summaryrefslogtreecommitdiff
path: root/src/strategy/monte_carlo.rs
AgeCommit message (Collapse)Author
2018-08-12Fixed logic error on when nothing should be avoidedJustin Worthe
2018-08-12Eliminated assuming opponents might do nothing in random movesJustin Worthe
2018-08-12Reduced more duplication and removed TODOsJustin Worthe
2018-08-12Removed unused importJustin Worthe
2018-08-12Folded duplicate code for player and opponentJustin Worthe
2018-08-12Collapsed player info into the rest of the 'buildings' objectJustin Worthe
2018-08-12Bumped energy cutoffs right up to allow iron curtainJustin Worthe
2018-08-12Test cases that do the iron curtainJustin Worthe
2018-08-12Allowed monte carlo search to use iron curtainsJustin Worthe
2018-08-11Clippy-suggested editsJustin Worthe
2018-08-09Rearranged random move selection to avoid expensive divisionJustin Worthe
2018-08-09Removed unnecessary Player fieldJustin Worthe
2018-08-09Removed dynamic settingsJustin Worthe
It worked really well for round 2 to set constants
2018-08-09Added more TODO ideasJustin Worthe
2018-07-24Fixed min and benchmark logging in discarding searchJustin Worthe
2018-07-24Modified pruning to be friendlier to calculating benchmarksJustin Worthe
2018-07-23Added option for discarding poor performing moves earlyJustin Worthe
2018-07-02Initial drop in replacement implementationJustin Worthe
It's faster than the other one! Doesn't tesla yet, but still! Yay!
2018-07-01Set up (failing) property test for working of new game stateJustin Worthe
Unfortunately, for this test to work, I still need to implement the function that reads the game state and goes from one to the other.
2018-06-30Initial stab at putting game engine behind a traitJustin Worthe
2018-06-30Removed deconstruction from the list of commands consideredJustin Worthe
I don't think deconstruction is a necessary part of a viable strategy for now.
2018-06-25Added rule on maximum two tesla towersJustin Worthe
2018-06-25Updated test cases and made engine work correctly according to testsJustin Worthe
I don't think a tesla appeared in this match. I need to contrive a bot to build one.
2018-06-25Compilation, allowing new moves to be chosen, and missile move orderJustin Worthe
2018-06-09Added pruning of buying energy buildings behind a feature flagJustin Worthe
2018-06-02Moved replay-based test to have convenience import from game engine replayJustin Worthe
2018-05-31Clippy suggested changesJustin Worthe
2018-05-19Inlined unnecessary mapJustin Worthe
2018-05-19Removed dependency on OS randomness for seedingJustin Worthe
I'm convinced that the cloud servers they're running on will be out of entropy, especially towards the end of the tournament. Rather have a less random initial seed than take that blocking performance hit.
2018-05-16Moved away from special benchmarking suiteJustin Worthe
Just using normal monte carlo. More iterations -> better.
2018-05-15Rayon for threadingJustin Worthe
2018-05-14Added running total of unoccupied cellsJustin Worthe
2018-05-14Changed invalid move checking to be a debug assertionJustin Worthe
2018-05-14Reduced number of needless allocations to improve perfJustin Worthe
Current iterations: 26486 in 10 seconds
2018-05-14Added profiling target with perfJustin Worthe
2018-05-12Limited bot to run within the 2 second windowJustin Worthe
2018-05-12Debugged and fixed the errors that had the monte carlo not workingJustin Worthe
Monte carlo now beats sample bot, if given plenty of time. I still need to put the max time tracking and enforcement in.
2018-05-12Initial stab at monte carlo implementationJustin Worthe
Doesn't seem to be working quite right... just sits there accumulating energy.