3

There are positions where no matter what legal move black or white does during this sequence of moves, an specific player, will always be mated after X moves (sometimes earlier, but after X moves its a certain thing).

What is the position with biggest value for X?

Colleg an Delta
  • 131
  • 2
  • 3
    Could you add an example of such a position? – Brian McCutchon Sep 15 '21 at 21:25
  • 1
    Do you mean something like this position? https://chess.stackexchange.com/questions/4963/longest-sequence-of-mutually-forced-moves/4965#4965 – bof Sep 16 '21 at 00:37
  • The problemist's MatPlus Forum surely will know! This is a known problem theme, bof gave an example that is forced but doesn't lead to a mate. A simple example (forced mate) would be Kh6 Bh5 - Kh8 and now add a lot of queens of both color attacking g7. After Qg7+ a big brawl ensues, the last queen mates. – Hauke Reddmann Sep 16 '21 at 09:58
  • Just confirming you dont mean something like the following;

    http://kirill-kryukov.com/chess/longest-checkmates/longest-checkmates-sorted-by-length.shtml

     – Dheebs Sep 16 '21 at 15:44
  • To make easier to understand, my question basically means, how many moves ahead would the AI need to see to be able to make sure it always lose the game. If the ai checks X moves ahead, it will know that if its plan was to lose the game, its plan failed, because he WILL mate the other player. – Colleg an Delta Sep 18 '21 at 19:22
  • David Bateman, this is not what I want, at position q7/7k/8/6R1/8/8/2P2R2/K7 w - - 0 1 of your list as some example, there is a possibility of white wining and the possibility of black wining. White, the player to play, wont ALWAYS win. – Colleg an Delta Sep 18 '21 at 22:43
  • The position linked by bof is one example, The alexey khanyan, tim krabble answer is another example. – Colleg an Delta Sep 19 '21 at 03:21

2 Answers2

7

The longest known sequence is 22 plies long.

[Title "Alexey Khanyan, Tim Krabbe's Website Diary Entry #267 2008, Mate In 11"]
[FEN "4Q2Q/4r3/4n1n1/1bbK1krn/RR1RR1RR/2qn1R1n/4n1nN/Q3Q3 b - - 0 1"]
[startflipped ""]

1... Ng2f4+ 2. Rfxf4+ N2xf4+ 3. Rgxf4+ Nh3xf4+ 4. Rhxf4+ Ndxf4+ 5. Rxf4+ Nhxf4+ 6. Rxf4+ Ngxf4+ 7. Rxf4+ Nxf4+ 8. Rxf4+ Kxf4+ 9. Qee5+ Qxe5+ 10. Qaxe5+ Rgxe5+ 11. Qxe5+ Rxe5+ 12. Qxe5#

Source: Diary Entry #267

Rewan Demontay
  • 17,514
  • 4
  • 67
  • 113
-1

We don't know-. For instance, it's entirely possible the starting position is just mate in 10000.

However, the obvious answer if we modify the question slightly to the longest forced mate known can be found in an endgame tablebase (a collection of endgame positions, with evaluations and time to mate). The Syzygy tablebase is an example. The current record (ignoring the 50 move rule) is 546 with this position (as mentioned on Wikipedia). Most tablebases ignore the 50 move rule, and as such I cannot find any note on the longest forced mate with the 50 move rule.

Rewan Demontay
  • 17,514
  • 4
  • 67
  • 113
pulsar512b
  • 1,858
  • 4
  • 12
  • 3
    My understanding is that the question is about situations where neither player can prevent mate, even if both players cooperate to avoid it. In such a position, there is no legal way to draw. The starting position is not such a position. – Brian McCutchon Sep 15 '21 at 21:23
  • @Brian-McCutchon is right. If that number X was found, it would be possible to create the worst ai ever. Lets say the value of X is 14, by calculating 14+1 moves ahead, he will be able to know with 100% certainty if any of the moves it is will be selecting for this move can lead to a win at the future or not and the ai would avoid such move.

    The 546 move is not a valid answer because mate is not forced to happen, its possible to black mate white, depending of what white/black do.

     – Colleg an Delta Sep 15 '21 at 23:01
  • Ah my bad, I misread the question- apologies. – pulsar512b Sep 16 '21 at 01:49