IVEE NBA — AI Prediction Market Automation

AI-powered NBA playoffs prediction market automation dashboard — built on Canon CLI for the DEGA NBA Playoffs Prediction Market Hackathon.

Live Demo: https://ivee-apps.vercel.app
Demo Script: DEMO.md
Hackathon: DEGA NBA Playoffs Prediction Market Hackathon

---

What It Does

IVEE NBA is a real-time intelligence dashboard that automates trading strategies on NBA playoff prediction markets (Polymarket). It integrates Canon CLI's 4-agent AI pipeline — market analyst, strategy architect, developer, and QA — to scan opportunities, generate trade signals, run risk checks, and log every decision to .canon/execution/.

Strategy focus: Arbitrage Detection across NBA championship, conference, and player award markets.

---

Features

6-Section Dashboard

Section	Panels	Data Source
Live	Games, Standings, Injuries, Schedule	balldontlie.io API
Markets	Markets, Arbitrage, Momentum, Speed Opps, Fear & Greed	Polymarket Gamma API
Analysis	Teams, Bracket, Cross-Market	balldontlie + Polymarket
Strategy	Strategy Dashboard, P&L Tracker	DEGA Rank + Canon Bridge
Automation	AI Agent Pipeline (4 strategies)	AutomationEngine + Canon
Logs	Execution Logs (JSONL)	localStorage / .canon/execution/

AI Trading Pipeline (Canon Architecture)

4-agent workflow running on every automation cycle:

market-analyst → strategy-architect → developer → qa
      ↓                ↓                  ↓          ↓
  Scan NBA         Design entry        Validate     Pre-flight
  markets          approach            signal       checks
      └──────────────────┴──────────────────┴──────────┘
                              ↓
                    TradeSignal (RiskInterface checked)
                              ↓
                    BrowserRiskAdapter.preTradeCheck()
                    • Max position: 5% of portfolio
                    • Daily loss circuit breaker
                              ↓
                    Execution Log → .canon/execution/YYYY-MM-DD.jsonl

4 Built-in Strategies

Strategy	Type	Description
Arbitrage Scanner	arbitrage	Detect price spreads across NBA playoff markets
Momentum Trader	momentum	Capitalize on sustained directional movements
Cross-Market Correlator	cross-market	Exploit lag between correlated outcome markets
Speed Opportunity Scanner	speed	Act on injury/stats data before odds adjust

Each strategy supports:

• Run Once — single pipeline execution
• Auto-Run — cron-based interval (e.g. */5 * * * *)
• Real-time agent messages and step-by-step pipeline view

Risk Management

• BrowserRiskAdapter implements RiskInterface from types/RiskInterface.ts
• Max position size: 5% of portfolio value
• Daily loss circuit breaker via localStorage
• All decisions logged with TradeSignal + RiskDecision types

---

Tech Stack

Layer	Technology
Frontend	TypeScript, Vite, Vanilla DOM (no React)
NBA Data	balldontlie.io (free + paid endpoints)
Markets	Polymarket Gamma API (public, no key needed)
Canon Integration	Canon bridge + 4-agent AI pipeline
Risk	RiskInterface + TradeSignal typed interfaces
Logging	localStorage JSONL (browser) + .canon/execution/ (CLI)
Deploy	Vercel

---

Getting Started

# Install
npm install

# Dev server (localhost:5173)
npm run dev

# Run demo pipeline (terminal)
npx tsx scripts/demo.ts

# Run a specific strategy
npx tsx scripts/demo.ts --strategy momentum

# Type check
npm run typecheck

# Full gate (typecheck + lint + tests)
npm run check

Environment Variables (optional)

VITE_NBA_API_KEY=    # balldontlie.io — enables /standings and /injuries
                     # Free endpoints (games, teams) work without key

Polymarket Gamma API is public — no key required. All panels gracefully fall back to mock data when APIs are unavailable.

---

Project Structure

src/
├── components/nba/          # 14 NBA panel components
│   ├── NbaAutomationPanel   # AI agent pipeline UI + auto-run
│   ├── NbaExecutionLogsPanel# JSONL log viewer + download
│   ├── NbaMarketsPanel      # Polymarket live markets
│   ├── NbaArbPanel          # Arbitrage opportunity scanner
│   └── ...                  # Live, Standings, Injuries, etc.
├── services/nba/
│   ├── automation-engine.ts # 4 strategies, 8-step pipeline, auto-run
│   ├── canon-bridge.ts      # Canon CLI integration + BrowserRiskAdapter
│   ├── browser-execution-log.ts  # localStorage JSONL log (500 entry cap)
│   ├── dega-rank.ts         # DEGA Rank positions + performance
│   ├── prediction-market.ts # Polymarket Gamma API + arbitrage
│   ├── client.ts            # balldontlie NBA API + mock fallback
│   └── predictions.ts       # Game prediction model
├── config/
│   ├── variant.ts           # SITE_VARIANT = 'nba'
│   └── panels.ts            # Panel registry
└── app/
    ├── panel-layout.ts      # Panel lifecycle + section nav
    └── app-context.ts       # Shared state

types/
├── TradeSignal.ts           # Signal interface (automation_id, direction, size…)
└── RiskInterface.ts         # Risk check interface

strategies/                  # Canon strategy templates
.canon/
├── execution/               # JSONL execution logs
└── state.json
scripts/
└── demo.ts                  # Terminal demo runner (Canon pipeline)

---

Automation Logs

Every pipeline run writes structured JSONL entries to .canon/execution/YYYY-MM-DD.jsonl:

{"type":"run_start","run_id":"run-1748389201","strategy":"Arbitrage Scanner","mode":"dry-run","timestamp":"2026-05-28T…"}
{"type":"data_fetch","source":"polymarket","count":6,"timestamp":"…"}
{"type":"agent_message","role":"market-analyst","content":"Found 3 spread opportunities…","timestamp":"…"}
{"type":"signal","market":"Will the Celtics win?","direction":"buy_yes","size":25,"confidence":0.78,"timestamp":"…"}
{"type":"risk_check","approved":true,"size":25,"limit":50,"timestamp":"…"}
{"type":"order_submit","status":"dry-run","order_id":"ord-mpp9ox1l","timestamp":"…"}
{"type":"run_complete","status":"success","expected_pnl":0.6,"timestamp":"…"}

Logs are also viewable in the Execution Logs panel in the dashboard and exportable as JSONL.

---

Hackathon Submission

• Event: DEGA NBA Playoffs Prediction Market Hackathon
• Strategy: Arbitrage Detection + Speed-Based Opportunity
• Canon Integration: 4-agent pipeline, TradeSignal, RiskInterface, execution logs
• Demo Video: DEMO.md — 3–5 minute walkthrough script
• Deadline: May 31, 2026

---

License

MIT — Free for personal and commercial use.