Prediction Markets — Vol. 1, Issue 1
About this issue: Each issue of Prediction Markets combines Claude-written synthesis with curated external sources. This is the landscape primer; future issues will track regulatory shifts, platform developments, notable markets, and new academic research as they arrive.
Introduction
Temerity Holdings trades actively on Kalshi. That activity is the lens through which this primer is written: not a survey of prediction markets in the abstract, but an orientation to the industry as it exists in May 2026 — a world that looks very different from the one that existed eighteen months ago.
In October 2024, the D.C. Circuit denied the CFTC’s emergency motion to stay a district court ruling that had let Kalshi list contracts on U.S. elections. In May 2025, the Commission voluntarily abandoned the appeal. In July 2025, the Department of Justice and the CFTC closed their separate investigations of Polymarket without charges; the same month, Polymarket acquired the small CFTC-licensed exchange QCEX for $112 million as its regulated U.S. on-ramp. By April 2026, the Third Circuit had ruled — 2 to 1 — that the Commodity Exchange Act preempts state gaming regulators from interfering with sports event contracts on federally regulated exchanges. None of this was inevitable. All of it has happened.
This issue covers five sections, in the order a trader thinks about the space: how we arrived here (history and regulation), where one can actually trade (platform ecosystem), whether the prices mean what they appear to mean (pricing efficiency), where edge actually comes from (strategy research), and what is happening right now (current developments). It is meant to be returned to when a term appears in a brief, a competitor surfaces, or a court rules on something that affects how Temerity does business. The next issue will track only what has changed since this one.
In This Issue
| Section | Sources |
|---|---|
| History & Regulatory Landscape | 6 |
| Platform Ecosystem | 7 |
| Pricing Efficiency & Known Biases | 7 |
| Strategy Research | 7 |
| Current Developments | 7 |
History & Regulatory Landscape
Synthesis by Claude Sonnet — May 2026
The frame to start with is this: betting on outcomes is not new in the United States, and the regulatory category we now call “prediction markets” is in some sense a return rather than an invention. From 1868 to 1940, organized betting on U.S. presidential elections ran out of New York’s Curb Exchange and downtown brokerages. Volume was enormous — the 1916 election alone saw an estimated $165 million wagered through formal markets, or roughly $3 billion in today’s dollars. Election odds were published in The New York Times, The Wall Street Journal, and the New York Sun. The markets were accurate; they were also openly understood as a way for political insiders to express conviction with capital. They faded through a combination of forces: the rise of scientific polling displaced their information-aggregation function, state governments progressively criminalized commercial bookmaking, and wartime mobilization suspended organized wagering in the 1940s. Regulation played its part — but there was no single decisive prohibition, just a slow displacement by competing forces. The Rhode and Strumpf historical record is the corrective to any narrative that treats today’s event contracts as a novel asset class — they are a recovered one.
The modern prediction-market era began in 1988 with the Iowa Electronic Markets, a research project run by the University of Iowa under a no-action posture from the CFTC. IEM was the proof of concept: small dollars, academic carve-out, real prices. Its predictive accuracy in subsequent presidential cycles became the foundation of the academic literature reviewed in Section 3.
The first commercial attempt to scale was Intrade, an Irish company that by the late 2000s was offering binary contracts on hundreds of events from elections to celebrity outcomes. On November 26, 2012 — twenty days after Intrade had successfully predicted the 2012 election — the CFTC sued, alleging unregistered options activity. The Commission’s enforcement chief framed the theory plainly: “It is against the law to solicit U.S. persons to buy and sell commodity options, even if they are called ‘prediction’ contracts, unless they are listed for trading and traded on a CFTC-registered exchange or unless legally exempt.” By December, U.S. customers were told to withdraw funds; by March 10, 2013, Intrade had shuttered amid “financial irregularities” tied to roughly $1.5 million in suspect payments to founder John Delaney. Katherine Mangu-Ward’s contemporaneous Reason essay characterized the episode as regulatory overreach; the doctrinal point, however, has held. The legal theory that binary contracts on events are options, and that listing them in the U.S. requires designation as a registered exchange, has remained the regulator’s default lens ever since.
