[Retros] Two-pawns chess

[Francois Labelle]

My computer finished looking for a win in two-pawns chess:
- White cannot win in 7 moves against perfect play.
- White can win in 8 moves with 1.e3.

At depth 8 I only tested 1.e3 which looked the most promising. I don't 
know if other opening moves can win in 8 too.

Below are optimal moves against all 20 black defenses. If multiple moves 
work for White, I pick the fastest win, and if there's a tie, I pick the 
first piece that works, from Ra1 to Rh1.

1. e3 a6 2. Bxa6 (white wins at move 4)
1. e3 a5 2. Qf3 (white wins at move 7)
1. e3 b6 2. Qh5 (white wins at move 6)
1. e3 b5 2. Bxb5 (white wins at move 4)
1. e3 c6 2. Bc4 (white wins at move 7)
1. e3 c5 2. Qh5 (white wins at move 6)
1. e3 d6 2. Nc3 (white wins at move 8)
1. e3 d5 2. Qh5 (white wins at move 5)
1. e3 e6 2. Qh5 (white wins at move 8)
1. e3 e5 2. Qg4 (white wins at move 7)
1. e3 f6 2. Qh5+ (white wins at move 5)
1. e3 f5 2. Qh5+ (white wins at move 5)
1. e3 g6 2. Qf3 (white wins at move 6)
1. e3 g5 2. Qg4 (white wins at move 5)
1. e3 h6 2. Qh5 (white wins at move 6)
1. e3 h5 2. Nc3 (white wins at move 6)
1. e3 Na6 2. Bc4 (white wins at move 7)
1. e3 Nc6 2. Bc4 (white wins at move 7)
1. e3 Nf6 2. Nc3 (white wins at move 8)
1. e3 Nh6 2. Qh5 (white wins at move 6)

- Francois

On 09/15/2012 09:02 AM, Francois Labelle wrote:

> By the way I've just programmed the following interpretation of Eric's 

> idea:

>  - players win by either checkmating or by capturing N pawns,

>  - moves must be orthodox-legal, i.e. players can't capture the Nth 

> pawn by putting (or leaving) their own king in check,

>  - promoting a pawn makes it no longer a target.

>

> For N=1, White wins in 3 moves and 1.c3 and 1.c4 are the only winning 

> moves (congrats Pascal!)

> For N=2, all I know is that White has no win in 6 moves. I'll see if I 

> can analyze more moves.

>

> My program is slow because my chess engine was not designed for 

> directmate (it has no transposition table). If someone is interested 

> in a computer analysis, it would probably be better to add the 

> condition to Popeye and run that.

>

>     Francois

>

> _______________________________________________

> Retros mailing list

> Retros at janko.at

> http://www.pairlist.net/mailman/listinfo/retros