GPT-6 Kalshi Trading Bot: Week 1 Performance Update
All About AIgo watch the original →
the gist
A custom AI trading bot using Google DeepMind's weather models on Kalshi generated a $34 profit in its first five days of operation by executing trades only when a 5-cent conservative edge was identified.
Performance Overview
The automated trading bot, deployed on a 24/7 VPS, generated a total profit of $34 over its first five days of operation. The system focuses on New York City weather markets using Google DeepMind weather data. While the first day resulted in a small loss, subsequent days yielded consistent gains, including a $14 profit on a single position. The creator notes that the current results may reflect beginner's luck and plans to monitor performance over a 30-day period to determine if the edge persists.
Trading Logic and Execution
The bot employs a strict risk-management strategy that prioritizes expected value over trade frequency. Key operational parameters include:
- Edge Threshold: The bot only executes a trade if it identifies a minimum 5-cent conservative edge between the model's prediction and the market price on Kalshi.
- Model Freezing: The system locks its predictions 15 minutes before the market opens to prevent late-stage volatility from skewing the entry logic.
- Conservative Bias: When multiple buckets are close, the bot selects the side with the largest conservative edge. If no bucket meets the 5-cent threshold, the bot defaults to a no-trade position.
During the update, the bot correctly identified that no market conditions met the required edge threshold for the day, resulting in zero trades. The creator emphasizes that avoiding trades during periods of low confidence is a core component of the system's design.