PredictIt followed a different path. In 2014, the CFTC’s Division of Market Oversight issued a no-action letter permitting Victoria University of Wellington to operate small-stakes political prediction markets in the U.S., subject to caps: 5,000 traders per market, $850 maximum position. The letter was withdrawn in August 2022 on allegations that PredictIt had operated as a de facto for-profit and expanded beyond political events. PredictIt sued; in Clarke v. CFTC, the Fifth Circuit broke from sister circuits to treat the staff no-action letter as the functional equivalent of a license, finding the withdrawal “arbitrary and capricious.” The Brookings analysis of the case is sharp on the second-order effect: the decision ironically harms regulated entities by incentivizing agencies to either eliminate no-action programs entirely, attach disclaimers that make letters less valuable, or provide less detailed analysis. The no-action regime that supported IEM and PredictIt is functionally over.
Kalshi took yet a third path: registration. In 2020, the CFTC designated KalshiEx as a designated contract market — the first DCM purpose-built for event contracts. The next four years were a slow negotiation over what kinds of events Kalshi could list; the Commission blocked Kalshi’s congressional control contracts in September 2023 on the theory that they “involve” gaming, a power asserted under Section 5c(c) of the Commodity Exchange Act. Kalshi sued. The U.S. District Court for D.C. ruled the agency had misread the statute. On October 2, 2024, a three-judge D.C. Circuit panel denied the CFTC’s stay motion, allowing Kalshi to list election contracts going into the November 2024 election. The opinion’s statutory analysis — that “involve” requires the contract to be one of the enumerated activities, not merely related to one — is the load-bearing precedent on which every subsequent platform now trades.
Polymarket’s regulatory history is shorter and uglier. On January 3, 2022, the CFTC issued a $1.4 million civil monetary penalty against Blockratize, Inc., d/b/a Polymarket, finding the platform had operated as an unregistered DCM and SEF, with its binary event contracts constituting swaps. Polymarket was required to wind down non-compliant markets; it pivoted to offshore-only access for U.S. residents and built the bulk of its 2024 election volume from foreign and crypto-native flow. The path back into the U.S. came not through litigation but through acquisition — the story Section 5 returns to.
Sources & Further Reading
| Title | Publication | ~Read | Why It’s Worth Reading |
|---|---|---|---|
| Historical Presidential Betting Markets | Rhode & Strumpf, Journal of Economic Perspectives (2004) | 40 min | The 1868–1940 record. Documents the scale of organized election betting on Wall Street and the gradual displacement by scientific polling. |
| Order: In the Matter of Polymarket (Docket 22-09) | CFTC (2022) | 20 min | The primary-source 2022 settlement — sets the legal theory (unregistered DCM/SEF, binary options as swaps) that still governs the space. |
| KalshiEx LLC v. CFTC, No. 24-5205 | D.C. Circuit (Oct 2024) | 25 min | The opinion that unlocked regulated U.S. election contracts; the statutory analysis of “involve” / “gaming” is what every subsequent platform now relies on. |
| The Death of Intrade | Katherine Mangu-Ward, Reason (2013) | 10 min | Post-mortem of the 2012 CFTC suit and Intrade’s collapse — the cautionary tale that shaped every subsequent platform’s compliance posture. |
| How PredictIt’s Legal Struggle Could Hamper Regulators | Connor Raso, Brookings | 15 min | The most credentialed walk-through of the 2014 PredictIt no-action letter, the 2022 withdrawal, and the Clarke v. CFTC second-order effects on the entire no-action regime. |
| Crackdown Forces Intrade to Drop Americans | The Regulatory Review (Penn Law) | 8 min | Contemporaneous legal-academic framing of the 2012 CFTC complaint; useful sibling to the Reason essay for the regulator’s theory. |
Go deeper: Rhode & Strumpf’s Historical Presidential Betting Markets is the single best correction to any “this is brand-new” framing of today’s event contracts. The KalshiEx v. CFTC D.C. Circuit opinion is the most consequential prediction-market document of the past decade and rewards a careful read of the statutory-interpretation reasoning.
Platform Ecosystem
Synthesis by Claude Sonnet — May 2026
Three platform archetypes matter today.
The CFTC-regulated DCM is the dominant U.S. retail venue. Kalshi sits at the center: a centralized matching engine, full CFTC rulebook, daily surveillance, position limits, and a fee schedule built on a single formula — for each contract traded, the taker pays the ceiling of 7¢ × C × (1 − C), where C is the contract price in dollars. The maximum fee is 1.75¢ per contract at the 50¢ midpoint; the minimum drops to under one cent at the tails. That curve is consequential: a trader who wants to express conviction near 90¢ pays meaningfully less per round-trip than one trading the noisier middle of the book, which is the inverse of what intuition would suggest and the same direction as informational efficiency would predict. As of mid-2025, Robinhood and Interactive Brokers route U.S. retail event-contract flow into Kalshi, and DraftKings and FanDuel have launched their own CFTC-regulated event-contract products under similar architectures.
