v1.0.1 is released. Includes a much improved AI that will give you a serious challenge. Suggest a fresh install is used to ensure the new AI is installed correctly.
Artificial Intelligence details
This game uses genuine machine learning AI. It will study how you play, build a statistical model and learn to counter your moves. This AI uses QTables to understand which move it should play given the current set of circumstances. v1.0.1 brings you a few things:
1) small bug fixes and defect remediation where the AI wasn't learning as fast has it could have.
2) the introduction of a second Qtable - or 2nd AI that can look ahead one move and try to outwit the player.
3) the primary AI for Jungeong now comes pre-loaded with a library of valid moves she has learned during play-testing.
v1 doesn't have a look-ahead function. It simply tried it's best based on it's current position. v1.0.1 now has a look ahead component.
The first AI, the "current move" AI, has priority and if it determines it can strike and win right now then it will take that option. if the "current move" AI can't determine a move that is appealing, then it defers to the "move-ahead" AI. This 2nd AI's responsibility is to assume it can't win this round, but can try to move to a better position for next round.
The only bot that uses AI is Junjeong. Play her for ten rounds and watch her noticeably improve. Play her twenty times and you may find yourself on a losing streak.
The bot called Kim Won-sul hasn't changed. He plays completely random cards and is a good bench mark. He will often try to play illegal cards (and fail) and can regularly beaten.
Kim Yu-sin will play random cards but only legal cards. This means he is much more effective as every play is a chance for you to lose.
Junjeong is the machine learning bot. Play her enough times and she'll give you a serious challenge.
This card game has gathered a lot of attention over the last 8 months and I wanted to refresh the AI for those that would like to revisit it. Thanks.
I've had a lot of fun defeating all three of the Hwarangs. I've downloaded and played the game on purpose, because I will implement a similar system in my text based game and I wanted to see how you have done it. Mine might be a little more complicated, but in essence they are very close to each other. The only thing that I couldn't agree with was the AI generated art, but that is a personal preference. Overall you have quite a fun game and you should try to increase its complexity.
Thanks for trying the game and finding time to comment. The beauty is in the simple short and sharp battles but I can see how some might be looking for more complexity. I have zero artistic skills so that's why the images are what they are. Hope your own vision works out for you. If you care enough, you can send your qtable.dat file to me via togfox @ gmail.com or discord. I can then use your fight style to train the AI. Cheers.
← Return to game
Comments
Log in with itch.io to leave a comment.
v1.0.1 is released.
Includes a much improved AI that will give you a serious challenge. Suggest a fresh install is used to ensure the new AI is installed correctly.
Artificial Intelligence details
This game uses genuine machine learning AI. It will study how you play, build a statistical model and learn to counter your moves. This AI uses QTables to understand which move it should play given the current set of circumstances. v1.0.1 brings you a few things:
1) small bug fixes and defect remediation where the AI wasn't learning as fast has it could have.
2) the introduction of a second Qtable - or 2nd AI that can look ahead one move and try to outwit the player.
3) the primary AI for Jungeong now comes pre-loaded with a library of valid moves she has learned during play-testing.
v1 doesn't have a look-ahead function. It simply tried it's best based on it's current position. v1.0.1 now has a look ahead component.
The first AI, the "current move" AI, has priority and if it determines it can strike and win right now then it will take that option. if the "current move" AI can't determine a move that is appealing, then it defers to the "move-ahead" AI. This 2nd AI's responsibility is to assume it can't win this round, but can try to move to a better position for next round.
The only bot that uses AI is Junjeong. Play her for ten rounds and watch her noticeably improve. Play her twenty times and you may find yourself on a losing streak.
The bot called Kim Won-sul hasn't changed. He plays completely random cards and is a good bench mark. He will often try to play illegal cards (and fail) and can regularly beaten.
Kim Yu-sin will play random cards but only legal cards. This means he is much more effective as every play is a chance for you to lose.
Junjeong is the machine learning bot. Play her enough times and she'll give you a serious challenge.
This card game has gathered a lot of attention over the last 8 months and I wanted to refresh the AI for those that would like to revisit it. Thanks.
I've had a lot of fun defeating all three of the Hwarangs. I've downloaded and played the game on purpose, because I will implement a similar system in my text based game and I wanted to see how you have done it. Mine might be a little more complicated, but in essence they are very close to each other. The only thing that I couldn't agree with was the AI generated art, but that is a personal preference. Overall you have quite a fun game and you should try to increase its complexity.
Thanks for trying the game and finding time to comment. The beauty is in the simple short and sharp battles but I can see how some might be looking for more complexity. I have zero artistic skills so that's why the images are what they are. Hope your own vision works out for you. If you care enough, you can send your qtable.dat file to me via togfox @ gmail.com or discord. I can then use your fight style to train the AI. Cheers.