kalshi-go

A Go trading bot for Kalshi prediction markets, with analysis and tools for the LA High Temperature market.

🎯 LA High Temperature Strategy

Tools and backtesting for the KXHIGHLAX (Highest Temperature in LA) market.

📚 Full documentation: docs/LAHIGH-STRATEGY.md

🏆 VALIDATED: Ensemble Strategy (82% Win Rate!)

After testing 20+ strategies, we found one that works:

Metric	Value
Win rate	81.8% (9/11 trades)
Total profit (21 days)	+$27.00
Sharpe ratio	4.52 (excellent!)
Kelly fraction	+40% (real edge!)

The Strategy

Trade only when 2+ signals agree:
  1. Market favorite (highest first price)
  2. METAR prediction (rounded to bracket)
  3. 2nd best bracket

If 2+ agree → BUY that bracket
If no consensus → SKIP the day (48% of days)

Real Backtest Results (14 Days)

Date	METAR	Predicted	Winner	Correct	Profit
Dec 25	66°F	66-67°	66-67°	✅	-$0.84
Dec 24	64°F	64-65°	64-65°	✅	-$1.50
Dec 23	63°F	64-65°	64-65°	✅	+$26.60
Dec 22	64°F	64-65°	64-65°	✅	+$0.06
Dec 21	64°F	64-65°	64-65°	✅	+$3.41
Dec 20	65°F	65-66°	65-66°	✅	+$1.49
Dec 19	69°F	69-70°	71-72°	❌	-$9.69
...
Total				50%	$0.77

Recommended Approach

┌────────────────────────────────────────────────────────────────┐
│  FOR LEARNING / THESIS TESTING                                 │
├────────────────────────────────────────────────────────────────┤
│                                                                │
│  1. Use small position sizes ($5-15)                          │
│  2. Hedge across 2-3 brackets                                 │
│  3. Track predictions vs outcomes                             │
│  4. Iterate and improve model                                 │
│                                                                │
│  This is DATA COLLECTION, not guaranteed profit.              │
│                                                                │
└────────────────────────────────────────────────────────────────┘

Quick Start

# Get today's trading recommendation
go run ./cmd/3signal/recommend/

# Run the validated ensemble strategy backtest
go run ./cmd/3signal/strategy/

# Run Monte Carlo simulation (10,000 runs)
go run ./cmd/3signal/montecarlo/

# Monitor today's temperature
go run ./cmd/lahigh-monitor/

Project Structure

kalshi-go/
├── cmd/
│   ├── 3signal/                 # 🏆 PRODUCTION STRATEGY
│   │   ├── recommend/           # Get daily trade recommendation
│   │   ├── strategy/            # Run backtest
│   │   ├── montecarlo/          # Monte Carlo simulation
│   │   └── edge-finder/         # Edge discovery
│   ├── kalshi-bot/              # Generic WebSocket bot
│   ├── lahigh-optimizer/        # Strategy optimizer (20+ strategies)
│   ├── lahigh-4signal-test/     # 4-5 signal experiments
│   ├── lahigh-autorun/          # Automated trading bot
│   ├── lahigh-trader/           # Manual trading bot
│   ├── lahigh-monitor/          # Real-time temperature monitor
│   └── lahigh-*/                # Other analysis tools
├── pkg/
│   ├── ws/                      # WebSocket client
│   └── rest/                    # REST API client
├── docs/
│   └── LAHIGH-STRATEGY.md       # Full strategy documentation
├── results/                     # Backtest output files
└── go.mod

Installation

go mod download

Configuration

Create a .env file with your Kalshi API credentials:

KALSHI_API_KEY=your-api-key-id
KALSHI_PRIVATE_KEY=-----BEGIN RSA PRIVATE KEY-----
...your private key...
-----END RSA PRIVATE KEY-----

Commands

LA High Temperature Trading

# Run rigorous backtest (simulates predictions, honest results)
go run ./cmd/lahigh-backtest-rigorous/

# Run real trade data backtest (shows price evolution)
go run ./cmd/lahigh-backtest-real/

# Monitor real-time temperature at LAX
go run ./cmd/lahigh-monitor/

# Run the trading bot
go run ./cmd/lahigh-trader/ -event KXHIGHLAX-25DEC27

# Run with Docker
docker-compose up --build -d

Generic Kalshi Bot

# Connect and subscribe to a market
go run ./cmd/kalshi-bot -market "KXBTC-25DEC31-T50000" -channel "ticker"

Packages

pkg/ws - WebSocket Client

Full-featured WebSocket client for Kalshi's streaming API.

client := ws.New(
    ws.WithAPIKeyOption("your-api-key", privateKey),
)
client.Connect(ctx)
client.Subscribe(ctx, "MARKET-TICKER", ws.ChannelTicker)

pkg/rest - REST API Client

REST client for order placement and market data.

client := rest.New(apiKey, privateKey)

// Get balance
balance, _ := client.GetBalance()

// Place an order
order, _ := client.BuyYes("KXHIGHLAX-25DEC27-B62.5", 10, 50)

// Get positions
positions, _ := client.GetPositions()

Data Sources

Source	Data	Used For
Iowa State ASOS	Historical METAR	Backtesting
Aviation Weather Center	Real-time METAR	Live monitoring
NWS API	Forecasts	Predictions
Kalshi API	Trade history, prices	Validation

Testing

# Run unit tests
go test ./pkg/ws/...

# Run integration tests (requires credentials)
go test -tags=integration ./pkg/ws/...

Key Learnings

1. Cheap brackets DON'T win: Brackets with first trade <30¢ have 0% win rate
2. Market IS efficient: High-priced brackets are priced high because they're likely to win
3. Single signals fail: +1°F calibration alone only works ~50% of the time
4. Signal consensus works: When 2+ signals agree, win rate jumps to 82%
5. Skip uncertain days: No trade is better than a bad trade

Validated Strategy

✅ ENSEMBLE STRATEGY WORKS

• Win rate: 81.8%
• Sharpe ratio: 4.52
• Kelly fraction: +40% (real edge)
• Trade only when market + METAR + 2nd agree

The key insight: Don't fight the market. Use METAR to confirm the market's pick.

License

MIT