Polymarket Arbitrage Bot: 5 Trading Strategies Explained
Python Polymarket arbitrage bot for crypto prediction markets (BTC, ETH, XRP, SOL). Runs five strategies in parallel - intra-market, combinatorial, cross-platform, endgame, and momentum/mean-reversion - with composite signal ranking and risk controls.
Landing page: casatrick.github.io/polymarket-arbitrage-bot-python (overview, quick start, FAQ) · Full docs (this page): strategy breakdown, configuration, and risk management · Source: github.com/casatrick/polymarket-arbitrage-bot-python
https://github.com/user-attachments/assets/a85d94c6-0f45-414b-8f49-70f44c4cb85e
Demo: Polymarket arbitrage bot scanning crypto markets and ranking multi-strategy signals.
A complete breakdown of the five arbitrage strategies used by this Polymarket arbitrage bot - covering intra-market arbitrage, combinatorial arbitrage, cross-platform arbitrage, endgame arbitrage, and momentum/mean-reversion trading across 5-minute, 15-minute, and 1-hour timeframes.
What I Offer
I build and maintain trading infrastructure like this for clients on Polymarket. Available as:
- Custom strategy builds - a bot tailored to your specific edge, risk tolerance, and market focus
- Licensing this bot - the five-strategy system below, configured and deployed for you
- Ongoing maintenance & new strategies - monthly retainer for tuning, monitoring, and extending an existing bot
All builds ship with paper-trading mode first, full source code ownership, and no black-box logic - you get the same level of parameter transparency shown in this doc.
Contact: Telegram @casatrick · GitHub: casatrick/polymarket-arbitrage-bot-python
Security & Key Handling
- Private keys are read from environment variables (
.env) and never hardcoded or logged. - Recommended setup uses a dedicated trading wallet with only the capital allocated to the bot - not a main wallet.
- Clients retain full source code and full custody of keys at all times; nothing runs through third-party infrastructure you don't control.
- Position size caps, daily loss limits, and consecutive-loss protection (see Risk Management) are enforced independently of strategy logic, so a bad signal can't bypass risk controls.
Table of Contents
- What I Offer
- Security & Key Handling
- Strategy 1: Intra-Market Arbitrage
- Strategy 2: Combinatorial Arbitrage
- Strategy 3: Cross-Platform Arbitrage
- Strategy 4: Endgame Arbitrage
- Strategy 5: Momentum / Mean-Reversion
- How Signal Ranking Works
- Risk Management
- How the Strategies Are Combined
- Configuration Reference
- Roadmap
- About the Developer
- License
- FAQ: Polymarket Arbitrage Bot
What Is a Polymarket Arbitrage Bot?
A Polymarket arbitrage bot is an automated trading system that scans Polymarket's prediction markets for pricing inefficiencies and executes trades to capture the resulting spread. This bot runs five distinct arbitrage and trading strategies in parallel, each targeting a different type of market inefficiency, then ranks and executes the highest-quality signals automatically.
This document covers how each strategy works, how signals are scored, and how to configure the bot's params.
Strategy 1: Intra-Market Arbitrage
How Polymarket YES/NO arbitrage works: in binary prediction markets, YES and NO token prices should always sum to $1.00. When combined YES + NO pricing falls below $1.00 after fees, buying both outcomes locks in a guaranteed, risk-free profit - the core mechanic behind most Polymarket arbitrage bot strategies.
Execution
- Buy YES and NO tokens separately through the CLOB API
- Merge tokens into complete sets
- Redeem complete sets immediately for $1.00 per share
- Profit is realized immediately - no waiting for market resolution
Configuration
| Parameter | Description | Default |
|---|---|---|
min_spread_pct |
Minimum profit % required | 1.5% |
max_position_usd |
Maximum position size | $500 |
fee_pct |
Polymarket fee percentage | 2.0% |
Strategy 2: Combinatorial Arbitrage
For multi-outcome Polymarket markets (e.g., price-range or bracket markets), all outcome prices should sum to $1.00. When the total falls below $1.00, buying every outcome guarantees a profit - a scalable extension of classic Polymarket arbitrage bot logic to markets with 3+ outcomes.
