Version the Resolution Wording Behind a Prediction-Market Signal
By DX Research Group · · Market data
A synthetic deadline amendment shows why a market probability needs the question and rules version that gave it meaning.
Attach the captured question and resolution wording to a prediction-market signal. A probability is meaningful only relative to the event the contract resolves. Two versions of the same market can change the interpretation of an observed price even when the market identifier and outcome labels remain stable. We would preserve the wording visible to the collector alongside each feature bundle.
Polymarket's resolution explanation states that markets resolve according to their predefined rules. The archival procedure below is our proposed collector design; the fixture makes no assertion that the platform actually amended a particular market.
A deadline amendment changes the forecast target
Assume a fictional market asking whether a company will announce a product by Friday. Version A specifies Friday 17:00 UTC. Version B, captured later, specifies Friday 23:59 UTC. Assume the contract price is 0.60 under A and 0.65 under B, and the collector can verify the wording change.
| Captured version | Deadline | Price interpretation |
|---|---|---|
| A | Friday 17:00 UTC | Price for the earlier cutoff |
| B | Friday 23:59 UTC | Price for the later cutoff |
The numerical price increase is 0.05. Treating it as a five-percentage-point improvement in belief about an unchanged event is an inference that requires more evidence. The target itself now allows nearly seven additional hours. The example separates a changed forecast question from movement in its market price.
Archive terms that determine the outcome
Retain the question, detailed rules, designated resolution source, deadline, and exception language when those fields are available. Bind them to a market ID and content version. Preserve the original formatting when punctuation or a nested clause affects interpretation, then create a structured summary for features.
Compare newly captured rules to the previous version. Classify changes into material target changes, source changes, and cosmetic edits using a reviewable procedure. A digest identifies a difference, while a semantic comparison explains why the difference matters. Keep the captured evidence behind that classification.
A market title may omit qualifications present in its rules. A research subagent should receive both when assessing relevance to a tradable asset. “Announcement,” “launch,” and “delivery” describe different events; compressing them into a shared label can create a misleading connection to price expectations.
Carry the version into a signal comparison
A proposed replay would freeze the price and change only the wording from A to B. The expected result is a changed event specification, with any asset-impact conclusion explained separately. A second case would change only cosmetic punctuation and require an unchanged structured target.
Our prediction-market provenance test handles information availability. Our social and prediction-market index note describes synthesis with other evidence. Wording versioning adds target stability to those checks. It can establish that a collector preserved the event definition, while a matched forecast and trading-policy study would still be needed to show that the signal improved decisions.