Polymarket is the second archetype: an on-chain crypto-native venue. Its order book runs on Polygon; resolution is handled by UMA’s Optimistic Oracle. The mechanic is worth understanding precisely. After an event resolves, anyone may post a $750 pUSD bond (pUSD is Polymarket’s dollar-denominated collateral token on Polygon) and assert an outcome. A two-hour challenge window opens; if no one disputes, the assertion finalizes and the market pays out. If challenged, both parties post matching bonds; if challenged again, the dispute escalates to UMA’s Data Verification Mechanism, where token holders vote over a 24-to-48-hour evidence period followed by a similar voting period. Worst-case resolution takes four to six days. Four DVM outcomes are possible: proposer wins, disputer wins, “too early” (favoring disputer), or — rarely — a 50/50 split where each token redeems for $0.50. The system is permissionless, but the cost is that the contract text is the only authority. Polymarket’s own docs put it bluntly: “the market title describes the question, but the rules define how it resolves.” Polymarket charges zero trading fees, which makes it superficially the cheaper venue; the catch is that costs show up elsewhere — in gas, in stablecoin on/off ramps, and in the wider implicit spread market-makers demand to compensate for resolution risk.
A second platform-economics fact, surfaced cleanly by Paradigm’s December 2025 audit, is that headline Polymarket volume is approximately twice the true figure. Each trade emits an OrderFilled event for both the maker and taker side; naive sums double the underlying activity. Commonly cited 2024 election figures of roughly $2.5 billion in monthly volume should be read as roughly $1.25 billion. Major data providers — DefiLlama, Allium, Blockworks — have since updated their dashboards. Any number cited from an older source should be assumed inflated by 2×. The diagnostic test: if a chart shows October or November 2024 monthly volume in the $2-3 billion range, it is summing both event sides; the corrected number is half.
The third archetype is play-money and research. Manifold Markets is the cleanest example: real users, fake currency, public calibration data. Its calibration page shows the empirical relationship between market prices and resolved outcomes across hundreds of thousands of contracts; markets priced at 70% resolve YES approximately 70% of the time, with predictable degradation at the tails. Manifold has become the de facto laboratory for studying market mechanism design, since researchers can introduce experimental contracts without regulatory friction. The Iowa Electronic Markets, the original research venue, remains live under its 1988 academic carve-out but trades at vanishingly small volumes.
Two derivative venues round out the landscape. PredictIt, post-Clarke v. CFTC, continues to operate under a court-ordered carve-out, though its position caps of $850 per market severely limit its usefulness for any non-trivial position. The decentralized prediction-market projects of the late 2010s — Augur, Gnosis Conditional Tokens, Omen — survive at the architectural level (much of UMA’s design borrows from them) but no longer host meaningful liquidity. Polymarket’s success on Polygon effectively absorbed the demand they were built to capture.
For honest cross-venue comparison, the Paradigm Prediction Markets dashboard is the cleanest free reference; it normalizes for the volume-double-counting issue and aggregates depth, open interest, and resolved-market accuracy across Kalshi, Polymarket, and Manifold. Pair it with Manifold’s calibration page for the accuracy question and a Kalshi-internal fee calculation for the cost question, and one has the three numbers a trader needs before committing capital to any venue.
