# Is the result in mini-chess known with perfect play?

I do not know, if "mini-chess" is the official name of the game I am talking about.

In mini-chess, the players only start with the pawns (In particular, there are no kings!)

• The players play the moves alternate, as in the usual chess.

• If a player manages to get a pawn to the opponent's back rank, he immediately wins.

• If a player cannot move, he loses.

• En passant is like in the usual chess.

Is it known which player wins with perfect play ?

• FYI - this game is often called "Pawn Wars" (Andrew Soltis, Studying Chess Made Easy) – user1108 Mar 9 '15 at 13:40
• I believe that it is a draw, but I cannot prove to you this. It just seems intuitively true as well as true from experience. – CognisMantis Mar 10 '15 at 3:23
• The rules I described allow no draw. Every game has a winner. – Peter Mar 10 '15 at 19:27
• What about drawing by repetition? – magd Mar 17 '15 at 13:35
• How should a repitition be possible in this game ? – Peter Mar 18 '15 at 10:22

The game is also known under the name pawns-only chess. There is a variation of this game called breakthrough that has more strategical depth by allowing the pawns to also move diagonally.

Since I had already implemented breakthrough, it was very straightforward to add an implementation of pawns-only chess to my Stockfish fork for fairy chess variants (actually I only had to define the rules in a simple structure, see the commit). I have run a single-threaded search for ~10 hours using 2GB of RAM, which yielded a mate score for all first moves, so the results should be reliable, although they are no strict mathematical proof, because Stockfish does forward pruning in its tree search.

The results of Stockfish's search state that the moves 1.b4, 1.c4, 1.f4, and 1.g4 are winning for white, whereas all other first moves are a win for black with perfect play. See the input and output below for details. Since the search was single-threaded and depth-limited, the results should be reproducible.

While writing this answer, I found that someone else independently had obtained the same results, see his chess stackexchange answer and github repository.

input:

``````setoption name hash value 2048
setoption name multipv value 16
setoption name UCI_Variant value pawnsonly
position startpos
go depth 40
``````

