# Thread: Longest possible chess game

1. ## Longest possible chess game

Originally Posted by Boris
Perhaps a real version of HAL 9000 would simply announce move 1.e4, with checkmate in, say, 38,484 moves.
I don't know about you, but I haven't seen a game of chess that lasts for 38000 moves...

2. Originally Posted by michael.mcguirk
I don't know about you, but I haven't seen a game of chess that lasts for 38000 moves...
Indeed not and I doubt it would be possible to get there without invoking an opportunity to claim a draw by 50 move rule.

16 pawn on the board and if they all make 7 moves that's 112.

32 pieces on the board, say 29 captures (to leave the kings and 1 piece to deliver mate).

Plus the 50 moves at the start of the game.

(112 + 29 + 1) * 50 = 7100 moves.

3. Originally Posted by Boris
Indeed not and I doubt it would be possible to get there without invoking an opportunity to claim a draw by 50 move rule.

16 pawn on the board and if they all make 7 moves that's 112.

32 pieces on the board, say 29 captures (to leave the kings and 1 piece to deliver mate).

Plus the 50 moves at the start of the game.

(112 + 29 + 1) * 50 = 7100 moves.
It's a bit more complex than that because some of the pawn moves need to be captures otherwise the pawns can never get past.

Also each pawn moves six times not seven.

Here is something I wrote about this that was published in Bruce Pandolfini's column on Chess Cafe in 2001:

Dear Bruce Pandolfini,

In your column you ask the length of the longest possible game assuming that the 50-move draw rule is used. I have often seen a figure of 5949 moves quoted. That was before King vs King was an automatic immediate draw, but the calculation was also incorrect anyway and I believe (though I'm not absolutely certain) that the longest game is presently drawn with Black's 5898th move. Note that the figure changes with slight changes in the Laws of Chess.

The calculation given by Mc Murray is wrong for two reasons. Firstly while
there can be 96 pawn moves and 30 captures, unless some of those captures
are by (not of!) pawns, then the pawns never get past each other and make
all their moves. It is necessary to have 8 captures by pawns so all the
pawns can pass each other and promote, so the figure to be multiplied is 118
not 126, as 8 of the pawn moves are also captures.

Secondly, while Mc Murray multiplies by 49.5, this is wrong. The game is
drawn only after 50 moves by *both* players without a pawn move or capture, so so long as the side making the pawn move or capture is the same one to make the last pawn move or capture, then that adds 50 moves to the total, not 49.5. So the base figure is 118x50, or 5900.

It's a bit trickier than that because there must be several changes in whose
turn it is to make the pawn move or capture through the game. Assuming
Black makes the first capture, we need a switch to White making the captures so that White can get pieces out and give them up on squares which double White's pawns on files to leave gaps for Black's pawns to pass through. Then we need another change back to Black making these captures. At this stage both sides have unpromoted pawns so we need another switch for White to promote those pawns and take Black's pieces, and a final switch for Black to take White's surviving pieces. Each switch costs half a move, so on Black's 5898th move, a king capture of White's remaining piece, the game is drawn as only two kings are left and FIDE Law 1.3 applies immediately. (Does USCF have this law too?)

To illustrate how to do a 5898 move game, here's an example. Both sides
just move other pieces around in the meantime without triple-repeating:

Black takes White's knights by gxh6 and bxa6 (100 moves)
Black's knights take White's queen and rooks (150 moves)
White plays d3 and e3 (99.5 moves)
White takes four Black pieces with pawns: hxg3, exf4, dxc4,axb3 (200 moves)
White takes Black's other three pieces with bishops (150 moves)
Black takes White's bishops : fxe6, cxd6 (99.5 moves).
White's pawns are on the b,c, f, g files, Black's are on the a, d, e and h
files.
Black makes 44 pawn moves including eight promotions (2200 moves)
White makes 42 pawn moves including eight promotions (2099.5 moves)
White takes Black's eight promoted pieces (400 moves)
Black takes White's eight promoted pieces (399.5 moves)
King vs King, game drawn immediately by FIDE law 1.3

TOTAL 5898 moves

If anyone thinks they can make one go for longer, I would like to see them
construct an outline game like the above to prove it rather than just
supplying an abstract "calculation".
Nobody wrote in disputing the above.

