DV-PMI · Press
Available for comment within 24 hours, weekdays
Quotes ready to cite, recent press, speaking engagements, and contact info. All quotes below released under CC BY 4.0; edit for length or style, keep the attribution. No interview required; an email heads-up is appreciated so the corresponding paper can be linked or context clarified.
Contact jdellavedova@sandiego.edu
1200×675 PNGs (16:9, optimized for Twitter / LinkedIn / Slack preview cards). Updated weekly. CC BY 4.0 with attribution. No interview required to use.
Required attribution: “Source: Della Vedova Prediction Market Indices (DV-PMI), jdellavedova.com.” All charts derived from 671M on-chain Polymarket trades since November 2022.
All quotes released under CC BY 4.0; edit freely as long as attribution remains.
“Bots now capture nearly every dollar of volume on Polymarket, up from about half two years ago. The retail trader's share of the market is vanishing.”
From Paper 1 and the weekly Bot Share of Volume index: bots execute 94% of weekly trades in recent weeks.
Joshua Della Vedova, Associate Professor of Finance, University of San Diego. Source: Della Vedova (2026), 'Who Profits from Prediction Markets? Execution, not Information.'
“About one in a hundred Polymarket wallets show trading signatures consistent with private information. The flag rate is three times higher in markets where humans control the outcome, such as elections and awards, than in markets driven by aggregate forces, such as crypto prices.”
From Paper 2: 6,032 of 450,048 wallets flagged at p < 0.01; flag rate is 2.79% in vote-category markets, 1.20% in stochastic markets.
Joshua Della Vedova, Associate Professor of Finance, University of San Diego. Source: Della Vedova (2026), 'Detecting Informed Trading in Prediction Markets: An Orthogonality Test,' SSRN abstract 6567238.
“Human traders on Polymarket distort probabilities in exactly the way Tversky and Kahneman predicted in 1992. The weekly pattern is visible, measurable, and remarkably stable.”
From Paper 4 (in progress): non-bot Prelec alpha averaged roughly 0.65 across 150 weeks, matching the canonical experimental estimate.
Joshua Della Vedova, Associate Professor of Finance, University of San Diego. Source: Della Vedova and Grant (2026, in progress), 'Probability Weighting from Prediction Markets.'
“Information asymmetry in prediction markets imposes real costs on uninformed traders. Approximately $150 million in profits flowed to wallets trading on private information; most of that was absorbed by automated market makers and passed through to retail as wider spreads.”
From Paper 2. Welfare interpretation is bounded: the market is zero-sum by construction, so transfer estimates are lower bounds under current fee structures.
Joshua Della Vedova, Associate Professor of Finance, University of San Diego. Source: Della Vedova (2026), 'Detecting Informed Trading in Prediction Markets,' SSRN abstract 6567238.
“Trading skill in prediction markets is two-dimensional. Forecasting accuracy and execution quality are different capacities drawing on different resources, and they are nearly uncorrelated at the trader level. Execution, not forecasting, is what determines who profits.”
From Paper 1. Bots achieve coin-flip forecasting accuracy (49.9%) yet earn positive returns via execution; active retail achieves 51.3% accuracy but loses money due to poor execution.
Joshua Della Vedova, Associate Professor of Finance, University of San Diego. Source: Della Vedova (2026), 'Who Profits from Prediction Markets? Execution, not Information.'
Upcoming and recent talks will appear here.
Joshua Della Vedova is an Associate Professor of Finance at the Knauss School of Business, University of San Diego. His research on prediction markets draws on 671 million on-chain Polymarket trades. See the research page for working papers, the methodology page for index construction, and the DV-PMI dashboard for live indices.
Quotes are released under Creative Commons Attribution 4.0. Attribution line included with each quote is the minimum citation; journalists are welcome to edit for length or style as long as the attribution remains.
