Analysis updated 2026-05-18
Run a paper trading simulation to test AI-driven predictions on Polymarket events without risking money.
Combine multiple AI model predictions with weighted averaging for a single trading signal.
Enforce automated risk limits like drawdown caps and position sizing on a trading strategy.
Monitor engine status, positions, and performance metrics through a web dashboard.
| dexoryn-china/polymarket-arbitrage-bot | 0xtotem/peek-dspy | ant-research/memdreamer | |
|---|---|---|---|
| Stars | 42 | 42 | 42 |
| Language | Python | Python | Python |
| Setup difficulty | moderate | moderate | hard |
| Complexity | 4/5 | 3/5 | 5/5 |
| Audience | developer | developer | researcher |
Figures from each repo's GitHub metadata at analysis time.
Requires API keys for AI providers and a web search provider, live trading needs Polymarket credentials.
This is an automated trading bot for Polymarket, a prediction market platform where users bet on the outcomes of real-world events like elections, economic reports, and sports results. The bot is designed to find markets, research them using web sources, predict probabilities using multiple AI models, manage risk, and place trades, all without requiring manual involvement. The README is written primarily in Chinese. The prediction system combines three large language models at once: GPT-4o, Claude, and Gemini. Each model independently estimates the probability of an outcome based on gathered evidence, and their predictions are combined using weighted averaging. The weights adjust over time based on which model has historically been more accurate in each category. The bot also pulls evidence from authoritative sources matched to the market type, for example government statistics sites for economic questions and official filings for corporate events, and scores each source by how reliable it is likely to be. Risk management is a central feature. Before placing any trade, the bot runs more than 15 checks covering things like maximum drawdown limits, position size caps, minimum edge after fees, minimum market liquidity, maximum spread, and category concentration limits. A four-level drawdown alert system progressively reduces position sizes as losses accumulate and halts all trading at a 20 percent drawdown. Live trading requires three separate configuration switches to all be enabled at once, with paper trading as the default mode. A nine-tab web dashboard runs on port 2345 and shows engine status, open positions, AI predictions with their evidence breakdown, risk metrics, smart-money wallet tracking, a liquidity scanning pipeline, and overall performance statistics including Sharpe ratio and model accuracy over time. Alerts go out through Telegram, Discord, or Slack. All trading decisions are logged with cryptographic checksums for audit purposes. Setup requires Python, API keys for at least one AI provider and one web search provider, and optionally Polymarket API credentials for live trading. Docker deployment is also supported. The disclaimer states that trading on Polymarket carries significant risk of loss and that the authors are not responsible for financial outcomes.
An automated Polymarket trading bot that combines predictions from GPT-4o, Claude, and Gemini with heavy risk management to research, size, and place bets on real-world events.
Mainly Python. The stack also includes Python, Docker, GPT-4o.
No license information is provided in the explanation.
Setup difficulty is rated moderate, with roughly 1h+ to a first successful run.
Mainly developer.
This repo across BitVibe Labs
Verify against the repo before relying on details.