5. Originally Posted by Kevin Bonham
Nobody wrote in disputing the above.
Well OK then: to avoid a possible draw claim, double repetition must be avoided.

6. ## Longest possible chess game

I decided to bite the bullet and try and construct the longest possible chess game. A Guinness record intent, perhaps? There are considerable content written on this in the internet, but not much of it very substantiated. In fact there is only one calculation that gets it right. Hats off for Kevin Bonham!

Here are a few links with the wrong results
http://blog.chess.com/kurtgodden/the...ble-chess-game
http://www.chess-poster.com/english/...d_you_know.htm
http://chessobserver.wordpress.com/t...st-chess-game/
http://infinityiitd.com/is-chess-an-...-complex-game/
http://www.chesscafe.com/text/bruce22.pdf
http://www.chesscafe.com/text/bruce23.pdf

In retrospect after doing my record attempt, it follows that I lost 6 halfmoves unnecessarily along the way, so the final result was 5895 moves instead of 5898. I only wish i had seen the below cookbook recipe by Kevin Bonham before I started!

I will try to add the two files FIDE_longest_possible_game.pgn (the actual Portable Game Notation file with the moves) and FIDE_longest_possible_game.png (graphics illustrating the handling with BabasChess and Waxman).

I quickly found out that there were few programs actually supporting such long games. Chessbase comments out all moves after move 300. SCID parses all 5895 moves correctly but then hangs and must be aborted. I probably tried a dozen programs (including web applets) but only BabasChess lived up to the promise of both editing and showing all moves, while making it possible to jump to given move numbers.
Waxman was useful because it allows for checking casual 3-times repetition of a position underway and 50-moves rule, so it helps eliminate mistakes in the move order, but also claims draw with (=) in the move list while allowing for continuation of the game. Waxman did have a small caveat, it would claim a draw after promotion thus not considering that this is another pawn move which zerosets the 50-move counter!

In essence this post from Kevin outlines the below

Originally Posted by Kevin Bonham

Here is something I wrote about this that was published in Bruce Pandolfini's column from Chess Cafe in 2001:

Originally Posted by Kevin Bonham to Pandolfini
Dear Bruce Pandolfini,

In your column you ask the length of the longest possible game assuming that the 50-move draw rule is used. I have often seen a figure of 5949 moves quoted. That was before King vs King was an automatic immediate draw, but the calculation was also incorrect anyway and I believe (though I'm not absolutely certain) that the longest game is presently drawn with Black's 5898th move. Note that the figure changes with slight changes in the Laws of Chess.

The calculation given by Mc Murray is wrong for two reasons. Firstly while
there can be 96 pawn moves and 30 captures, unless some of those captures
are by (not of!) pawns, then the pawns never get past each other and make
all their moves. It is necessary to have 8 captures by pawns so all the
pawns can pass each other and promote, so the figure to be multiplied is 118
not 126, as 8 of the pawn moves are also captures.

Secondly, while Mc Murray multiplies by 49.5, this is wrong. The game is
drawn only after 50 moves by *both* players without a pawn move or capture, so so long as the side making the pawn move or capture is the same one to make the last pawn move or capture, then that adds 50 moves to the total, not 49.5. So the base figure is 118x50, or 5900.

It's a bit trickier than that because there must be several changes in whose
turn it is to make the pawn move or capture through the game. Assuming
Black makes the first capture, we need a switch to White making the captures so that White can get pieces out and give them up on squares which double White's pawns on files to leave gaps for Black's pawns to pass through. Then we need another change back to Black making these captures. At this stage both sides have unpromoted pawns so we need another switch for White to promote those pawns and take Black's pieces, and a final switch for Black to take White's surviving pieces. Each switch costs half a move, so on Black's 5898th move, a king capture of White's remaining piece, the game is drawn as only two kings are left and FIDE Law 1.3 applies immediately. (Does USCF have this law too?)

