Background: I’ve made a solver that works in-game and in realtime by taking a screenshot, converting it to a board, solving the board, and speaking the results. It has a mode for opponents and one for alone types (like capture as you have done). The solver works very quickly for opponent mode, even checking against the opponents next possible moves. However, the alone mode generates so many results with my method that it just takes FOREVER.

In case you prefer to see what I’ve done first, this is the general flow for alone mode:

*Generate all meaningful swaps for a given board
*Queue them up for processing
*Execute the boards as the game would
*Only for meaningful results:
–add them to a results tree
–queue them up for further processing (FILO to get depth first behavior)
*Repeat
*Analyze all the tree’s leaves to find the best path (it short-circuits if it finds an empty board)

Various optimizations are in there, but the biggest one is searching the existing results to make sure only unique results are added to the tree. With that optimization or without it, I’ve never seen my program end on one of the capture boards :) although it does work on a test board.

Thanks for checking it out! No, I wasn’t aware of Tower Puzzles, but rest assured that I’ll give it a try as soon as possible. In fact, it’s downloading right now :-)

Mark

Have you played Aurora Feint II: Tower Puzzles? It’s a whole game with just these kinds of puzzles that we launched earlier this week. I wonder if your solver will work on the last puzzle… ;)

Happy Holidays
- Jason