Sources & Further Reading
| Title | Publication | ~Read | Why It’s Worth Reading |
|---|---|---|---|
| KalshiEX LLC Rulebook (v1.18, July 2025) | CFTC / Kalshi | 60 min (reference) | The authoritative DCM rulebook — market structure, settlement, surveillance, member obligations. Reference rather than read-through. |
| Kalshi Fee Schedule | Kalshi | 5 min | The canonical taker/maker formula (7¢ × C × (1 − C)); essential for any cost-of-trading comparison. |
| Resolution (UMA Optimistic Oracle) | Polymarket Docs | 15 min | Official spec for on-chain settlement — the single most load-bearing mechanism on the platform. |
| Polymarket Volume Is Being Double-Counted | Storm Slivkoff, Paradigm (Dec 2025) | 10 min | Required reading before quoting any Polymarket volume statistic; explains the OrderFilled double-emission and how to compute corrected one-sided volume. |
| Paradigm Prediction Markets Dashboard | Paradigm | 10 min (interactive) | Live cross-venue volume / open-interest dashboard from a serious quant shop; the cleanest free comparison data. |
| Platform Calibration | Manifold Markets | 5 min | Live calibration curve and Brier-score data — the empirical anchor for “do these markets actually work.” |
| Prediction Markets — Everything You Need to Know (Tabarrok & Kominers) | a16z crypto | 45 min listen (optional) | Background reference. Strongest free survey of mechanism-design tradeoffs and the AI-oracle direction. Not required reading. |
Go deeper: Paradigm’s Polymarket Volume Is Being Double-Counted is the single most important platform-data correction of the past year. Pair it with the Polymarket Resolution doc to understand the full settlement pipeline that produces the volumes everyone is now learning to read correctly.
Pricing Efficiency & Known Biases
Synthesis by Claude Sonnet — May 2026
Twenty years of academic literature on prediction markets converges on a defensible-but-narrow claim: at sufficient liquidity and short horizons, prices behave as reasonable probability estimates, and the markets in aggregate beat alternative forecasting methods. The most-cited summary remains Wolfers and Zitzewitz’s 2008 Palgrave entry, which surveys evidence across the Iowa Electronic Markets, Tradesports, and corporate prediction markets at Hewlett-Packard and Eli Lilly. Their conclusion — that prices on liquid contracts approximate the underlying probability with calibration errors generally under five percentage points — has been replicated repeatedly. Page and Clemen’s 2013 Economic Journal study refined the picture: short-horizon contracts (resolving within days to weeks) are well-calibrated; long-horizon contracts (resolving in months to years) systematically over-predict change relative to status quo. The distortion appears to come from a combination of trader impatience, opportunity-cost-of-capital effects on far-dated contracts, and a thin marginal supply of patient capital willing to underwrite long-tail mispricings.
The clearest documented bias is the favorite-longshot effect: long-odds contracts trade at prices that imply higher win probability than realized outcomes warrant. Snowberg and Wolfers’s 2010 decomposition, using horse-track data covering millions of races, attributes the bias roughly 80% to misperceptions of small probabilities and 20% to risk preferences. The behavioral mechanism dominates the utility mechanism, which is the more pessimistic answer for arbitrageurs: if the bias were primarily risk preference, taking the other side would compensate a rational arbitrageur for risk; if it is primarily misperception, the bias is exploitable but capital-constrained.
The question that matters in 2026 is whether favorite-longshot bias survives in a regulated, transparent venue with sophisticated market-makers. The May 2026 working paper Makers or Takers: The Economics of the Kalshi Prediction Market from the George Washington University Forecasters Forum analyzes contract-level data from Kalshi and answers yes: the bias persists. Low-probability outcomes are systematically overpriced; high-probability outcomes underpriced. The magnitude attenuates in high-volume contracts but does not disappear. The paper’s secondary contribution is a clean maker-vs-taker decomposition: makers — passive liquidity providers — earn positive expected returns gross of fees; takers earn negative expected returns. Once Kalshi’s 7¢ × C × (1 − C) fee schedule is applied, taker-side strategies require substantial directional edge to clear costs, while maker-side strategies retain positive expectancy in mid-volume contracts. The fee curve concentrates pain at the 50¢ midpoint, which is exactly where the favorite-longshot bias is least pronounced and the maker edge is most contested.
Manipulation has been studied most carefully by Rhode and Strumpf, whose 2008 paper combined a century of observational data on attempted market manipulations with a field experiment on the Iowa Electronic Markets. The pattern is consistent: large one-time orders move prices temporarily but revert within hours as informed capital reasserts. Sustained manipulation requires sustained capital, and once arbitrageurs identify the source, the cost-to-move-price ratio rises sharply. The implication for fund operators is that liquid markets are difficult to manipulate at scale, but illiquid markets — particularly short-fuse contracts on niche outcomes — remain vulnerable, especially in the final hours before resolution when capital cannot easily be redeployed.