To illustrate how to do a 5898 move game, here's an example. Both sides
just move other pieces around in the meantime without triple-repeating:

Black takes White's knights by gxh6 and bxa6 (100 moves)
Black's knights take White's queen and rooks (150 moves)
White plays d3 and e3 (99.5 moves)
White takes four Black pieces with pawns: hxg3, exf4, dxc4,axb3 (200 moves)
White takes Black's other three pieces with bishops (150 moves)
Black takes White's bishops : fxe6, cxd6 (99.5 moves).
White's pawns are on the b,c, f, g files, Black's are on the a, d, e and h
files.
Black makes 44 pawn moves including eight promotions (2200 moves)
White makes 42 pawn moves including eight promotions (2099.5 moves)
White takes Black's eight promoted pieces (400 moves)
Black takes White's eight promoted pieces (399.5 moves)
King vs King, game drawn immediately by FIDE law 1.3

TOTAL 5898 moves

If anyone thinks they can make one go for longer, I would like to see them
construct an outline game like the above to prove it rather than just
supplying an abstract "calculation".
Nobody wrote in disputing the above.
I will only dispute 2 small things. First the above calculation is correct in claiming a base figure of 118x50 = 5,900 only because of the pecularity that first 50 moves can be made, then 118 times the life of the game is prolonged with 50 moves by moving a pawn or capturing a piece, however the last time this is done the count stops *before* the next 50 moves because with bare kings it is an instant draw. I am sure you made the right thinking but it is not so clear that the first free 50 moves are matched by the last 50 moves that can't be moved with bare kings.

Secondly the sequence you presented

Black makes 44 pawn moves including eight promotions (2200 moves)
White makes 42 pawn moves including eight promotions (2099.5 moves)
White takes Black's eight promoted pieces (400 moves)

Here I see a problem, although it is not clear to me if it is or not, of maintaining two kings out of check, and still have enough room to make 50 moves for each pawn move without repeating the position, while still shuffling around those eventually 16 queens. I would say there is some technical difficulty in that, while not claiming it is impossible. But consider the following rearrangement:

Black makes 44 pawn moves including eight promotions (2200 moves)
White takes Black's eight promoted pieces (400 moves)
White makes 42 pawn moves including eight promotions (2099.5 moves)

Now it's much easier! Before white attempts to queen, he gobbles up all the black queens, but losing no halfmoves with this rearrangement.
I conclude that the longest possible game is in fact possible as above, with the last moves e.g. 5898. Qh1+ Kxh1 ½-½
Here is my outline based on the above:

Black takes White's knights by gxh6 and bxa6 (50.Nh6 gxh6 ... 100.Na6 bxa6)
Black's knights take White's queen and rooks (150...Nxd1 200...Nxh1 250...Nxa1) [1.halfmove lost from here]
White plays d3 and e3 (300.d3 350.e3)
White takes four Black pieces with pawns: hxg3, exf4, dxc4,axb3 (400.hxg3 450.exf4 500.dxc4 550.axb4)
White takes Black's other three pieces with bishops (600.Bxa8 650.Bxh8 700.Bxa8 after Qa8 for instance) [1. halfmove lost from here]
Black takes White's bishops : fxe6, cxd6 (749.Be6 fxe6 799.Bd6 cxd6).
White's pawns are on the b,c, f, g files, Black's are on the a, d, e and h files.
Black makes 44 pawn moves including eight promotions (849...a5 ... 3049...h1Q) [1. halfmove lost from here]
White takes Black's eight promoted pieces (3098...Qd4+ 3099.Kxd4 3149.Kxd4 3199.Kxd4 ... 3499.Kxd4)
White makes 42 pawn moves including eight promotions (3549.f5 ... 5599.g8Q) [1. halfmove lost from here]
Black takes White's eight promoted pieces (5648.Qd8+ Kxd8 5698.Qd8+ Kxd8 ... 5898.Qd8+ Kxd8 ½-½)
King vs King, game drawn immediately by FIDE law 1.3

But now that I have seen how this can be done, I am probably too lazy to produce the game. After all it is quite a lot of nitty-gritty.
Half the challenge was also to show it work in real life, my PGN file should be an indication of the feasibility of the method.

