Methodology — Polymarket Wire

What this system is

Polymarket Wire is an automated monitoring and distribution system. It observes probability changes on Polymarket markets and publishes structured alerts when those changes exceed defined thresholds. No analysis is added. No prediction is made. The system records what the market did, not what it means.

Prediction markets aggregate collective belief into a single probability. When that probability shifts, something has changed — new information, changing sentiment, or updated expectations. This system detects those shifts and makes them observable in near real-time.

System principles

Rule-based detection — Signals are triggered only when defined probability thresholds are exceeded. Human judgement is not involved in deciding whether a signal is published.

Signal quality over volume — The system prioritizes meaningful probability movement and includes safeguards that prevent repetitive or low-value alerts.

Transparency of method — Detection thresholds and signal types are publicly documented so users understand how events are classified.

Observation, not interpretation — Signals record what the market did. They do not attempt to explain why it happened or what will happen next.

Signal types and detection thresholds

Markets are sampled at regular intervals. At each sample, the current probability is compared against historical samples within each detection window. If the change exceeds the threshold for that window, a candidate event is generated. When multiple signal types qualify simultaneously, the highest-priority classification is used.

Signal	Window	Threshold	Priority	Description
SHIFT	6 hours	≥ 40 pts	Highest	A large structural repricing over the prior 6-hour window. Indicates a sustained and significant change in collective belief. SHIFT events bypass most daily caps and are exempt from some rate limits that apply to lower-priority signals.
FAST MOVE	30 minutes	≥ 10 pts	High	A sharp probability change within a 30-minute window. Indicates rapid repricing, often correlated with breaking information entering the market. News context from trusted sources is attached when a matching headline is found.
MOVE	60 minutes	≥ 5 pts	Standard	A meaningful directional change sustained over a 60-minute window. Represents a moderate but consistent drift in market probability.

These thresholds apply to most market categories. Sports and other lower-signal categories use higher minimum move sizes: sports MOVE events require ≥ 10 pts; sports SHIFT events are only reported within a 2-hour detection window. The system prioritises quality of signal over volume.

Candidate selection and rate limits

When a market qualifies, it is added to a publish queue, not posted immediately. Before publication, candidates are subject to several rate controls:

Per-market cooldown — 120 minutes between posts for the same market.

Event cooldown — Additional safeguards prevent repeated alerts about the same underlying event within short time windows: at most 2 alerts per event per 24 hours (SHIFT events are exempt from the daily limit), and at most 4 per event per 7 days across all signal types.

Hourly feed cap — at most 3 posts per hour across all markets, to prevent flood events from dominating the feed.

When multiple candidates are queued, the system selects using a priority system that considers signal type, magnitude of movement, and overall feed balance.

News context

For FAST MOVE and SHIFT events, the system queries Google News RSS for headlines from a list of trusted sources (Reuters, AP, BBC, Bloomberg, Financial Times, and others). A trusted source match is a hard gate — if no qualifying headline is found, no context is attached.

Headline matching uses entity overlap: proper nouns in the market question are extracted and matched against the headline. The first matched entity must appear in the headline for the context to qualify. Headlines are normalized before matching.

Context injection is fail-open. If the lookup fails, times out, or returns no trusted match, the signal is still published without context. The absence of a context link does not indicate the absence of a news event.

What this system does not do

The system does not predict outcomes. It records what already happened to a market's probability.

The system does not weight signals by perceived importance. A 40-point shift in a sports market and a 40-point shift in a political market are both reported as SHIFT events.

The system does not editorialize. No commentary, sentiment, or interpretation is added at any stage of the pipeline.

Prediction market probabilities reflect collective belief, not ground truth. High volume on a question does not imply correctness. Probability movement may or may not precede real-world events.