Documentation

Platform Guide

Complete reference for CeeEdge — understanding markets, parameters, and platform rules.

monitoring

Weather Market Monitor

Tracks daily high temperature prediction markets on Kalshi and Polymarket for 5 US cities (NYC, Chicago, Miami, LA, Denver). Overlays GFS ensemble weather model probabilities to identify mispriced brackets and cross-platform pricing discrepancies.

swap_horiz

Arbitrage Monitor

Detects arbitrage across Kalshi, Polymarket, and Opinion using advanced scanning and LLM-validated event matching.

Platform Feature Comparison

Capability	Kalshi	Polymarket	Opinion
Core Data Type	CLOB orderbook	AMM + CLOB hybrid	Event contracts
Settlement Logic	NWS CLI official reports	Weather Underground / UMA Oracle	Platform-specific
Source Feed	REST API (RSA-PSS auth)	Public REST + WebSocket	Cross-platform scan
Scan Frequency	Every 30s	Every 30s	Every 30s
CeeEdge Integration	check_circle	check_circle	check_circle

Platform Overview

CeeEdge is a real-time monitoring platform for prediction markets across multiple exchanges. It combines weather market monitoring with cross-platform arbitrage scanning to surface actionable pricing discrepancies.

thermostat

Weather Markets

Tracks daily high temperature prediction markets for NYC, Chicago, Miami, LA, and Denver. Overlays GFS ensemble probabilities to identify mispriced brackets.

compare_arrows

Arbitrage Scanner

Scans markets across Kalshi, Polymarket, and Opinion for cross-platform pricing discrepancies using fuzzy matching and LLM-validated event pairing.

Weather Dashboard

The main dashboard shows weather temperature prediction markets organized by city and date. Each row represents a temperature bracket — a range of possible daily high temperatures.

Reading the Market Table

Bracket — The temperature range being traded (e.g., "48-49°F").

Kalshi — Current YES price on Kalshi. A price of $0.22 means the market implies a 22% probability.

Polymarket — Current YES price on Polymarket. Same interpretation as Kalshi.

GFS Model — Probability calculated from the GFS 31-member ensemble. This is the "fair value" estimate.

Edge — GFS Model - Market Price. Positive = underpriced, negative = overpriced.

GAP — |Kalshi - Polymarket|. Large gaps indicate potential cross-platform arbitrage.

cloud

GFS Ensemble Explained

The Global Forecast System (GFS) ensemble runs 31 independent simulations with slightly perturbed initial conditions. By counting how many members predict the daily high falling in each bracket, a probability distribution is derived. The ensemble approach captures forecast uncertainty — when members disagree widely, uncertainty is high.

Understanding Parameters

Parameter	Definition	Example
Edge	Model Probability minus Market Price. Positive = underpriced, negative = overpriced.	+25.6%
Cross-Platform Gap	Absolute price difference between Kalshi and Polymarket for the same bracket.	6.2%
Model Probability	Fraction of GFS ensemble members where the daily high falls in this bracket.	47.6%
YES Price	Cost to buy a YES contract. Pays $1 if the event occurs, $0 if not.	$0.22
Arbitrage Cost	YES price on Platform A + NO price on Platform B. If < $1.00, arbitrage exists.	$0.93
ROI	Return on investment: (payout - cost) / cost.	7.5%
Annualized ROI	ROI scaled to yearly rate based on days until close.	~195%
Settlement Source	Official data source used to determine outcome. Differs by platform.	NWS CLI

Edge Signals

Signals are triggered when pricing discrepancies exceed configurable thresholds.

analytics

Model Edge Signal

Triggered when GFS model probability differs from market price by more than 8% (configurable). A positive edge means the weather model assigns higher probability than the market — the bracket may be underpriced.

swap_horiz

Cross-Platform Gap Signal

Triggered when Kalshi and Polymarket prices for the same bracket differ by more than 5%. Large gaps may indicate stale pricing, different settlement expectations, or genuine arbitrage opportunities.

bolt

True Arbitrage Signal

Triggered when buying YES on one platform and NO on another costs less than $1.00 total. This guarantees a profit regardless of outcome. Requires sufficient liquidity at both price levels to be executable.

Arbitrage Monitor

The arbitrage monitor scans all markets across Kalshi, Polymarket, and ForecastEx. It works in three stages:

1. Market Matching

Uses fuzzy string matching (token_set_ratio) plus LLM validation to pair identical events across platforms. Accounts for different naming conventions, closing dates, and event types. A match score of 88+ is auto-approved; lower scores go through LLM validation.

2. Orderbook Hydration

For the top 12 prioritized pairs, fetches real-time orderbook depth (bid/ask ladders) from each platform. This determines actual executable prices at various position sizes.

3. Arbitrage Testing

Tests both strategies per pair: (A) Buy YES on Platform 1 + Buy NO on Platform 2, and (B) the reverse. Factors in platform fees (Kalshi: 0.7%, Polymarket: 0.3%). Reports true arbitrage (cost < $0.98, ROI > 3%), near-misses, and micro-edges.

Platform Rules

Each platform has unique settlement rules that can affect outcomes. Understanding these differences is critical for cross-platform strategies.

gavelView Settlement Rules

GFS Model

The Global Forecast System (GFS) ensemble consists of 31 independent model runs with slightly perturbed initial conditions. By counting how many members predict the daily high falling in each bracket, we derive a probability distribution.

Source: Open-Meteo Ensemble API (free, no auth required)

Model: GFS Seamless (combines HRRR/RAP/GFS for best resolution)

Members: Up to 31 ensemble members

Refresh: Every 6 hours (aligned with GFS run times: 00Z, 06Z, 12Z, 18Z)

Forecast range: Today + 2 days

Probability: count_in_bracket / total_members

Limitation: The GFS model is a numerical weather prediction, not a market prediction. Model probabilities may diverge from market prices due to model bias, time lag between runs and pricing, and different settlement stations vs. model grid points.

Settlement

Weather markets settle based on the official daily high temperature recorded by the designated weather station. Critical differences exist between platforms.

Kalshi Settlement

chevron_rightSource: NWS CLI (National Weather Service Climate Logs)
chevron_rightPublished next morning (~6-8 AM local time)
chevron_rightUses calendar day in local timezone
chevron_rightStations: KNYC, KMDW, KMIA, KLAX, KDEN

Polymarket Settlement

chevron_rightSource: Weather Underground (weather.com)
chevron_rightUpdates more frequently, may use different stations
chevron_rightNYC: May use KLGA (LaGuardia) instead of KNYC
chevron_rightStation difference can result in 1-3°F discrepancy

Known Trap: During DST transitions, settlement windows may shift. Some platforms may record the temperature from a 25-hour or 23-hour calendar day, affecting the daily high.

Glossary

Term	Description
METAR	Meteorological Terminal Aviation Routine weather report from airports
NWS CLI	National Weather Service Climate Log Issued — official daily weather summary
GFS	Global Forecast System — NOAA's primary numerical weather prediction model
Ensemble	Multiple model runs with perturbed initial conditions providing probability distributions
Bracket	A temperature range in a prediction market (e.g., 48-49°F)
Threshold Market	Market betting on above/below a temperature (e.g., ≥54°F)
YES Price	Cost of a contract that pays $1 if the event occurs
Edge	Difference between model probability and market price
Arbitrage	Risk-free profit from buying complementary contracts across platforms
Near-Miss	Cross-platform cost between $0.98-$1.08 — close to arbitrage but not guaranteed
Micro-Edge	Cost between $0.995-$1.005 — essentially breakeven after fees
RSA-PSS	Signature scheme used by Kalshi API for authenticated requests
CLOB	Central Limit Order Book — trading engine for token-based markets

For the interactive settlement rules comparison, visit the Rules page. For live arbitrage scanning, visit the Arbitrage page.