Execution
- Buy all outcome tokens separately through the CLOB API
- Merge tokens into complete sets
- Redeem complete sets immediately for $1.00 per share
- Profit is realized immediately - no waiting for resolution
Configuration
| Parameter | Description | Default |
|---|---|---|
min_deviation_pct |
Minimum deviation from $1.00 | 2.0% |
max_position_usd |
Maximum position size | $300 |
min_outcomes |
Minimum number of outcomes required | 3 |
Strategy 3: Cross-Platform Arbitrage
This strategy compares Polymarket prediction prices against real-time spot prices from exchanges like Binance and CoinGecko. When a significant discrepancy appears, the bot trades on the assumption that Polymarket pricing will converge toward fair value - a directional, convergence-based approach rather than risk-free arbitrage.
Fair Probability Calculation
Fair probability is estimated using:
- Distance of spot price from the market's strike price
- Recent price momentum
- Time remaining until market expiry
Execution
Buy the underpriced outcome (YES or NO) and hold for market correction.
Configuration
| Parameter | Description | Default |
|---|---|---|
min_price_diff_pct |
Minimum price difference to trigger | 3.0% |
max_position_usd |
Maximum position size | $1000 |
stale_threshold_sec |
Maximum age of price data | 30s |
Strategy 4: Endgame Arbitrage
When a market approaches resolution and one outcome shows very high probability (>93%), buying that outcome captures a small but near-certain return. This is one of the more capital-efficient strategies in a Polymarket arbitrage bot, since time-to-resolution is short and predictable.
Execution
Buy the high-probability outcome and hold to resolution.
Configuration
| Parameter | Description | Default |
|---|---|---|
min_probability |
Minimum outcome probability | 0.93 |
max_time_to_resolution_hrs |
Maximum hours until resolution | 48 |
min_annualized_return_pct |
Minimum annualized return | 100% |
max_position_usd |
Maximum position size | $2000 |
Strategy 5: Momentum / Mean-Reversion
This strategy tracks Polymarket YES/NO prices as a time series and applies standard technical indicators: Z-score, RSI, rate of change, and VWAP divergence - the same tools used in crypto and equities momentum trading, adapted for prediction markets.
Entry Conditions
Mean Reversion Buy (Oversold):
- Z-score below -threshold
- RSI < 35
- Rate of change > -0.5%
- Action: buy YES (expect reversion upward)
Mean Reversion Sell (Overbought):
- Z-score above +threshold
- RSI > 65
- Rate of change < 0.5%
- Action: buy NO (expect YES price to revert down)
Timeframe Parameters
| Timeframe | Lookback | Entry Z-score | Take Profit | Stop Loss |
|---|---|---|---|---|
| 5-minute (scalping) | 12 candles (1 hr) | ±2.0σ | 1.5% | 1.0% |
| 15-minute (swing) | 16 candles (4 hrs) | ±1.8σ | 3.0% | 2.0% |
| 1-hour (position) | 24 candles (1 day) | ±1.5σ | 5.0% | 3.0% |
Configuration
Each timeframe has its own parameter set in MomentumConfig:
tf_5m_lookback,tf_5m_entry_zscore,tf_5m_take_profit_pct,tf_5m_stop_loss_pcttf_15m_*(same pattern)tf_1h_*(same pattern)
How Signal Ranking Works
The StrategyAggregator scores every signal (0.0-1.0) using a weighted composite:
score = (profit_score × 0.30) +
(confidence_score × 0.25) +
(strategy_priority × 0.20) +
(urgency_score × 0.15) +
(risk_reward_score × 0.10)
| Component | Description |
|---|---|
profit_score |
Expected profit % ÷ 10 (capped at 1.0) |
confidence_score |
Signal confidence (0.0-1.0) |
strategy_priority |
Strategy-type weight (see below) |
urgency_score |
HIGH = 1.0, MEDIUM = 0.67, LOW = 0.33 |
risk_reward_score |
Risk/reward ratio × 5 (capped at 1.0) |
Strategy Priority Weights
| Rank | Strategy | Priority | Reliability |
|---|---|---|---|
| 1 | Intra-market | 1.00 | Risk-free arbitrage |
| 2 | Combinatorial | 0.95 | Risk-free arbitrage |
| 3 | Endgame | 0.90 | High probability |
| 4 | Cross-platform | 0.80 | Directional, requires convergence |
| 5 | Momentum / mean-reversion | 0.70 | Technical, less certain |
Risk Management
The bot's risk layer includes:
- Maximum position size per trade
- Maximum portfolio exposure
- Daily loss limits
- Stop-loss and take