Play basketball, and coach a bot to be a superstar! 🤖🏀
First, Tell Us About You!
Select a Level to Start!
🎯
Level 1 One Spot, One Brain
Fixed position. Teach the bot one perfect shot!
1 state · 100 actions
📖
Level 2 New Spot, New Page
Ball moves. Bot needs a brain page per position!
10 states · 1,000 actions
📚
Level 3 Dodge the Blocker
Ball + blocker move. The bot's brain becomes a library!
80 states · 80,000 actions
Inside HoopsBot's Brain: Q-Table 🧠
👁️ State
What the bot can see
e.g. ball position, blocker height
🎯 Action
What the bot can choose
e.g. angle + power combo
The Q-Table stores a score for every State → Action pair. Higher = bot thinks that action is better.
📄 You're on Level 1
One grid. That's it. The ball never moves, so HoopsBot only needs a single 10×10 grid to remember every shot it's tried. What you see is the entire brain.
📖 You're on Level 2
The ball moves — so the brain needs pages. Each ball position gets its own grid. You're only seeing one page of the flipbook right now.
💡 Cells flip green→pink? Same shot, different position = different result. That's why separate pages exist.
📚 You're on Level 3
Ball + blocker = the brain becomes a library. One flipbook per blocker position, each with pages per ball position. You're seeing one page out of hundreds. 🤯
Reading the Grid 🎨
Each cell = one angle+power combo (Action).
Colour = how good the bot thinks it is:
Green: Positive Action✅ (Good Shot)
Pink: Negative Action❌ (Bad Shot)
Gray: No Idea🤷 (Not Trained Yet)
How It Learns 🧠
After every shot, the bot updates that cell using the Bellman equation: