Toward Black–Scholes for Prediction Markets: A Unified Kernel and Market-Maker's Handbook

Battle Count: 70

Highly relevant to quantitative trading, particularly in the emerging prediction market space (Polymarket, Kalshi). The paper directly addresses market making, inventory management, hedging, and risk transfer in event contracts. The Avellaneda-Stoikov quoting framework, variance/correlation hedging, and PnL attribution are directly applicable to trading desk operations. The institutional interest (ICE's $2B investment in Polymarket) and growing volumes make this immediately relevant. However, it is more specialized than general equity/FX quant trading, focusing on binary event contracts rather than continuous price processes.

High complexity due to: (1) multi-stage calibration pipeline (heteroskedastic Kalman filtering, EM for jump-diffusion separation, RN drift enforcement with Monte Carlo inner expectations, re-smoothing); (2) PIDE solvers for exotic derivatives with jump integrals; (3) surface construction with shape constraints across (τ, m); (4) real-time quoting systems with toxicity filters, news guards, and inventory-aware reservation prices; (5) multi-event correlation and co-jump estimation. The mathematical framework is sophisticated (Itô-Lévy calculus, martingale restrictions, compensated jump measures). However, the paper provides clear equations and algorithmic steps that reduce ambiguity.

The paper provides detailed mathematical formulations, calibration pipeline steps, and experimental setup (20 Polymarket trades, synthetic RN-consistent paths, fixed seeds, specific hyperparameters). However, no code repository is explicitly linked. The methodology is well-described but implementation details (exact EM convergence criteria, specific jump law parameters, PIDE solver settings) would require additional work to reproduce fully. The synthetic experiment is well-specified but real-data experiments are limited.