The State Of Prediction Markets
The first trading crowd where every claim gets graded, and almost no one is grading it.
Nobody knows if the account posting picks in your Discord is actually good. Nobody knows if their stock-picking subscription is worth the $100 a month. Estimates exist: screenshots, win-rate claims, “trust me.” Ledgers do not. The verification problem is the same one every trading community has faced since trading communities existed, with one exception.
Every position taken on a federally regulated U.S. prediction market resolves to a single, public, undisputed number: correct or incorrect, paid out at $1 or $0. We call this property resolution, and this paper argues that it, not the size of the market or the novelty of the assets, is the load-bearing fact of this category. Nearly everything that matters follows from it: why skill is measurable here in a way it isn’t anywhere else in retail trading, why almost nobody is actually measuring it yet, and why the gap between those two facts is where Chad sits.
The market, 2026
Estimates of market size vary with the boundary drawn: regulated event contracts alone versus event contracts plus offshore and sportsbook-adjacent volume. We present the range rather than the most flattering point in it. The dispersion is itself a datum: this is a young market measured with unstandardized definitions.
The sports book is scaling faster than the market it sits inside. That growth is the reason the questions that follow matter now: a much larger share of participants is new, unlogged, and pricing contracts off social consensus.
Growth, and why now
The measurement layer is worth building now because the thing being measured is compounding faster than the tooling around it. Volume grew roughly 5x in a single year, and the 2030 projections are not one bank’s outlier; three independent analyst firms land in the same range.
The category is also still maturing. A December 2025 Vanderbilt University study of the 2024 election cycle found Kalshi correctly predicted 78% of its markets against 67% for Polymarket, and that prices sometimes moved on “herd behavior” rather than pure information. That is context about the market, not a claim about any individual: nothing here suggests that keeping a record makes a trader more likely to beat the market.
Claims without ledgers
Retail trading communities have run on unverifiable authority for as long as they've existed. A caller posts a win, a following grows, money changes hands for access, and no one downstream can check the record behind the claim. This isn't a prediction-market problem specifically. It's structural to U.S. retail trading generally, and regulators are now documenting it directly.
Unverifiable skill. Social media users and finfluencer followers answered an average of just 42% of an objective investment-knowledge quiz correctly, while 63% rated their own investment knowledge as high. The people most confident in the advice they follow are the least equipped to evaluate it.
Escalating, predictable losses in sports betting. Across 183,821 U.S. households and 4 million+ household-quarters, frequent sports bettors deposited $1,138 per quarter, over 5% of their income, versus $93 for occasional bettors. Twelve quarters after a first bet, the average frequent bettor was depositing roughly 8x their original amount. These bettors already showed worse financial health before their first bet: lower income, less investing, higher credit card balances.
An official warning, largely unheeded.FINRA’s own investor alert lays out the risk plainly for a U.S. audience. The volume of retail day trading has only grown since.
Any market that could fix the underlying problem (verifiable skill, measured behavior, a real feedback loop) would be categorically different from the retail trading crowds that came before it. One already has the raw material to.
The resolution
The claim of this paper is that one property separates prediction markets from every other trading crowd. Stated precisely: every position is a claim about a future event, and every claim resolves publicly, on a fixed date, to an outcome no one can dispute after the fact. A stock “call” is never actually settled; you can argue forever about whether it was a good trade. A Kalshi position is settled the moment the event happens, at a price the exchange itself published in advance.
Three consequences follow. Skill is auditable: a track record is a database query against public settlement data, not a screenshot. Confidence is scoreable: because a trader can state how surethey were, not just which side they took, calibration, not just win rate, becomes measurable for the first time in a retail trading context. And the record can’t be edited after the fact: unlike a deleted tweet or a quietly unpinned losing call, a settled market stays settled.
What resolution enables
Treat resolution as a primitive and the missing infrastructure of retail trading follows from it.
Calibration scoring.With confidence stated up front and an undisputed outcome on the back end, a trader’s accuracy can be scored against their own stated certainty: a Brier score, not a win-rate, the same measurement professional forecasters are held to.
Auditable track records.A source’s history becomes a ledger, not a highlight reel: every call, every outcome, timestamped before resolution so it can’t be edited into a better story afterward.
