.

To be fair, there is some personal value in attempting to tame a very long sequence (note: I expect your "golden sequence" criteria, if pursued at all, would result in the exact opposite of taming a position).

But, is it worthy of publication?

Many years ago, I sent Milan Vukcevich a #7, with a queen deflection key which leaves the white King in check (and leaves black well ahead in material).
In fact, at first glance, a mate looks hopeless after the sacrificial key is accepted.

Milan was kind enough to explain why I should not publish such a problem.

He showed me a #3 problem that achieved the same idea.
I admit, at the time that comparison problem seemed banal (it was hardly a sacrifice at all, by comparison).
But, he put some good questions to me:
1) if this idea is "better" in 7 moves (versus 3), why not 8 moves, or 9, or ... 30?
2) what kind of problems would such a publication encourage?
3) would that encouragement (which signals that composers should completely ignore the concept of ECONOMY OF MOVES) benefit or harm problem chess?

You have turned ECONOMY OF MOVES completely on its head.
And then, you turned ECONOMY OF MATERIAL completely on its head.

What have you gained for it?
Where is the artistry?

If you really want to pursue this anti-economical path...
a) why are you limiting yourself to an 8x8 board?
b) why are you limiting yourself to illegal orthodox positions?
c) why are you not ALL-IN?

You should not limit yourself to two or three fairy elements -- load up dozens of fairy elements (whatever allows your sacrifices to appear all the more spectacular).
Go whole hog.
Burn down the ECONOMY OF SQUARES.
Burn the ECONOMY OF FAIRY ELEMENTS to the ground, too (it's always the last fig leaf).

Take a flame thrower to any guideline you can find for economical chess composition.
Just keep asking yourself: What Would Montag Burn?

I found this path to be deeply unsatisfying. It's empty. There's no enduring beauty down that road.
If you don't already see that reflected in your problem, keep going.

I hope that, in the chess problem world, the paths are still numerous and varied! Some composers strive to construct problems that improve on records (while those such as myself can only sit on the sideline an admire the experts' work). But our world still has room for those who don't wish to take that path.

Turning now to tasks, once a task has been defined, I think it a fair enough question to ask: "How economical can a problem be, that shows this task?". With most tasks, economy can be measured in the two ways Kevin has mentioned: economy of moves and economy of material. But seeking the longest sound problem meeting certain criteria is a valid, well defined task. The longest sound directmate with both queens in the diagram and with no repeating manoeuvre is such a task. That does not, however, entail abandoning economy: problem-length measures the task achieved, but we may still impose a limit on material to achieve economy.

(13) Posted by Tom [Wednesday, Feb 7, 2024 12:00]

@Joost de Heer, @Steffen Slumstrup Nielsen,

Thanks, I see now. That's 20-21 moves with at least two rooks and three pieces on the board. I think that's a lot!

@Kevin Begley, I don't think my position should be published. Only orthodox positions are allowed, illegal or not.

@Rosie Fay,
QUOTE
But seeking the longest sound problem meeting certain criteria is a valid, well defined task. The longest sound directmate with both queens in the diagram and with no repeating manoeuvre is such a task. That does not, however, entail abandoning economy: problem-length measures the task achieved, but we may still impose a limit on material to achieve economy.

My motivation is a bit more abstract, less well-defined: finding the longest non-endgame forced variations. However, we can explore this abstract topic through more concrete tasks (such as yours). That's why I defined "golden sequences", to give an example.

Maybe it's too early to limit ourselves to legal positions? Because getting long middlegame variations is very hard and right now we have little understanding (or maybe I'm just ignorant) what ideas can be used for getting such variations. We do have legal examples from human and computer games, but they are not 100% proven or not 100% sound. One 100% sound and 100% proven example is only 13 moves...

MatPlus.Net Forum General Find the longest "golden sequence" (non-trivial non-endgame attack/defense)