search

SanchoGGP / ggp-base

The General Game Playing Base Package
8 stars 4 forks source link

Suspect king pinning in Checkers #13

Open arr28 opened 10 years ago

arr28 commented 10 years ago

Move 55 in http://www.ggp.org/view/tiltyard/matches/f2a734a9519a4ab089e04463d2f711728b4ea4e6/ appears suspect. Sancho effectively pinned both his kings despite their being other moves which don't appear terrible.

Although Sancho went on to win, his evaluation dropped significantly after this move. (At this point, Sancho was still 10s of moves from determining a forced win, so it wasn't a case of "it doesn't matter which of these moves I do, I'm going to win anyway".)

SteveDraper commented 10 years ago

Agreed. I suspect a general mobility heuristic (assessed in metagaming with correlation tests as we do for piece heuristics) is the answer here, but that position would be a good one to extract into the test case library.

On Wed, Mar 12, 2014 at 8:16 AM, Andrew Rose notifications@github.comwrote:

Move 55 in http://www.ggp.org/view/tiltyard/matches/f2a734a9519a4ab089e04463d2f711728b4ea4e6/appears suspect. Sancho effectively pinned both his kings.

Reply to this email directly or view it on GitHubhttps://github.com/SteveDraper/ggp-base/issues/13 .

arr28 commented 10 years ago

Already extracted into library. (I used this as my test case for the tool to generate test cases from the Tiltyard.)

On 12 March 2014 15:16, SteveDraper notifications@github.com wrote:

Agreed. I suspect a general mobility heuristic (assessed in metagaming with correlation tests as we do for piece heuristics) is the answer here, but that position would be a good one to extract into the test case library.

On Wed, Mar 12, 2014 at 8:16 AM, Andrew Rose <notifications@github.com

wrote:

Move 55 in

http://www.ggp.org/view/tiltyard/matches/f2a734a9519a4ab089e04463d2f711728b4ea4e6/appearssuspect. Sancho effectively pinned both his kings.

Reply to this email directly or view it on GitHub< https://github.com/SteveDraper/ggp-base/issues/13> .

Reply to this email directly or view it on GitHubhttps://github.com/SteveDraper/ggp-base/issues/13#issuecomment-37420301 .

arr28 commented 10 years ago

With 1.54, fails the testcase ~50% of the time, moving King E8 -> D7 otherwise (which seems much better to me).

arr28 commented 9 years ago

With 1.60w, on snap-ci, we're still failing this case fairly often with the same move we made in the actual game.