kaiyotech

I make bots to play games using reinforcement learning.

to watch training
for videos

[kaiyotech @kaiyotech]

Nice article about rlgym and Ripple and Ragnarok and sorta Opti.

https://t.co/HYtKpjPsQ4

[RLGym @RLGymOfficial]

A new foe has appeared!

Ripple is a new bot from Rolv that uses a two-stage approach: first learning to imitate human gameplay, then playing against itself to improve🦾

[kaiyotech @kaiyotech]

You can follow the training 24/7 on my twitch. Currently working on air dribble/reset/double taps for a bit. Full run of the hierarchy maybe next week.

[kaiyotech @kaiyotech]

Opti bot working on that clean movement.

[kaiyotech @kaiyotech]

@RLGymOfficial Opti is a ton of fun to make because of things like this. Cheers for the tweet.

[kaiyotech @kaiyotech]

@A1Mato @RLGymOfficial Different bot. This is Opti, not Nexto, and even if rlgym was improving Nexto (they're not) they wouldn't release it, just like I'm not releasing Opti.

[kaiyotech @kaiyotech]

@6haunRL @RLGymOfficial Check my YouTube. This is just for movement. It's not the full bot.

[kaiyotech @kaiyotech]

@RLGymOfficial Folks can see other clips from Opti here.

https://t.co/KaROGVn8dj

[RLGym @RLGymOfficial]

🌊🌊🌊🌊🌊

[kaiyotech @kaiyotech]

@xiao_ted There are a number of interesting bots in development using the same rlgym framework, obviously they'll stay unreleased though.

My project is Opti, an options hierarchical model with many mechanic specific tasks. You can find it training 24/7 at https://t.co/reiQnvNhkq