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.
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 parameters.
For strategy consulting or purchase inquiries, contact me at Telegram: @casatrick
- 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
- FAQ: Polymarket Arbitrage Bot
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.
- 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
| 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% |
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.
- 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
| 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 |
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 is estimated using:
- Distance of spot price from the market's strike price
- Recent price momentum
- Time remaining until market expiry
Buy the underpriced outcome (YES or NO) and hold for market correction.
| 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 |
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.
Buy the high-probability outcome and hold to resolution.
| 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 |
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.
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 | 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% |
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)
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) |
| 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 |
The bot's risk layer includes:
- Maximum position size per trade
- Maximum portfolio exposure
- Daily loss limits
- Stop-loss and take-profit levels
- Consecutive-loss protection
See risk_manager.py for implementation details.
The bot runs every applicable strategy on every scanned market simultaneously, then merges and ranks the resulting signals.
- Market discovery - the bot discovers all crypto prediction markets for BTC, ETH, XRP, and SOL.
- Parallel scanning - for each market, applicable strategies run concurrently:
- Intra-market (binary markets)
- Combinatorial (3+ outcome markets)
- Cross-platform (where a strike price can be extracted)
- Endgame (all markets)
- Momentum/mean-reversion (all binary markets, per timeframe)
- Signal collection - all signals from all strategies are pooled into a single list.
- Filtering - low-quality signals are discarded:
- Minimum confidence: 0.35 (35%)
- Minimum profit: 0.5%
- Ranking - remaining signals are ranked by composite score.
- Execution - the top-ranked signals execute (max 3 per scan cycle).
Market: "Will BTC be above $100k by Friday?"
| Signal | Detail | Score |
|---|---|---|
| Endgame | YES at $0.96, resolves in 2 hours | 0.91 |
| Intra-market | YES=$0.45, NO=$0.50 → combined $0.95 | 0.85 |
| Cross-platform | Spot implies 70% probability vs. $0.45 YES | 0.72 |
| Momentum (5m) | Oversold condition detected | 0.58 |
Result: the bot executes Endgame first, then Intra-Market if capital allows - ranked purely by composite score.
Multiple strategies can generate signals for the same market. The risk manager blocks duplicate positions in a single market, so only the highest-ranked signal executes.
Cross-Platform and Momentum strategies run across 5m, 15m, and 1h timeframes simultaneously, letting the bot capture short-term scalps, swing trades, and longer position trades from the same market data.
Strategy parameters live in config.py, one configuration class per strategy:
IntraMarketConfigCombinatorialConfigCrossPlatformConfigEndgameConfigMomentumConfig
Each strategy can be enabled or disabled independently via its enabled flag.
Modify STRATEGY_PRIORITY in the StrategyAggregator class:
STRATEGY_PRIORITY = {
"intra_market": 1.0, # Increase for more risk-free arb focus
"combinatorial": 0.95,
"endgame": 0.90,
"cross_platform": 0.80, # Increase for more directional trades
"momentum_mean_reversion": 0.70, # Increase for more technical trades
}Modify the weights in _composite_score():
composite = (
profit_score * 0.30 + # Increase for profit-focused ranking
confidence_score * 0.25 + # Increase for confidence-focused ranking
strategy_score * 0.20 + # Increase for strategy-type preference
urgency_score * 0.15 + # Increase for time-sensitive trades
rr_score * 0.10 # Increase for risk/reward focus
)What is Polymarket arbitrage? Polymarket arbitrage exploits pricing inefficiencies in prediction markets - most commonly when YES + NO token prices don't sum to exactly $1.00, allowing a risk-free profit by buying both sides.
Is a Polymarket arbitrage bot risk-free? Some strategies (intra-market, combinatorial) are structurally risk-free if executed correctly, since both outcomes are purchased for guaranteed redemption. Others (cross-platform, momentum) are directional and carry market risk.
What timeframes does this Polymarket arbitrage bot support? The bot supports 5-minute, 15-minute, and 1-hour timeframes for its momentum and cross-platform strategies, allowing scalping, swing, and position-style trading from the same infrastructure.
Which markets does the bot scan? Crypto prediction markets for BTC, ETH, XRP, and SOL, across all five strategies described above.
Interested in strategy consulting or licensing this Polymarket arbitrage bot? Contact at Telegram: @casatrick