Prediction Market Arbiter — Cross-Platform Divergence Scanner

Overview

Prediction Market Arbiter is a systematic price comparison engine for prediction markets. It scans Kalshi and Polymarket simultaneously, identifies identical or near-identical events priced differently across platforms, and alerts you to potential arbitrage opportunities or market mispricings.

What It Does

Core Function: Compare YES prices for the same event across Kalshi and Polymarket. When the same market exists on both platforms but with significant price divergences, alert the trader to potential inefficiency.

Example:

• - Market: "Will Bitcoin exceed $100k by Dec 2026?"
• Kalshi price: 72¢ (72% implied probability)
• Polymarket price: 58¢ (58% implied probability)
• Divergence: 14 percentage points = arbitrage signal

The arbiter detects these inefficiencies automatically, filters by volume and match quality, and ranks them by spread size.

Key Features

• - Fuzzy Title Matching — Uses Jaccard similarity (word overlap) to identify the same event across different platform vocabularies
• Configurable Divergence Threshold — Default 8%; tune for your risk tolerance
• Volume Filtering — Combines Kalshi + Polymarket volume; minimum 1000 default
• Match Confidence — Only pairs markets with >60% title similarity (configurable)
• Pagination — Fetches top 1000 Kalshi markets (5 pages × 200) + top 200 Polymarket (by volume)
• Cache Results — Saves matches to JSON for programmatic access and follow-up analysis
• Multi-Platform Support — Uses Kalshi API (with client authentication) and Polymarket Gamma API (no auth required)

When to Use This Skill

• - You trade across Kalshi and Polymarket and want to find price inefficiencies automatically
• You're arbitraging prediction markets and need real-time divergence alerts
• You want to understand market structure: which platform leads on certain event types?
• You're analyzing prediction market microstructure and efficiency
• You need a scheduled divergence scanner (every 4 hours, daily, etc.)

Requirements

API Keys & Credentials

1. 1. Kalshi API Key (free, required)

- Sign up at https://kalshi.com - Navigate to Settings → API - Generate API credentials (key ID + private key file) - Cost: Free tier, unlimited reads

1. 2. Polymarket Access (free, no authentication)

- Gamma API is public: https://gamma-api.polymarket.com - No API key required - No cost

Python & Dependencies

• - Python 3.10 or higher
• Required packages:

CODEBLOCK0

Configuration

Create or update your config.yaml:

CODEBLOCK1

Threshold Tuning Guide

Parameter	Default	When to Adjust
INLINECODE1	8.0%	Lower (5%) for more alerts; higher (15%) for only major spreads
INLINECODE2

0.6 | Raise to 0.7+ to be stricter (fewer false matches); lower to 0.5 for loose matching | | min_volume | 1000 | Lower for illiquid opportunities; raise to 5000+ for only liquid pairs |

How It Works

Phase 1: Fetch Markets

Kalshi: Paginate through active markets (status=open), fetching up to 5 pages of 200 markets each (max 1000 markets).

• - Extract: title, ticker, yes_price (cents), volume, open interest

Polymarket: Fetch top 200 markets sorted by volume from Gamma API.

• - Extract: question text (title), yes_price (in cents), volume

Phase 2: Fuzzy Title Matching

For each Kalshi market, compare against all Polymarket markets using Jaccard similarity:

CODEBLOCK2

Only pairs with similarity >= threshold advance to comparison.

Example — Successful Match:
CODEBLOCK3

Example — No Match (Different Events):
CODEBLOCK4

Tuning note: Markets with similar topics but different vocabulary (e.g., "Bitcoin" vs "BTC") may not match at default thresholds. Lower fuzzy_match_threshold to 0.4-0.5 for broader matching, or use the aggressive config profile.

Phase 3: Filter by Volume

Combined volume = Kalshi volume + Polymarket volume

• - If combined volume < min_volume threshold, drop the pair
• Filters out low-liquidity pairs where arbitrage isn't actionable

Phase 4: Calculate Divergence

For each matched pair:

CODEBLOCK5

If delta_pct >= threshold, add to alerts.

Phase 5: Sort & Alert

Sort divergences by largest spread first. Return top N (default 5 in alerts).

