Polymarket Bot Arena

An adaptive trading bot system that runs 4 competing strategies on Polymarket's BTC 5-minute up/down markets. Bots learn from every resolved trade and continuously improve their win rates.

How It Works

4 competing bots trade every active BTC 5-min market on Polymarket (via Simmer for paper trading):

Bot	Strategy	Description
momentum-v1	Trend Following	Trades in the direction of short-term BTC price momentum
meanrev-v1	Mean Reversion	Bets against overextended moves using z-score and RSI
sentiment-v1	Sentiment	Uses social/news sentiment signals
hybrid-v1	Ensemble	Weighted combination of all three strategies

Adaptive learning: Each bot tracks win rates by market conditions (price bucket, BTC momentum, time of day). After every resolved trade, outcomes feed back into a Bayesian learning model that adjusts future decisions. More data = smarter bots.

Evolution: Every 12 hours, bots are ranked by P&L. The bottom 2 are replaced with mutated versions of the winners, carrying forward learned knowledge.

Architecture

trading_bot/
  arena.py           # Main loop: market discovery, trading, resolution, evolution
  learning.py         # Bayesian learning engine (feature extraction + win rate tracking)
  db.py               # SQLite: trades, bot configs, evolution history, learning data
  config.py           # Risk limits, API config, paper/live toggle
  setup.py            # Account setup & verification
  polymarket_client.py # Direct Polymarket CLOB client (for live trading)
  bots/
    base_bot.py       # Abstract base with make_decision() (strategy + learning)
    bot_momentum.py   # Momentum strategy
    bot_mean_rev.py   # Mean reversion strategy
    bot_sentiment.py  # Sentiment strategy
    bot_hybrid.py     # Ensemble strategy
  signals/
    price_feed.py     # Real-time BTC prices via Binance WebSocket
    sentiment.py      # Social sentiment scoring
    orderflow.py      # Polymarket order flow signals
  copytrading/
    tracker.py        # Track top-performing wallets
    copier.py         # Mirror trades from tracked wallets
  dashboard/
    server.py         # FastAPI dashboard backend
    index.html        # Real-time web dashboard with market timers

Setup

Prerequisites

• Python 3.10+
• A Simmer account (free, for paper trading)

Install

pip install websocket-client requests fastapi uvicorn

Configure

1. Get your Simmer API key from https://simmer.markets
2. Save it:

mkdir -p ~/.config/simmer
echo '{"api_key": "your-key-here"}' > ~/.config/simmer/credentials.json

3. Run setup to verify:

python setup.py

Run

# Start the arena (paper trading)
python arena.py

# Start the dashboard (separate terminal)
python dashboard/server.py
# Open http://localhost:8501

Dashboard

Real-time web dashboard showing:

• P&L stats (today / week / all time)
• Per-bot performance with win rates
• Active BTC 5-min market countdown timers
• Recent trades with outcomes
• Evolution history
• Daily earnings chart

Paper vs Live Trading

The system starts in paper mode using Simmer's $SIM currency. To switch to live trading with real USDC on Polymarket:

1. Save your Polymarket wallet private key to ~/.config/polymarket/credentials.json
2. Toggle via the dashboard button or --mode live flag
3. Live mode has stricter risk limits ($10/trade vs $50)

Risk Limits

Setting	Paper	Live
Max per trade	$50 SIM	$10 USDC
Max daily loss per bot	$200 SIM	$50 USDC
Max daily loss total	$500 SIM	$100 USDC

License

MIT