output (truncated):

``````info depth 40 seldepth 38 multipv 1 score mate 18 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv b2b4 g7g5 c2c4 f7f5 d2d4 g5g4 e2e3 g4g3 f2g3 e7e6 g3g4 f5g4 e3e4 d7d6 d4d5 e6d5 e4d5 h7h6 a2a4 g4g3 h2g3 h6h5 b4b5 b7b6 g3g4 h5h4 g4g5 h4h3 g2h3 a7a5 b5a6 c7c6 a6a7 c6d5 a7a8
info depth 40 seldepth 36 multipv 2 score mate 18 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv g2g4 b7b5 f2f4 b5b4 e2e4 c7c5 d2d3 h7h6 a2a4 b4a3 b2a3 g7g6 h2h4 f7f6 f4f5 g6g5 h4h5 a7a6 c2c3 e7e5 f5e6 d7e6 d3d4 c5d4 c3d4 f6f5 e4f5 e6f5 g4f5 g5g4 f5f6 a6a5 f6f7 g4g3 f7f8
info depth 40 seldepth 40 multipv 3 score mate 20 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv f2f4 c7c5 e2e4 h7h5 c2c4 d7d6 g2g3 g7g6 f4f5 a7a6 a2a3 b7b5 b2b3 g6f5 e4f5 e7e6 c4b5 a6b5 f5e6 f7e6 h2h3 d6d5 g3g4 c5c4 b3c4 h5g4 h3g4 b5b4 a3b4 d5c4 g4g5 c4c3 d2c3 e6e5 g5g6 e5e4 g6g7 e4e3 g7g8
info depth 40 seldepth 40 multipv 4 score mate 20 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv c2c4 f7f5 d2d4 a7a5 f2f4 e7e6 b2b3 b7b6 c4c5 b6c5 d4c5 h7h6 h2h3 g7g5 g2g3 d7d6 c5d6 c7d6 f4g5 h6g5 a2a3 e6e5 b3b4 a5b4 a3b4 f5f4 g3f4 g5g4 h3g4 e5f4 b4b5 f4f3 e2f3 d6d5 b5b6 d5d4 b6b7 d4d3 b7b8
info depth 40 seldepth 39 multipv 5 score mate -19 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv b2b3 c7c5 c2c4 g7g5 a2a3 f7f5 b3b4 d7d6 h2h3 f5f4 g2g4 f4g3 f2g3 h7h5 g3g4 h5g4 h3g4 e7e6 e2e4 c5b4 a3b4 e6e5 d2d3 a7a6 c4c5 d6c5 b4c5 a6a5 c5c6 b7c6 d3d4 e5d4 e4e5 d4d3 e5e6 d3d2 e6e7 d2d1
info depth 40 seldepth 37 multipv 6 score mate -18 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv e2e3 b7b5 g2g4 e7e5 f2f4 e5f4 e3f4 c7c5 f4f5 d7d6 a2a3 d6d5 b2b3 h7h6 a3a4 b5b4 a4a5 a7a6 h2h3 g7g6 f5g6 f7g6 h3h4 c5c4 g4g5 h6h5 b3c4 d5c4 d2d3 c4d3 c2d3 b4b3 d3d4 b3b2 d4d5 b2b1
info depth 40 seldepth 37 multipv 7 score mate -18 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv a2a4 g7g5 f2f3 c7c5 a4a5 d7d5 e2e4 d5d4 e4e5 h7h5 g2g3 g5g4 f3f4 e7e6 f4f5 e6f5 d2d3 b7b5 a5b6 a7b6 c2c4 d4c3 b2c3 b6b5 d3d4 c5d4 c3d4 b5b4 e5e6 f7e6 d4d5 b4b3 d5e6 b3b2 h2h3 b2b1
info depth 40 seldepth 37 multipv 8 score mate -18 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv d2d3 g7g5 d3d4 b7b5 d4d5 a7a5 e2e3 h7h5 f2f4 g5f4 e3f4 h5h4 c2c3 a5a4 b2b3 a4b3 a2b3 e7e6 c3c4 b5b4 g2g4 h4g3 h2g3 f7f5 c4c5 e6d5 g3g4 f5g4 c5c6 d7c6 f4f5 g4g3 f5f6 g3g2 f6f7 g2g1
info depth 40 seldepth 37 multipv 9 score mate -18 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv h2h4 b7b5 g2g4 c7c5 c2c3 e7e5 f2f3 d7d6 d2d3 d6d5 e2e3 c5c4 d3d4 e5d4 e3d4 f7f5 g4f5 h7h5 f3f4 a7a6 b2b4 c4b3 a2b3 a6a5 f5f6 g7f6 c3c4 d5c4 b3c4 b5c4 f4f5 c4c3 d4d5 c3c2 d5d6 c2c1
info depth 40 seldepth 37 multipv 10 score mate -18 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv f2f3 g7g5 c2c4 f7f5 e2e3 e7e5 b2b4 f5f4 e3f4 e5f4 c4c5 d7d6 d2d4 d6c5 d4c5 c7c6 a2a3 a7a6 a3a4 h7h6 g2g4 f4g3 h2g3 h6h5 b4b5 h5h4 g3h4 g5h4 b5a6 b7a6 f3f4 h4h3 f4f5 h3h2 a4a5 h2h1