- Jesper Nørgaard Welen, 2010-11-02

7. http://ficsforum.110mb.com/ficsforum...c.php?f=14&t=8 has the right answer too and for the right basic reasons (though I haven't tried to verify or refute the suggested method) and is the only other one I can find in English that has it right without quoting my solution. It's amazing how many wrong answers there are; I've seen a few in books and they were all wrong too.

But consider the following rearrangement:

Black makes 44 pawn moves including eight promotions (2200 moves)
White takes Black's eight promoted pieces (400 moves)
White makes 42 pawn moves including eight promotions (2099.5 moves)
I'm confident my method is safe and that it would be quite easy to keep the promoted pieces in a bunch somewhere and thus avoid checks, but I agree that yours is more obviously safe and therefore better.

8. Originally Posted by Kevin Bonham
http://ficsforum.110mb.com/ficsforum...c.php?f=14&t=8 has the right answer too and for the right basic reasons (though I haven't tried to verify or refute the suggested method) and is the only other one I can find in English that has it right without quoting my solution.
That piece of analysis was utterly unconvincing to me. Especially the reasons why switching turns between white making draw-avoiding moves and black making them, seemed cloudy arguments to me. Your cookbook recipe was much easier to proof watertight.
Originally Posted by Kevin Bonham
I'm confident my method is safe and that it would be quite easy to keep the promoted pieces in a bunch somewhere and thus avoid checks, but I agree that yours is more obviously safe and therefore better.
I now realize I was obsessed with queening - but underpromoting for instance to knights this becomes child play. The only requirement is that the last piece is at least a rook, or else a "dead" position occurs before the last knight or last bishop is taken. So say 15 knights and a white rook, that would be pretty easy. The rook finishes the ball, eaten by the black king.

If insisting on queens the problem of lack of room to make maneuvers for wasting time beetween the 50-moves cycles seem to me to become quite big, since any little rotation of either king or queens will repeat the position too fast. If white could just move positively each move like 1.b4 and 2.b5 and 3.b6 then it would be rather straight-forward, but because both players need to shuffle too for 50 moves between each pawn push, I would really like to see a cookbook recipe on that!

By the way, is Babaschess really the only PGN reader that reads the damn game? I have tried Rybka, Keith Fuller pgn-reader, Chesspad, Chessdb, SCID, PGN Mentor, etc. etc. It occurs to me that it would be impossible in all of these programs to see the famous mate in 524 moves (7-piece position counting kings) in a PGN file, or even proprietary game file for that matter.

9. If you two ever meet OTB you should play it out.

10. Originally Posted by Boris
If you two ever meet OTB you should play it out.
With clocks and increment.

11. He, he, Boris. Good one. But I have my secret weapon, I will make Kevin dizzy with 15 knights on board - think of all the knight forks! It's not every day you can claim a family check to a king and 7 knights with one single move!

12. longest games, is every possible position taken to repetition 2 times?

13. Originally Posted by antichrist
longest games, is every possible position taken to repetition 2 times?
There isn't time to try that because the 50-move rule gets in the way. Without the 50-move rule the longest possible game would be incredibly long. You might have the same 118 phases but the number of moves in each phase would be enormous.

14. Originally Posted by Jesper Norgaard
It's not every day you can claim a family check to a king and 7 knights with one single move!
I've moved the task I came up with in response to this (find the greatest number of pieces you can place under attack with a single move) to its own thread here.

15. Originally Posted by Kevin Bonham
There isn't time to try that because the 50-move rule gets in the way. Without the 50-move rule the longest possible game would be incredibly long. You might have the same 118 phases but the number of moves in each phase would be enormous.
But where repetition is possible just short of incurring 50 move rule has it been applied?