🏏 Captain Cool OS
Captain Cool OS is an AI-powered tactical decision engine for cricket captains, built using Google Gemini and a multi-agent orchestration system.
The idea was simple:
What if an AI could think through high-pressure IPL moments like a real captain?
Not just predict outcomes.
Actually debate tactics.
The Idea
Cricket is one of the few sports where a captain can completely change the game with a single decision.
Who bowls the next over?
Do you trust pace in heavy dew?
Do you go defensive or attack for wickets?
Do you protect square boundaries or long-off?
Do you stick with the matchup or adapt to pressure?
In real matches, these decisions happen ball-by-ball under insane pressure.
We wanted to build a system that could simulate that kind of tactical thinking.
What We Built
Captain Cool OS is a multi-agent AI system that behaves like a cricket team’s tactical war room.
A user provides the current match situation:
- score
- wickets
- venue
- dew conditions
- batter matchups
- bowlers remaining
- death-over pressure
The system then runs an internal debate between specialized AI agents before locking in a tactical decision.
Instead of generating one generic response, the agents challenge each other, identify weaknesses, revise plans, and finally arrive at a tactical consensus.
The Multi-Agent System
We built multiple Gemini-powered agents, each with a clear responsibility.
📊 Stats Analyst
Looks at:
- venue trends
- bowler vs batter matchups
- recent form
- pressure metrics
- death-over effectiveness
This agent acts like the analytics department.
🧠 Strategist
Creates the initial tactical plan.
This includes:
- who bowls next
- line and length
- field setup
- contingency plans
- execution strategy
⚠️ Devil’s Advocate
This is where things get interesting.
Instead of agreeing blindly, this agent actively attacks the proposed plan:
- identifies tactical risks
- exposes weak execution points
- challenges matchup assumptions
- questions field placements
This creates a real reasoning loop instead of a single-pass AI answer.
🧩 Reflection Evaluator
Acts like a head coach reviewing the plan.
It checks:
- contradictions
- execution feasibility
- tactical instability
- risk exposure
If the strategy looks flawed, the system forces a revision loop.
🎙 Narrative Translator
Converts structured tactical reasoning into natural cricket language.
The final output feels closer to commentary or dugout strategy discussion rather than raw AI output.
Tactical Orchestration Flow
The orchestration pipeline works like this:
- Stats Analyst gathers evidence
- Strategist proposes a tactical plan
- Devil’s Advocate stress-tests it
- Reflection Evaluator checks for contradictions
- Strategist revises the plan if needed
- Final tactical consensus is generated
The system doesn’t just answer.
It debates before deciding.
Tactical Visualization Layer
One of the most fun parts of the project was building the cinematic tactical interface.
We added:
- confidence evolution tracking
- tactical instability alerts
- live orchestration timelines
- bowling trajectory targeting
- field placement radar visualizations
The goal was to make the reasoning process feel visible and understandable instead of hidden behind a chatbot response.
Why We Think This Is Interesting
Most sports AI tools focus only on predictions.
Captain Cool OS focuses on decision-making under pressure.
We wanted the system to:
- explain itself
- challenge itself
- adapt its reasoning
- communicate like a cricket strategist
The debate between agents became the most interesting part of the project.
Tech Stack
- Google Gemini 2.5 Flash
- Google Antigravity
- FastAPI
- Next.js
- TypeScript
- TailwindCSS
- WebSockets
- Custom SVG tactical visualizations
Future Scope
There’s a lot we’d love to explore further:
- live Cricbuzz/ESPN integration
- ball-by-ball adaptive tactics
- live win probability shifts
- voice-enabled captain interaction
- reinforcement learning memory
- multimodal pitch analysis