The newest empirical anchor is a July 2025 arXiv paper comparing Polymarket prices to polling aggregators for the 2024 U.S. presidential election. The authors find Polymarket was superior to polling at predicting the swing-state outcomes, framed within Wisdom-of-Crowds theory: aggregated financial incentives produced more accurate forecasts than aggregated stated preferences. The paper does not adjudicate why — the candidate explanations include the French-whale-style information edge discussed in Section 4, social-desirability bias in polls that markets effectively price out, and selection effects in which voters are willing to be polled. The result is suggestive rather than conclusive, but it is the first peer-reviewed-quality evidence that liquid prediction markets out-predicted modern polling in a U.S. election with a non-trivial state-by-state sample.
Scott Alexander’s Prediction Market FAQ is the operator-facing complement to the academic literature, and the source most worth reading carefully on why mispricings persist. His framework runs as follows. First, capital and opportunity costs: correcting a small absolute mispricing requires tying up capital that could earn the equity-market return elsewhere, and for far-dated contracts the comparison is brutal. Second, regulatory caps that prevent any single arbitrageur from clearing a bias — PredictIt’s $850 position limit is his canonical example. Third, transaction costs that consume modest edges, particularly at the taker level. Fourth, counterparty and resolution risk premia: when participants assign even a 10–15% probability that the market “will go bust, steal your money, or wrongly resolve the question against you,” the math on correcting a 10% mispricing turns negative. Fifth, attention gaps: small markets fall below the monitoring threshold of sophisticated capital. The Kalshi paper’s finding that favorite-longshot bias survives in a regulated venue with deep liquidity is consistent with Alexander’s frame — the bias is, at most prices, too small in absolute dollar terms to attract the marginal capital that would close it. Mispricings persist not because the market is broken but because the friction costs of correction are real and quantifiable.
Sources & Further Reading
| Title | Publication | ~Read | Why It’s Worth Reading |
|---|---|---|---|
| Prediction Markets (Palgrave) | Wolfers & Zitzewitz (2008) | 45 min | The canonical primer; defines accuracy benchmarks, contract types, and known biases. Every other source builds on it. |
| Do Prediction Markets Produce Well-Calibrated Probability Forecasts? | Page & Clemen, Economic Journal (2013) | 40 min | Empirical calibration test on a large prediction-market dataset; short-horizon markets well-calibrated, long-horizon systematically biased. |
| Explaining the Favorite-Longshot Bias | Snowberg & Wolfers, NBER WP 15923 (2010) | 35 min | Cleanest decomposition of favorite-longshot bias — ~80% misperception, ~20% risk preferences. |
| Manipulating Political Stock Markets | Rhode & Strumpf (2008) | 50 min | Empirical baseline on manipulation cost and durability: prices revert within hours; sustained manipulation requires sustained capital. |
| Makers or Takers: The Economics of the Kalshi Prediction Market | GWU Forecasters Forum WP 2026-001 | 40 min | The most current Kalshi-specific empirical work: favorite-longshot bias survives in a regulated venue, makers outperform takers gross of fees, the fee curve compresses taker edge. |
| Are Betting Markets Better than Polling in Predicting Political Elections? | arXiv (July 2025) | 25 min | Polymarket vs. polling aggregators in 2024 swing states; the rigorous answer to “did the markets actually beat the polls?” |
| Prediction Market FAQ | Scott Alexander, Astral Codex Ten | 25 min | The operator-facing lens on why mispricings persist — capital costs, position caps, fees, counterparty risk, attention gaps. |
Go deeper: Page & Clemen for the calibration question, Makers or Takers for the current Kalshi evidence. Alexander’s FAQ is the cheapest read in the slate and the one most likely to change how a trader thinks about why an obvious mispricing is still there.
Strategy Research
Synthesis by Claude Sonnet — May 2026
For a fund already trading these venues, the toolkit splits into two categories. The first two are analytical foundations — frameworks that shape how to think about event contracts. The second two are operational edge sources — strategies that identify where profit actually comes from.
The first is mechanism understanding. Event contracts are digital (binary) options, and the mathematics inherited from the FX and equity-exotics literature applies cleanly. The cleanest replication is the call-spread limit: a digital paying $1 if the underlying ends above a strike is the limit, as spread , of a long call at and short call at , scaled by . This view illuminates two operational hazards. First, near expiry, delta spikes — for a contract priced near 50¢, an infinitesimal move in the underlying probability can move the contract from 51¢ to 49¢ or vice-versa, producing leverage that does not appear in the pre-trade Greeks. Second, vega becomes negative near the at-the-money boundary as implied volatility rises, because higher dispersion in the underlying distribution pulls probability mass away from the digital’s strike. These are not exotic edge cases; they are the everyday risk profile of any near-expiry election or sports contract held into resolution. Pin risk — the discrete jump from $1 to $0 across the resolution boundary — is the practical manifestation, and disciplined operators either close before resolution or size as if the position were fully exposed.
