Monday, September 30, 2013

Chess960: SP395 beats SP491 in the competition for the most number of viable starts

Turns out that SP395 beats SP491 in the competition to find the highest number of viable moves *here*. It beats standard chess in variety and also only takes four moves to castle queen-side like standard chess.

SP395: Sixteen viable first moves!?

Depth 24 search using Houdini-3 with a variation of only +/-0.15 for:
  1. a3 - yes because it develops a queen
  2. a4 - yes because it develops a queen, threatens a5 and controls b5
  3. b3 - yes because it develops a queen and Nc3 shields the queen
  4. b4 - yes because b4 is supported by Be1xb4 lining up black's king
  5. c3 - yes because it shields the queen and enables d4
  6. c4 - yes because it controls d5 and threatens c5
  7. d3 - yes because it develops a bishop
  8. d4 - yes because it develops a bishop and threatens d5
  9. e3 - yes because it enables developing Nc1-e2
  10. e4 - yes because it allows Ne3 and controls d5
  11. f4 - yes because it develops a bishop and could support a pawn on d5
  12. g3 - yes because it develops a bishop
  13. g4 - yes because it develops a bishop (can white castle queenside!?)
  14. Nb3 - yes because if 1...a5 white has a4/d3 attacking with development
  15. Nc3 - yes because it is flexible and rapidly develops for sake of possible O-O-O
  16. Ne3 - yes because it rapidly develops and controls d5

Rejected moves:
  1. f3?! - no because it traps the bishop and weakens the king
  2. Nd3?! - no because it over commits to a congested position
  3. h3?! - no because it achieves too little
  4. h4?! - no because it achieves too little
  5. O-O?! - no because it makes little sense to reveal castling intentions so soon
I'm starting to think that the positions with queen in the corner and a bishop in the corner produce such massive first move choices because there is no compulsion for white to play out pawns into the center when it might be wise to keep the long diagonals open. Also, another possible reason is that both rooks are on the b-g files which supports pawns that could be useful all the way up to the 5th rank.

1 comment:

  1. I don't know if I agree with your conclusions, Harry. What you might be seeing is further proof that engines are weak in the early opening, especially when there are no tactics to consider. I doubt that many (any?) engines have the generic evaluation functions needed for chess960, and I suspect they mainly on the same opening heuristics used in SP518.

    I propose we play a two-game match with one of these SPs that you've flagged, where we each take White in one game. In the game where you are White, I will select your first move from your list of viable moves. In the game where I am White, I will choose my first move myself. I will even give you draw odds for the match, i.e. if it finishes 1-1, you will have won the match. Do you really have so much confidence in all of the moves you list? - Mark