Guess which stone will make the win!

Look at http://www.littlegolem.net/jsp/game/game.jsp?gid=1302803&nmove=12

Can you imagine, which stone will make the winning move?

Suprise! Suprise!

A surprising development of the game :-) Such things happen from time to time in this unpredictable game.

[game;id:1302803;move:12]

Assuming I've typed the grid in correctly and interpreted the output correctly, RoRoRo slightly disagrees with 17.1/d4-d3, prefering d4-e3 at every evaluation depth apart from 2-ply.

that was a neat quarterback sneak!

@FatPhil: Can you give a sequence of moves, where it makes any difference? I see 13% in both cases(*). The chances after the next seems to be equal(*) too (65% for a 1, 85% for a 2, 93% for 3-6).

(*) The estimated winning rates (in percent) are rounded to integers.

I can't give percentages, as RoRoRo evaluates things on an abstract scale from -12 (red win) to +12 (blue win). Here's some raw data

Initial position:

phil@geespaz:EinStein$ ./esbot.g5 -8 r 1 FO@R@@ QN@@@@

moveRedEval(FO@R@@,QN@@@@,8) QK = 8.93184

moveRedEval(FO@R@@,QN@@@@,8) QP = 9.45995

moveRedEval(FO@R@@,QN@@@@,8) QL = 8.81970

Selected=QL Score=8.81970

Taking RoRoRo's advice:

phil@geespaz:EinStein$ ./esbot.g5 -8 b 1 FO@R@@ LN@@@@

moveBlueEval(FO@R@@,LN@@@@,8) FL = 0.00000

moveBlueEval(FO@R@@,LN@@@@,8) FG = 2.44444

moveBlueEval(FO@R@@,LN@@@@,8) FK = 2.40586

Selected=FG Score=2.44444

phil@geespaz:EinStein$ ./esbot.g5 -8 b 2 FO@R@@ LN@@@@

moveBlueEval(FO@R@@,LN@@@@,8) OT = 8.53755

Selected=OT Score=8.53755

phil@geespaz:EinStein$ ./esbot.g5 -8 b 4 FO@R@@ LN@@@@

moveBlueEval(FO@R@@,LN@@@@,8) RX = 10.48405

moveBlueEval(FO@R@@,LN@@@@,8) RS = 10.48405

moveBlueEval(FO@R@@,LN@@@@,8) RW = 2.71142

Selected=RX Score=10.48405

Ignoring RoRoRo:

phil@geespaz:EinStein$ ./esbot.g5 -8 b 1 FO@R@@ KN@@@@

moveBlueEval(FO@R@@,KN@@@@,8) FL = 2.83025

moveBlueEval(FO@R@@,KN@@@@,8) FG = 2.44444

moveBlueEval(FO@R@@,KN@@@@,8) FK = 0.00000

Selected=FL Score=2.83025

phil@geespaz:EinStein$ ./esbot.g5 -8 b 2 FO@R@@ KN@@@@

moveBlueEval(FO@R@@,KN@@@@,8) OT = 8.76286

Selected=OT Score=8.76286

phil@geespaz:EinStein$ ./esbot.g5 -8 b 4 FO@R@@ KN@@@@

moveBlueEval(FO@R@@,KN@@@@,8) RX = 10.49949

moveBlueEval(FO@R@@,KN@@@@,8) RS = 10.49949

moveBlueEval(FO@R@@,KN@@@@,8) RW = 3.29321

Selected=RX Score=10.49949

The slight disagreement seems to be of the order of less than .5%, so rounding to integers will lose a lot of information.

Naively linearly mapping -12 to 0, and +12 to 100% probability of blue winning:

roll hanfried : RoRoRo

1 . . . 61.79% : 60.19%

2 . . . 86.51% : 85.57%

3-6 . . 93.75% : 93.68%

avg . . 87.22% : 86.75%

So the biggest deviation between RoRoRo is shown when blue rolls 1.

NOTE: I think I swapped the co-ordinates and the colours over as I typed the grid in upside down, and it was easier to just switch colours than work out what the grid was again. So you probably need to interpret Y..U as the top row, E..A as the bottom row, red as blue, and blue as red.

I wish I'd remembered about that before doing the sims again. It might be best to just start with the non-flipped grid

@FatPhil: If I understand your notation correctly, _you_ see a stone on position F, but there is no stone. There is just a stone on position G.