- Such variations can be found in computer games.
- Such variations can be used for testing engines.
- Maybe we can even use computers to generate such variations.
Motivation
People love long middlegame attacks. Kasparov's Immortal (~14-move attack), Serper's Immortal (~23-move attack), Polish Immortal (~13-move attack).
However, strangely no one seems to ask how long can such middlegame attacks get. Even though it seems like an important topic about exploring the limits of chess' complexity.
The current length records are all about the endgame (tablebase records) or are in some sense "trivial", using very specific tricks to achieve length (e.g. repeated zugzwang). Such length records have zero traits in common with immortal games.
Defining middlegame and triviality
How do we define a "middlegame"? How do we define "triviality"?
Instead of defining it directly, I think a better approach is to consider parameters which make the length of a checkmate objectively harder to achieve:
- (P) The amount of pieces on the board.
- (P) The amount of pieces which make a move in the checkmating variation.
- (P) The amount of heavy pieces. Rooks and queens.
- (P) The amount of non-check moves.
- (P) The size of material disadvantage of the checkmating side.
- (P) The degree of naturalness of the position.
- (N) The amount of obvious repetitions.
- (N) The degree of isolation of pieces. For example, if a piece is 100% isolated from the game forever, it's bad.
So, instead of a single definition of a "non-trivial middlegame checkmate", think about a PARAMETER SPACE we can explore. I know, it's inconvenient that there isn't a single simple definition for the task at hand. But that doesn't make the topic less important. If we care about chess, we should care about the maximal length of middlegame attacks. We should care about DIFFERENT types of length records, not just about 1 or 2 types.
How to research that?
I think there's two main avenues of research.
The first avenue is analyzing computer games. They contain the longest and most sound middlegame attacks. The second avenue is creating arbitrary chess positions.
I'll share some of my puzzles. Then some of the other people's puzzles and some computer games.
My puzzles
Disclaimer: all positions here were verified with 40MB browser Stockfish, sometimes it misses a lot of stuff. I think the positions are valuable even if they have refutations.
Above is a checkmate in 34 moves. At every half-move White is at least 16 points of material down. White never gains material advantage. Queens never get traded off. No piece is forever isolated from the game. Checkmating the Black King requires finding A LOT of completely original moves, no repetitions. Not every White's move is a check or even a capture. The entirety of the checkmate takes place in the "middlegame".
8 | |||||||||
7 | |||||||||
6 | |||||||||
5 | |||||||||
4 | |||||||||
3 | |||||||||
2 | |||||||||
1 | |||||||||
a | b | c | d | e | f | g | h |
qrqqqqqq/qqqqqBqN/qrrPnpqq/Nnr1nppr/qrBNPq1b/rP1B1p1k/pPQPQQRP/nNBNKR1b w - - 1 1
Above is a checkmate in 32 moves. At every half-move White is at least 126 points of material down. If you want to try guessing it move by move: https://lichess.org/study/hBcBtt33. Click "reveal the solution" to reveal the next move, it doesn't reveal the whole thing.
Above is a fragment of my "checkmate in 42 moves" puzzle. White is never up material. Most of the half-moves White is at least 3 points down.
8 | |||||||||
7 | |||||||||
6 | |||||||||
5 | |||||||||
4 | |||||||||
3 | |||||||||
2 | |||||||||
1 | |||||||||
a | b | c | d | e | f | g | h |
rn2Q1rb/qq1n1b2/Nqp3Bk/1q2Pppq/RprpR1rb/brpP1rQ1/qPPN2PQ/rbBRK2Q w - - 6 1
Above is a checkmate in 43 moves. At every half-move White is at least 26 points of material down. There's one fixable flaw though: White can find a shorter checkmating variation on the 3rd move.
8 | |||||||||
7 | |||||||||
6 | |||||||||
5 | |||||||||
4 | |||||||||
3 | |||||||||
2 | |||||||||
1 | |||||||||
a | b | c | d | e | f | g | h |
qRnrnrkn/rb1q1bqp/Rqnp1Qbq/nR2QnQP/rR1P1BN1/BNPQpb1R/qPB2P2/qNNNK2n w - - 3 1
Position above shows what a non-trivial checkmate in 52 moves could look like. Unfortunately the position has multiple flaws without any obvious fix. White has shorter checkmates (10. Rf5+, 12. Bf6) and Black has a draw (25. ... Rxg7).
Others' compositions
This composition by Steffen Slumstrup Nielsen has the following properties. White is at least down a piece for the first 26 moves. At least two heavy pieces are on the board for the first 21 moves. It IS a long non-trivial attack, despite being closer to an endgame.
8 | |||||||||
7 | |||||||||
6 | |||||||||
5 | |||||||||
4 | |||||||||
3 | |||||||||
2 | |||||||||
1 | |||||||||
a | b | c | d | e | f | g | h |
8/7Q/2r1p3/2rkr3/2rrr2Q/7K/8/8 w - - 0 1
This composition by Philip Bondarenko is unique, because it shows the longest checkmate given the biggest material disadvantage (13 points of material) with the maximal freedom of movement (all Black's rooks are centralized in a completely open position). It's "trivial" in terms of all moves being checks and there being an obvious repeating pattern, but it maximizes other non-trivial parameters.
8 | |||||||||
7 | |||||||||
6 | |||||||||
5 | |||||||||
4 | |||||||||
3 | |||||||||
2 | |||||||||
1 | |||||||||
a | b | c | d | e | f | g | h |
b2b1Rqb/Nnp2brR/1pN1N1Pq/n2K4/pP6/2k1Pp1n/Rpp3br/rbBbrrqq w - - 0 1
Above is a bastardized version of a composition by Otto Blathy and Heinrich Meyer. White gives checkmate in 45 moves. Being at least 67 points down at every half-move. However, it's "trivial" in terms of all moves being checks and there being an obvious repeating pattern.
The composition above is unique, because despite having a giant material advantage White needs ~24 moves to get a dominating position.
In this puzzle by David Zimbeck White launches an insane sacrificial attack. In some variations White takes slightly more than 31 moves to gain material advantage. The puzzle is not based on a repeating pattern.
Computer games
In the game Stoofvlees vs. Igel, there happened a non-trivial 21 move sequence. Throughout which White were down at least 5 points of material (not every half-move, but every move), at least 2 queens and 5 heavy pieces overall remained on the board.
In the game Torch vs. Leela Chess 0, there happened a non-trivial 20 move sequence.
So, even with computer games finding a 25+ move attack down material is not easy. But my knowledge of computer chess is extremely limited.
Conclusions and Resources
Here are Lichess studies with the positions I talked about:
https://lichess.org/study/sTon08Mb
https://lichess.org/study/XZjpjz0c
Even illegal positions don't make creating a long non-trivial middlegame checkmate easy.
The longest non-trivial checkmate may be only 50-60 moves long. We never gonna know where a hard ceiling will hit, but it's definitely got to be less than 100 moves.
Finding "the longest non-trivial middlegame draw" has to be more complicated than finding the longest non-trivial middlegame checkmate.
Many positions with long (30+ moves) non-trivial checkmates were built from positions from real games. It gives many positions a kind of unique and recognizable look. Underneath all the insanity those positions look "vaguelly typical".
What you can do without much effort
You can post here a legal middlegame position with a 10+ checkmate (the longer, the better). The entirety of the checkmate doesn't have to be non-trivial (because we can make it non-trivial later, by modifying the position). I could try expanding it (by making the position illegal). That's how I created most of my puzzles, the original positions came from my bullet games. And one of the positions I was lucky to find on Reddit.