Source accountability.Followers can finally check the thing FINRA’s own research shows they currently can’t: whether the person they are paying for picks actually knows more than they do, or is simply more confident.
The same research quantifies the downside of trusting the wrong source. Among investors targeted for fraud, 68% of social media users and 69% of finfluencer followers reported actually losing money to it, compared with 26–29% of investors who don’t use these channels.
The evidence that structured measurement, not more content or more confidence, actually improves judgment over time is not new. It’s just never been applied here.
The intervention in that research wasn't more information. It was structured feedback on a forecaster's own calibration, over time, the same mechanism Chad is built around, applied to a market that, unlike the one those forecasters were tested in, settles every position automatically and in public.
What could kill this
A thesis that doesn’t examine its own weak points isn’t a thesis. Here’s where this one is genuinely exposed.
The Good Judgment Project’s results come from a structured, incentivized, multi-year research program, not casual retail bettors checking an app between trades. It’s strong evidence the mechanism can work. It is not proof it transfers cleanly to this context, and the honest position is that Chad is a bet on that transfer, not a guarantee of it.
Regulatory status has shifted before and can again; a product built on Kalshi’s API inherits Kalshi’s regulatory risk, not just its data.
The whole thesis depends on Kalshi's settlement data being timely and correct. A dispute, an outage, or a contract that settles ambiguously (combo markets awaiting review are a real, existing case) breaks the ledger the same way a corrupted record would break any other verification system.
A calibration score is a mirror, not an intervention. Some traders will look at it and improve. Some won't look at all. The mechanism is evidence-backed; the behavior change is not guaranteed for any individual user.
Falsifiable predictions
- 01
Calibration, not win-rate, becomes the standard metric serious U.S. prediction-market traders quote about themselves within three years, the same way Brier scores are already standard in professional forecasting.
- 02
At least one prediction-market source-verification product (Chad or a competitor) reaches meaningful U.S. adoption before a major platform builds the equivalent natively. Resolution data is public; the measurement layer on top of it is not yet built by anyone at scale.
- 03
FINRA or the SEC issues further formal guidance on finfluencer-driven trading before 2028, extending the pattern already visible in FINRA’s 2024–2026 research and enforcement activity.
The measurement, not the market
Prediction markets solved price discovery: a Kalshi contract prices an event in real time, continuously, verifiably. What they haven't solved is the trader. Every other measurable retail-trading crowd in the U.S. (stock-picking Discords, sports tout services, finfluencer feeds) runs on claims nobody can check, a pattern FINRA itself now documents. This one runs on a ledger nobody's reading yet.
The asymmetry is worth stating plainly: every claim in this paper can be verified today, from public settlement data and published U.S. regulatory and academic research, by anyone who bothers to look. Almost nobody has.
Start your own ledgerSources
- FINRA Investor Education Foundation, “Finfluencer Followers and Social Media Scrollers: The Profile, Patterns, and Pitfalls of Social-Media-Informed Retail Investors” (2026), drawing on the 2024 National Financial Capability Study.
- Baker, S., Balthrop, J., Johnson, M., Kotter, J., & Pisciotta, K., “Gambling Away Stability: Sports Betting’s Impact on Vulnerable Households,” Northwestern University (Kellogg) & Brigham Young University (March 2026).
- FINRA, “Day Trading: Your Dollars at Risk” (2020), FINRA.org.
- Clinton, J.D. & Huang, T., Vanderbilt University, prediction-market accuracy study, 2024 U.S. election cycle (December 2025).
- Good Judgment Project, Philip Tetlock & Barbara Mellers, University of Pennsylvania (Wharton School); research funded by IARPA. Track record updated through 2023.
- Bernstein, Macquarie, and Eilers & Krejcik Gaming 2030 prediction-market volume projections (2026); FalconX volume data (2026); Pew Research Center (2026); exchange volume reports and CFTC filings (2026).
This document is a statement of research opinion by Chad, provided for informational purposes only. It is not investment, legal, or tax advice, and nothing here should be read as a recommendation to trade any specific market. Prediction markets involve real financial risk; past patterns in cited research do not predict any individual’s future results. Figures are drawn from U.S. regulatory bodies and academic sources believed reliable but not independently re-verified by Chad.