Research
Research papers on financial markets, machine learning and human behavior.
2026
Measuring Trade Direction in a Prediction Market: Settlement Ground Truth and Trading-Cost Measurement on Polymarket
Dubach validates Polymarket's public trade direction against on-chain taker settlement, finding exact side and token agreement on 24.5 million matched pairs across twelve selected days. Comparing conventional trade-sign rules on the same cached tape shows how classification errors, sample selection and weighting change measured trading costs.
Read on arXiv A Buy in the Data, a Sell in the Model Code Dataset
Synthetic Consumer Panels Without Microdata: Iterative Proportional Fitting from Open Aggregate Statistics
Dubach shows that a nationally representative synthetic consumer panel can be built from only freely published aggregate cross-tabulations via iterative proportional fitting, with no individual-level microdata seed. The method is instantiated for Switzerland from two free Federal Statistical Office cubes, producing 10,000 personas whose marginals reproduce the population register and validated against real Swiss human ratings from the ENERGYSCAPE panel.
The Anatomy of a Decentralized Prediction Market: Microstructure Evidence from the Polymarket Order Book
Dubach analyzes 30 billion Polymarket order-book events across 52 days to identify eight microstructure patterns, including a longshot spread premium and the finding that public-feed trade direction matches on-chain ground truth only 59% of the time. Under review at the Journal of Financial Markets.
Read on arXiv The Anatomy of a Decentralized Prediction Market: Notes from the Polymarket Order Book Code
The Online Gambling Fairness Paradox: Cryptographic Verification, Behavioral Harm, and Consumer Protection
Dubach presents a statistical analysis of 20,038 cryptocurrency crash game rounds, confirming cryptographic fairness while showing that mathematical fairness alone does not ensure consumer safety.
Read on SSRN Against All Odds: The Mathematics of 'Provably Fair' Casino Games Code
2025
Attention Dynamics in Online Communities: Power Laws, Preferential Attachment, and Early Success Prediction on Hacker News
Dubach studies attention dynamics in online communities through 38 million Hacker News submissions, examining power laws in content visibility and early predictors of engagement.
Read on SSRN 65% of Hacker News Posts Have Negative Sentiment, and They Outperform
Modeling Postprandial Glycemic Response in Non-Diabetic Adults Using XGBRegressor
Dubach applies a machine-learning approach to predict individual postprandial glycemic responses using continuous glucose monitoring data and meal composition.
2021
A Python Integration of Practical Asset Allocation Based on Modern Portfolio Theory and Its Advancements
Dubach and Hilber present an open-source Python implementation of modern portfolio theory and its extensions, including Black-Litterman, risk parity, and hierarchical risk parity.