Prediction Markets Trading at Quant Firms

Two things changed at once: the legal path opened, and the volume arrived.

Prediction markets in the US sit under the Commodity Futures Trading Commission (CFTC) as event contracts, a type of binary derivative, rather than under state gambling law. The pivotal fight was Kalshi versus the CFTC over election contracts:

1. 1
   September 2023. The CFTC orders Kalshi not to list congressional control contracts, a majority of commissioners calling them akin to gaming.
2. 2
   September 2024. A US District Judge grants Kalshi summary judgment, finding the CFTC exceeded its authority. The DC Circuit briefly stays the order.
3. 3
   October 2024. The DC Circuit denies the CFTC's emergency stay and Kalshi's election contracts go live. More than 500 million election contracts trade on Robinhood within a week of its presidential contract launch.
4. 4
   May 2025. The CFTC voluntarily drops its appeal, ending the case in Kalshi's favour.
5. 5
   2025 onward. Kalshi self-certifies sports contracts, opening a second legal front with state gaming regulators, which remains live today.

Once elections were cleared, the volume followed. Kalshi reported $23.8 billion of notional volume in 2025, up 1,108% year on year across 97 million transactions. December 2025 alone set a monthly record of $6.38 billion. Combined Kalshi and Polymarket volume then reached a record $44 billion in June 2026, nearly double May's total, driven by the NBA Finals and the World Cup. Sports dominate: roughly 80% of Kalshi's volume, with its World Cup contracts alone generating around $27 billion.

Why do market makers care? The economics are the familiar market making economics, adapted to binaries. Exchanges court professional liquidity with reduced fees and higher position limits: Susquehanna became Kalshi's first official market maker on exactly those terms. Interactive Brokers' ForecastEx pays an incentive coupon, linked to the Fed funds rate, on capital tied up in positions. And the fee structure itself shapes where quoting is profitable:

The structural attraction, in the words of one firm, is a market defined by inefficiency, fragmentation and the absence of mature pricing, which is exactly where quant strategies historically thrive.

The regulatory opening produced a land grab for the regulated stack: a designated contract market (DCM) plus a derivatives clearing organisation (DCO).

The evidence divides cleanly into committed participants, hedgers, and firms that have deliberately stayed out.

Susquehanna deserves specific attention. SIG is culturally a probabilities-and-options firm with deep roots in sports analytics, and it treats event contracts as an asset class, not a novelty. Its head of prediction markets has described the firm as risk-neutral by default but willing to hold positions to provide hedges to clients, and its sports desk trades live in-game markets, correlated combinations and player-level props. It is the single best example of the thesis that an options market maker's skill set transfers directly to binaries. Separately, Susquehanna partnered with Robinhood to acquire MIAXdx as a DCM and DCO for a joint prediction markets exchange. If you are targeting the firm, our SIG interview guide covers the process.

The firms that stayed out matter for an honest assessment. As of 2026, Citadel Securities, IMC and Hudson River Trading were reported as having stayed away, citing lack of regulatory clarity and volumes that are still light relative to their core markets. Citadel Securities has since softened in public comments, while making clear it is not looking at sports. The reputational and gambling-classification risk is a real reason serious firms hesitate.

A specialist ecosystem is forming alongside the giants. Crypto-native market maker Wintermute, which processes more than $3.5 trillion in annual volume across 70-plus venues, confirmed in May 2026 that it quotes two-sided markets on both Kalshi and Polymarket. Small specialist funds in New York, Chicago and Geneva have been named as hiring, though most are thinly documented beyond a single round of reporting. The more interesting structural story is professional sports betting syndicates crossing over: the financial firms bring capital and execution infrastructure but often lack specialist sports pricing models, while the syndicates have spent decades pricing sport and modelling correlation. Kalshi's launch of combination contracts (parlays) in September 2025 is the specific product that pulled them in, because pricing correlated multi-leg risk is exactly their expertise. The crossover runs both ways: Jump hired a researcher who built sports betting models in his own time while working at a Big Four accounting firm.

