Skip to content

Releases: Stermere/Checkers-Engine

v1.2.0

08 Sep 20:26
Compare
Choose a tag to compare

Since the last release, a lot has changed I'll go over the highlights. First, the engine no longer uses GB's of RAM to store the game tree and now grabs the best move from the transposition table. Secondly, the nodes per second have been doubled and many odd bugs with the transposition table have been squashed. Lastly and most excitingly there is now a web app you can visit to play the engine without the need to download anything! This limits the most challenging difficulty to just a half second of thinking time but that should be enough for all but the most experienced players. If you'd like to play it visit https://ckees1.pythonanywhere.com (it will take a second to load if it hasn't been visited in a while). In the future, you will be able to run the web app locally and set the thinking time to whatever you want.

1.0.2

28 Aug 19:34
Compare
Choose a tag to compare

Added search extensions to increase the depth and modified the evaluation algorithm to temp the engine to play better in the opening.
The new command line args are: player_type search_time max_depth.

v1.0.1

13 Jun 18:57
Compare
Choose a tag to compare

To set up the program download Main.exe, and the difficulty you want (make sure they are in the same file together). To play against the bot run the .bat file corresponding to the difficulty you want to play against. To anyone trying to beat it on expert mode good luck!

v1.0.0

12 Jun 22:27
Compare
Choose a tag to compare

To set up the program download neural_net, Main.exe, and the difficulty you want (make sure they are in the same file together). To play against the bot run the .bat file corresponding to the difficulty you want to play against. The bot uses a neural network trained on 500 games of self-play with a much simpler evaluator. To anyone trying to beat it on expert mode good luck! you should be able to tie it easily it plays very passive on expert. The static evaluator is still stronger since the neural net is not trained on enough data and or it is overfitting. Try out version 1.0.1 for that engine version.

v1.1.1

31 Dec 21:37
Compare
Choose a tag to compare

This update brings much better branching control when searching helping it not miss tactical sequences. The command line args are player_type, search_time, and max_depth. where player type is 'human' or 'bot', search_time is the number of seconds to allow a search, and max_depth is an integer denoting the furthest a search can look into the future.

Unrelated to the strength of the engine it now has a name! Marcher