info depth 40 seldepth 37 multipv 11 score mate -18 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv g2g3 g7g5 g3g4 c7c5 a2a4 d7d5 a4a5 e7e6 b2b3 e6e5 c2c3 e5e4 d2d3 e4d3 e2d3 d5d4 c3d4 c5d4 b3b4 a7a6 h2h3 f7f6 f2f3 h7h6 b4b5 a6b5 h3h4 g5h4 a5a6 b7a6 f3f4 h4h3 f4f5 h3h2 g4g5 h6g5
info depth 40 seldepth 39 multipv 12 score mate -18 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv h2h3 b7b5 h3h4 f7f5 a2a3 c7c5 b2b3 d7d5 c2c3 c5c4 b3c4 d5c4 h4h5 e7e5 d2d3 c4d3 e2d3 a7a5 g2g3 a5a4 f2f3 h7h6 g3g4 f5f4 g4g5 h6g5 h5h6 g7h6 d3d4 e5e4 f3e4 f4f3 d4d5 f3f2 e4e5 f2f1
info depth 40 seldepth 37 multipv 13 score mate -18 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv d2d4 b7b5 h2h3 g7g5 e2e3 h7h5 e3e4 f7f6 d4d5 a7a5 c2c3 h5h4 b2b3 b5b4 c3c4 c7c6 d5c6 d7c6 c4c5 e7e6 g2g4 h4g3 f2g3 f6f5 e4f5 e6f5 h3h4 g5h4 g3h4 f5f4 h4h5 f4f3 h5h6 f3f2 a2a4 f2f1
info depth 40 seldepth 35 multipv 14 score mate -17 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv e2e4 g7g5 d2d4 b7b5 f2f3 h7h5 c2c3 f7f6 b2b3 e7e6 d4d5 e6d5 e4d5 f6f5 g2g3 g5g4 f3f4 a7a5 a2a3 a5a4 d5d6 c7d6 b3a4 b5a4 c3c4 d6d5 c4c5 d5d4 c5c6 d7c6 h2h3 g4h3 g3g4 h5g4
info depth 40 seldepth 35 multipv 15 score mate -17 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv c2c3 b7b5 e2e4 g7g5 d2d4 a7a5 f2f3 c7c6 g2g3 d7d6 d4d5 c6d5 e4d5 e7e6 h2h4 g5h4 d5e6 f7e6 g3h4 h7h5 b2b3 b5b4 c3b4 a5b4 f3f4 d6d5 f4f5 e6f5 a2a3 b4a3 b3b4 a3a2 b4b5 a2a1
info depth 40 seldepth 35 multipv 16 score mate -17 nodes 52290435219 nps 1359031 hashfull 1000 tbhits 0 time 38476253 pv a2a3 b7b5 e2e4 g7g5 d2d3 a7a5 d3d4 h7h5 f2f3 f7f6 e4e5 f6e5 d4e5 a5a4 g2g3 e7e6 h2h3 h5h4 g3h4 g5h4 b2b4 a4b3 c2b3 c7c5 a3a4 b5a4 b3a4 c5c4 a4a5 c4c3 a5a6 c3c2 f3f4 c2c1
``````

The game you describe reminds me of Dawson's chess where the pawns are just one step apart and no en passant is allowed.

This game is instensively analysed and the winner depends in a non-trivial way on the length of the pawn chain!

I suspect that Black should win.

A strategy for Black is to repeat moves and keep symmetry. Since W has no breakthrough idea, W's best bet is to keep sacrificing pawns. Black could grab the first W pawn and then when W sacs a second one, offer B's fifth rank pawn as a counter-sac - using the idea that this pawn can reach W's back-rank faster. B recaptures W's second pawn only when W captures this pawn. After W sacs enough pawns, B can queen first.

• I also suspect that black wins, but keeping symmetry doesn't work. White doesn't have to sacrifice pawns, he can prepare pawn levers like 1.e4 e5 2.g3 g6 3.f4 now 3…f5 would lose for black and 3…ef 4.gf is also problematic because white will play e5 and freeze either the d- or the f-pawn. – BlindKungFuMaster Mar 21 '15 at 22:35
• Start analysing with fewer pawns: One pawn each: black wins because of lock-up. Two pawns: black wins with break-through. Three pawns? – jk - Reinstate Monica Mar 27 '15 at 10:52