The second is reasoning in log-odds space. A contract priced at 91¢ and one priced at 99¢ are not 8 percentage points apart in any operationally meaningful sense; they are roughly and in log-likelihood-ratio space — a factor of two in evidence. Bayesian updating, expressed in log-odds, becomes additive: posterior log-odds equal prior log-odds plus the log-likelihood ratio of the new evidence. This collapses what otherwise requires computing into simple arithmetic, and — more importantly for sizing — it spreads out the extremes in a way that matches the actual information content of the prices. Kelly sizing follows naturally: bet fractions are linear in log-odds advantage, not probability advantage. The practical implication is that a trader who reasons in probabilities will systematically under-react to mispricings at the tails (treating 92¢ and 94¢ as nearly identical) and over-react to mispricings in the middle (treating 48¢ and 52¢ as substantially different). Both errors compound over many trades. The Bawa essay translates this into Kalshi/Polymarket-specific Kelly formulas and a worked example — at market price 60¢ with a 75% fair value, the full-Kelly fraction is 37.5% of bankroll; he recommends fractional Kelly at 0.25–0.5× given that full Kelly has roughly a one-third probability of halving the bankroll before doubling it.
The third is market-making and rebate capture. Polymarket publishes its liquidity-rewards formula explicitly: a maker’s score , where is the maximum eligible spread from the adjusted midpoint, is the maker’s actual spread, and is an in-game multiplier. Tighter spreads earn quadratically more; two-sided depth is rewarded with a multiplier that single-sided makers cannot capture; orders below ten cents or above ninety cents require strict two-sidedness to score at all. The April 2026 sports-market rewards pool exceeded $5 million, distributed across NFL ($7,700/game), NBA, Champions League ($24,000/game), EPL ($10,000/game), and CS2 contracts. The opportunity is real and quantifiable — but the systematic sports market-maker profiled on Polymarket’s own Oracle blog makes clear that it is not easy. He runs an automated system over several hundred gigabytes of historical data, deploys up to $300,000 across NFL Sundays, and reports performance volatility that includes a $100,000 drawdown over ten days in hockey season and an $80,000 single-game win on the Super Bowl. His operational philosophy is to avoid being clicked — to be the maker who has updated his quotes before injury news propagates rather than the maker whose quotes are stale when the news breaks. Hanson’s foundational 2003 LMSR paper established the theoretical basis for automated market maker (AMM) designs used in early on-chain prediction markets — including Polymarket’s original architecture before its 2023 migration to a central limit order book (CLOB). Today all major venues run CLOB structures, so the LMSR paper is historical context rather than operational blueprint; it remains worth understanding for anyone reasoning about why CLOBs displaced AMMs in high-volume markets.
The fourth is information edge. The clearest 2024 example is the French trader known publicly as Théo or Fredi9999, who cleared more than $80 million across multiple Polymarket accounts on Trump’s 2024 election. The edge was not insider information — it was a methodological insight: he commissioned private YouGov polls using neighbor prompts (“who do you think your neighbors are voting for?”) rather than direct-preference prompts, on the hypothesis that social-desirability bias was depressing reported Trump support in Pennsylvania, Michigan, and Wisconsin. The neighbor data showed substantial Trump margins where direct polling showed near-ties. The market disagreed for several weeks; by election night, it did not. Polymarket CEO Shayne Coplan framed the incentive structure exactly: “If this guy was not able to make over $80 million but rather able to make $80,000, he would have never gone through the hassle.” Information edge in prediction markets requires the same thing it requires elsewhere — a non-consensus hypothesis, a data-acquisition strategy that cannot be replicated cheaply, and the conviction to size into it. The size of the prize justifies the cost of the research, but only above a threshold.
