Author Topic: Kyoto Shogi: theoretical win for sente?  (Read 14 times)

HGMuller

  • Global Moderator
  • Full Member
  • *****
  • Posts: 120
    • View Profile
Kyoto Shogi: theoretical win for sente?
« on: July 30, 2018, 10:42:48 am »
Kyoto Shogi is a very peculiar chess variant, because the pieces change identity every time you move them. E.g. a Rook turns into a Pawn, and when the Pawn moves it turns again back into a Rook. Similarly a Gold General toggles to Knight, a Silver General to Bishop, and a Tokin (which also moves as Gold) to Lance. Only the King always remains itself. A weird property of the game is that pieces can get stuck (i.e. left without moves) when they advance too far, and that for some pieces this cannot be avoided: in Shogi a Knight always moves two ranks forward, and the Gold it morphs to then can at most move one rank back. A Tokin on last rank can stay alive arbitrarily may moves, but only if it moves back an forth (the latter as Lance) betwee two squares, which also isn't very useful.

For this reason the value of pieces is very much dependent on their location on the board. A Knight on the forelast rank (which has no moves) might even have negative value, as the enemy King can shelter behind it, and it would bind one of your other pieces to keep it protected against being taken by that King (which would give him the piece in hand).

I taught my engine CrazyWa (which was really designed to play Wa Shogi on an 11x11 board with 31 piece types) to play this game. I used hand-picked piece-value guestimates, so there is no guarantee its evaluation is very accurate. But when I use it to analyze the start position, it sees a big advantage for the sente player. When I analyze the positions to which the suggested moves lead, this advantage grows even more. This is indicative of a won position.

For CrazyWa the value of a piece in hand is slightly over 3. (It doesn't matter very much which piece; in hand they are about equally useful.) I could easily find lines of play for sente against any possible defense that would lead to a position that would analyze as above +6. Now I am working on demonstrating that from each of these positions an analysis score above +9 can be forced against any defense. Even if the scores given by CrazyWa are not very accurate, +9 must be way beyon the threashol for a certain win.

Social Buttons

Like Like x 2 View List