search

SanchoGGP / ggp-base

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

Failure to efficiently solve Hidato-37 #361

Open arr28 opened 9 years ago

arr28 commented 9 years ago

We don't engage the partitioning logic, so we fail to solve this game. I strongly suspect that if we did engage the partitioning logic, we'd ace it.

arr28 commented 8 years ago

Following the fix for #372, we do now engage the partitioning logic for the Hidato variants. We do ace the 19 cell version. To say that we ace the 37 cell version is stretching it, but given reasonable clocks (e.g. 60/30) we do solve it.

Hidato is a slightly interesting one in that we'd really like to do the partitions both ways round - i.e. simultaneously ask...

As written, we only ask the former question.

I don't know how the analysis would need to change to form partitions that embody questions of both forms. However, having stared at the logic that uses the analysis results, I think it could cope with simultaneously having partitions of both kinds. (We already know that inputs can appear in more than one partition because the Futoshiki "quit" move appears in all partitions.)

I'm going to leave this one open, but downgrade it since we can solve both the existing puzzles. Oh, and I also need to get the puzzles up on Tiltyard.

arr28 commented 8 years ago

Visualization written and sent to Sam to publish to Tiltyard.

arr28 commented 8 years ago

Thinking about this a little more, I suspect we possibly solve it due to pair latches rather than partitioning. And I've seen that sometimes it takes more like 150s to solve, which we might well not get. Would definitely be interesting to get the partitioning to split the partitions the other way round.