Launch-Time Contract Information Predicts Trading Volume and Dispute Risk in Prediction Markets
ORIGINAL / What Prediction-Market Contracts Reveal at Launch: Trading Volume and Dispute Risk
Using a near-census of resolved Polymarket markets, this study predicts lifetime trading volume and settlement disputes solely from listing-time contract information (title, description, resolution source, metadata). It assesses interpretable features versus neural text models and documents performance degradation under platform change. The work offers a contract-text-based screening approach for market designers and researchers, but dispute prediction precision is too low fo
01 ABSTRACT
Using a near-census of 802,863 resolved markets, the study shows that listing-time contract information predicts lifetime trading volume well (out-of-sample R²=0.747), with market topic contributing most; dispute risk can be ranked (AUC=0.78) but individual prediction is imprecise at a 0.31% base rate (average precision 0.012). The authors find that volume prediction performance drops sharply after a platform format change, with losses concentrated in question forms absent from training and attributed more to a changed relationship between contract characteristics and demand than to an outdated text-to-feature mapping. These findings are specific to one platform and period.
02 KEY FINDINGS
- Launch-time contract information explains most variation in trading volume; market topic is the strongest predictor, while wording and resolution-rule design add smaller measurable gains.
- Dispute risk is rankable (AUC 0.78) but at a 0.31% base rate average precision is only 0.012, making case-by-case prediction unreliable.
- Platform format change causes large performance drops in volume prediction (AUC from 0.940 to 0.577), with losses concentrated in question forms unseen during training.
AI GENERATED SUMMARY / DISCOVERED BY ARXIV CS.LG · ARXIV Q-FIN