Sources & Further Reading
| Title | Publication | ~Read | Why It’s Worth Reading |
|---|---|---|---|
| Logarithmic Market Scoring Rules | Robin Hanson | 45 min | The foundational LMSR paper — historical context for the AMM-based designs that preceded today’s CLOB venues; explains why CLOBs eventually won. |
| Binary Options: Pricing, Replication and Skew Sensitivity | Quant Next | 20 min | Rigorous derivation of digital payoff as the limit of a call spread; delta spike near expiry, skew sensitivity, replication-based hedging. Directly maps to pin risk on Kalshi/Polymarket. |
| Log-Odds Are Better Than Probabilities | Robert Huben, LessWrong | 10 min | The conceptual scaffolding for additive Bayesian updates, log-loss, Kelly sizing, and stable sizing on event contracts. |
| Meet Your Market Maker | The Oracle by Polymarket | 10 min | First-person profile of a systematic sports market-maker — what it actually costs to capture rebate edge. |
| Liquidity Rewards (market-maker docs) | Polymarket Documentation | 10 min | The exact rewards formula and the April 2026 pool sizes. Required to model rebate-based MM strategies. |
| How a French “whale” made over $80M on Polymarket | CBS News / 60 Minutes | 8 min | Best free write-up of the Théo / Fredi9999 trade — concrete case study of an information-edge play executed at scale. |
| The Math of Prediction Markets | Navnoor Bawa | 15 min | Practitioner-grade walkthrough tying Kelly sizing to CLOB mechanics on Polymarket and Kalshi. Worked examples. |
Go deeper: Meet Your Market Maker is the most operationally honest piece in the entire slate — read it before deploying capital into rebate strategies. The Huben essay on log-odds is short, foundational, and will change the way Greeks and Kelly sizing are evaluated.
Current Developments
Synthesis by Claude Sonnet — May 2026
The 2025–2026 arc has three threads: federal-level consolidation, platform consolidation, and a state-by-state preemption fight that is not yet resolved.
At the federal level, the direction has been unambiguous. In May 2025, the Commodity Futures Trading Commission voluntarily dropped its appeal of KalshiEx v. CFTC, leaving the district court’s holding — that the Commission had exceeded its authority in blocking Kalshi’s election contracts — as the controlling precedent. That single decision converted election event contracts from a contested category to a permanently lawful one absent new legislation. In July 2025, the Department of Justice and the CFTC both closed their separate investigations of Polymarket’s founder Shayne Coplan and the platform itself, ending the FBI-raid saga that had clouded the company’s U.S. re-entry plans. The same month, Polymarket announced its acquisition of QCEX — comprising the CFTC-licensed exchange QCX, LLC and clearinghouse QC Clearing LLC — for $112 million in cash. Coplan framed it as “demand is greater than ever — not just in user growth and trading volume, but in how mainstream audiences are turning to Polymarket to separate signal from noise”; the regulatory translation is that Polymarket bought rather than built its U.S. regulatory infrastructure, compressing what would have been a multi-year DCM application into a one-step transaction. The combined entity reported roughly $6 billion in predictions made on the platform in the first half of 2025. Polymarket’s regulated U.S. product launched in late 2025 alongside an integration with X (formerly Twitter).
In March 2026, the CFTC issued an Advance Notice of Proposed Rulemaking on prediction markets — the formal start of a rulemaking process that, when complete, will codify the boundary between event contracts and “gaming” or “sports competition” under Section 5c(c). The same month, the Commission’s new Director of Enforcement, David Miller, gave a high-signal speech at NYU Law shifting the agency away from “regulation by enforcement” and identifying five priority areas, with insider trading on event contracts named explicitly. The misappropriation theory under Section 6(c)(1) of the Commodity Exchange Act applies directly: a congressional staffer who trades on non-public legislative outcomes, a team trainer who trades on undisclosed injuries, a YouTuber who trades on content not yet released. The “Eddie Murphy Rule” — barring federal employees from trading on confidential government information — extends to event contracts. Miller also introduced a new declination framework that requires self-reporting, full cooperation, and complete remediation as prerequisites — his framing was that cooperation is binary, “like jumping into a lake; you’re either in a hundred percent or you’re out.” For any institutional participant, information barriers and pre-trade compliance review are now standard expectations.
The state-by-state picture, as of mid-May 2026, is the live battlefield. The April 7, 2026 Third Circuit decision in NJ v. Kalshi — a 2-1 panel — held that the Commodity Exchange Act grants the CFTC exclusive jurisdiction over sports event contracts on federally regulated exchanges, neutralizing New Jersey’s cease-and-desist letters. The court rejected New Jersey’s “strong presumption against preemption” argument in traditionally state-regulated gambling areas, finding sports event contracts qualify as swaps under federal law when traded on CFTC-licensed markets. The dissent identified the central tension — that Kalshi’s sports offerings are “virtually indistinguishable” from products on DraftKings and FanDuel sportsbooks — and was supported by amicus briefs from 34 states plus D.C. and the Northern Mariana Islands. The preemption question will likely reach the Supreme Court.