Example output:
CODEBLOCK6

Cache Results

Write all matches (top 20) to JSON:

CODEBLOCK7

Divergence Interpretation

Common Divergence Patterns

1. Polymarket Leads on Hype Events

• - Recent news → PM moves first (higher volume of retail traders)
• Look for Kalshi underpriced relative to PM on breaking news

2. Kalshi Leads on Structural Events

• - Predictable political cycles, economic data → Kalshi often ahead
• PM sometimes lags

3. Liquidity-Driven Spreads

• - If one platform is much more liquid, it's often the "fair price"
• Illiquid platform = potential mispricing

4. Time-of-Day Effects

• - US markets hours: Kalshi often tighter
• Asian hours: PM may be the pricing engine

Scheduling

Command-Line Usage

CODEBLOCK8

As a Cron Job (Every 4 Hours)

CODEBLOCK9

As an OpenClaw Scheduled Task

CODEBLOCK10

Example Output

Alert Message

CODEBLOCK11

Cache Output

File: INLINECODE5

CODEBLOCK12

API Reference

Main Function

CODEBLOCK13

Configuration Object

CODEBLOCK14

Troubleshooting

No Matches Found

Possible causes:

1. 1. Threshold too strict — Raise divergencethresholdpct or lower fuzzymatchthreshold
2. Low market overlap — Most Kalshi markets don't have Polymarket equivalents
3. Both platforms down — Check status pages

Wrong Matches (False Positives)

If fuzzy matching is pairing unrelated markets:
CODEBLOCK15

API Failures

Kalshi API error:

• - Check API key validity: kalshi_api_key_id and private_key_file path
• Verify key has "read-only" permission (sufficient for this skill)

Polymarket API error:

• - Polymarket Gamma API is public; if it fails, their servers are down
• Check https://gamma-api.polymarket.com/markets directly in browser

Performance

Typical run: 30-90 seconds

• - Kalshi fetch: 10-30s (paginating 5 pages)
• Polymarket fetch: 5-15s
• Matching: <5s
• If taking >2 minutes, reduce kalshimaxpages or polymarket_limit

Performance & Cost

Typical Run Metrics

• - Runtime: 30-90 seconds
• API calls:

- Kalshi: 5 calls (page 1-5) - Polymarket: 1 call (top 200 by volume)

• - Cost: $0 (all reads, no premium APIs)

Scaling

For 6 runs per day (every 4 hours):

• - Kalshi: 30 API calls/day (free tier handles 1000s)
• Polymarket: 6 API calls/day (public API)
• Total cost: $0

OpenClaw Ecosystem Integration

Cross-platform price comparison engine for the Prediction Market Trading Stack.

Connected Skill	How It Connects
Market Morning Brief	Divergences appear in your daily morning digest
Kalshi Command Center

Trade the Kalshi side of flagged divergences |
| Polymarket Command Center | Browse the Polymarket side for context |
| Kalshalyst | Compare contrarian edges against cross-platform pricing |

Install the complete stack:
CODEBLOCK16

Implementation Notes

Battle-tested in production trading environments. Design principles:

1. 1. Standalone config/logging — works with any OpenClaw setup
2. Direct alerts — sends divergences straight to you
3. All thresholds and matching logic — refined through live cross-platform trading
4. Standalone scripts — zero external dependencies beyond listed packages

Configuration Example

conservative.yaml (Wide Spreads Only)

CODEBLOCK17

aggressive.yaml (Early Detection)

CODEBLOCK18

Further Reading

• - See references/fuzzy-matching.md for detailed explanation of title matching and threshold tuning

Support & Iteration

Common iteration paths:

1. 1. Threshold Tuning — Adjust divergencethresholdpct and fuzzymatchthreshold based on your alert frequency
2. Volume Filtering — Increase minvolume to focus on only liquid pairs
3. Scheduling — Change checkintervalminutes (240 = 4 hours, 60 = 1 hour, etc.)
4. Platform Expansion — Add new platforms (Manifold Markets, Betfair, etc.) by implementing fetchXmarkets() functions

License & Attribution

Author: KingMadeLLC

---

Feedback & Issues

Found a bug? Have a feature request? Want to share results?

• - GitHub Issues: github.com/kingmadellc/openclaw-prediction-stack/issues
• X/Twitter: @KingMadeLLC

Part of the OpenClaw Prediction Stack — the first prediction market skill suite on ClawHub.