The DOJ and CFTC have responded by filing affirmative preemption suits against Illinois, Connecticut, and Arizona — the three states that had pressed hardest on prediction-market operators. Illinois issued cease-and-desist letters to Kalshi and Polymarket US, alleging unlicensed gambling. Connecticut characterized sports event contracts as unlicensed wagering and alleged violations of its Unfair Trade Practices Act. Arizona escalated furthest, filing a 20-count criminal information against KalshiEx in March 2026 that includes four counts of unlawful election wagering — the only state to bring criminal charges against a CFTC-regulated exchange. Nevada, Maryland, Tennessee, Michigan, Washington, and Massachusetts have issued lower-temperature challenges that vary in posture from cease-and-desist letters to ongoing review. The Norton Rose Fulbright synthesis is the cleanest legal summary; for case-by-case granularity, the Lines.com state-by-state breakdown and the Action Network lawsuit tracker are the strongest free trackers — both are sports-betting trade press, which is the only place a true state-by-state map is published in free form.
Two adjacent developments deserve flagging. First, sportsbook entry: DraftKings Predictions and FanDuel Predicts have launched CFTC-regulated event-contract products in selected states, leveraging the same legal foundation as Kalshi. The competitive implication is real — the major sportsbooks bring orders-of-magnitude more retail flow than Kalshi or Polymarket have ever seen — and the substantive product overlap is what motivates the dissent in NJ v. Kalshi. Second, the no-action regime has continued to deteriorate under Clarke v. CFTC’s long shadow. PredictIt operates under a court-ordered carve-out rather than agency discretion; the Commission has not granted a new no-action letter to a prediction-market operator since 2014, and is unlikely to.
The strategic implication for Temerity is that the regulatory environment for U.S. event-contract trading is the most favorable it has been since Intrade’s 2012 collapse, but the favorability is concentrated at the federal level and contested at the state level. Capital deployment that depends on Kalshi continuing to operate as a CFTC-regulated DCM with election and sports contracts is well-protected. Capital deployment that depends on any specific state’s posture — particularly in Arizona, Illinois, Connecticut, or New Jersey — is exposed to the preemption fight that the next twelve months will adjudicate.
Sources & Further Reading
| Title | Publication | ~Read | Why It’s Worth Reading |
|---|---|---|---|
| NJ cannot regulate Kalshi’s prediction market, U.S. appeals court rules | CNBC (April 2026) | 5 min | The Third Circuit’s preemption ruling — neutralizes the state cease-and-desist wave for sports event contracts. |
| CFTC drops Kalshi election bet case appeal | CNBC (May 2025) | 4 min | Trump-era CFTC voluntarily dismissed the appeal; election contracts permanently lawful absent new legislation. |
| Polymarket investigations ended by DOJ, CFTC | CNBC (July 2025) | 4 min | The declination notices that unlocked Polymarket’s U.S. re-entry path. |
| Polymarket Acquires CFTC-Licensed Exchange and Clearinghouse QCEX for $112M | PR Newswire / Polymarket | 3 min | Primary-source announcement of the July 2025 deal that became Polymarket’s regulated U.S. on-ramp. |
| Prediction markets at a crossroads: Preemption, enforcement and rulemaking | Norton Rose Fulbright | 12 min | Big-law synthesis of the 2025–2026 landscape: CFTC posture, state preemption fights, sports-contract litigation, rulemaking outlook. The single best one-stop primer for a fund operator. |
| U.S. Prediction Market Legal Status: State-by-State Breakdown for 2026 | Lines.com | 34 min | The strongest free state-by-state map (updated April 2026): per-state risk levels, cease-and-desist dates, AG actions. Sports-betting trade press but no better free alternative exists. |
| Tracking Every Prediction Market Lawsuit Involving Kalshi, Polymarket & Others | Action Network | 15 min | Case-by-case companion: 30+ active suits with court status and next hearing dates across Kalshi, Polymarket, Robinhood, Coinbase. |
Go deeper: Norton Rose Fulbright’s Prediction markets at a crossroads is the most efficient single read on the federal landscape; pair with the Lines.com state map for the territorial picture. If only one is read, choose the Norton Rose piece.
My Notes
Added after reading