The postings are concrete and public. The core role types:

Representative postings, with real detail:

• DRW, Prediction Markets Trader. A dedicated desk focused on Kalshi and Polymarket. Strategies listed: market making with dynamic skew, order-flow and book-imbalance microstructure plays, cross-platform arbitrage, sub-second news momentum, and statistical pairs. Base salary $175,000 to $200,000 plus discretionary bonus. The process runs a technical screen, a strategy discussion, and a live coding and modelling exercise on real prediction market scenarios.
• DRW, Software Engineer, Prediction Markets (Python). Base $150,000 to $225,000, owning pricing pipelines, cross-venue reconciliation and risk.
• Akuna Capital, Junior Quantitative Researcher, Prediction Markets (Chicago). Sports pricing models, data capture pipelines, in-play quoting. Minimum base $145,000 plus bonus, with prior trading-firm experience "valued but not necessary", and an explicit preference for people who have been trading Kalshi or Polymarket as a personal project.
• Susquehanna is recruiting to detect incorrect fair values and identify inefficiencies, plus a dedicated sports trader.

Compensation. The documented base ranges ($145,000 to $225,000) sit below top-tier mainstream quant trading new-grad total compensation, which runs roughly $250,000 to $450,000-plus at the very top firms per self-reported data. But those are base figures with undisclosed discretionary bonuses on top, so the comparison is not clean, and total comp for these specific desks is simply not publicly documented. One quant recruiter characterised 2026 hiring as selective and exploratory rather than large-scale desk building. Be sceptical of any precise total-comp number for this niche.

Geography. Chicago (DRW, Akuna, Jump), New York (the native funds), the Philadelphia region (Susquehanna), London (Wintermute), Amsterdam (Flow Traders) and Singapore. Several roles are explicitly remote or flexible, which is unusual for trading and reflects how new and small these desks are.

This is what separates a credible interview answer from a naive one. Each contract settles at $1 if the outcome happens and $0 if it does not, so a price of 40 cents is the market's implied 40% probability. That changes the geometry of trading:

• The payoff is bounded. Your maximum loss and gain per contract are both capped, unlike a linear instrument.
• Spreads and risk live in probability space. A 1 cent spread is a 1 percentage point spread in probability, which is enormous near the middle and trivial near the extremes.
• Quoting near 0 and 1 is dangerous. Selling a contract at 2 cents collects 2 cents to risk 98. You are writing deep out-of-the-money optionality, and one surprise wipes out hundreds of correct trades. This is the practical face of the favourite-longshot bias.
• There is often no continuous hedge. Many event contracts have no correlated liquid instrument to delta-hedge against, so inventory risk is carried all the way to resolution. Where a hedge exists, it is a different asset with basis risk.
• Adverse selection is acute. In a market about a real-world event, the person lifting your offer may simply know something you do not. The insider trading cases covered below are the extreme version of the everyday adverse-selection problem a maker faces.

Most regulated US event contracts are also fully collateralised: every position is backed dollar for dollar with no margin, which is capital-intensive relative to futures and a real drag on returns. That is starting to change now that Kalshi has approval for an affiliate to operate as a Futures Commission Merchant, the first step toward margin.

The bread and butter of the desks is relative value, not outcome prediction:

• Cross-venue arbitrage. The same event priced differently on Kalshi and Polymarket, or versus a sportsbook line. Buy Yes on the cheaper venue and No on the other so the two legs cost less than $1. The frictions are real: you must pre-fund both venues, fees eat thin edges, and settlement risk is the killer, because if the two venues resolve the same event differently you lose both legs.
• Mutually exclusive sets that must sum to 1. Where exactly one outcome resolves Yes, the Yes asks across all outcomes should sum to about 100 cents. If they sum to less, buy the whole field and lock a profit; if the bids sum to more, sell the field.
• Arbitrage against traditional instruments. CPI contracts against inflation swaps or TIPS breakevens; Fed decision contracts against fed funds futures and OIS; sports contracts against sportsbook lines. This is where the incumbents' existing books give them a structural advantage.
• The longshot bias. Systematic overpricing of low-probability contracts is a documented, persistent edge source.

The calibration evidence is worth quoting accurately in an interview. Prediction markets are generally well calibrated, with typical Brier scores in the 0.15 to 0.25 range, better than expert panels. They are most efficient in the middle and at the high-probability end, and least efficient in the low-probability tail: a study of Kalshi unemployment contracts found contracts priced below 30 cents significantly overpriced, with an actual win rate of about 4%. Efficiency also varies by category, with Fed and interest-rate markets near-perfectly calibrated and employment markets the weakest. The practical implication: respect the base rate, hunt the low-probability tail and thin markets for edge, and track your own Brier score against the market's as a feedback loop.

A decision-ready view has to weigh the bear case fairly.

• Regulatory risk is the dominant risk. State gaming regulators in New Jersey, Nevada, Maryland, Ohio, New York and elsewhere, plus tribal gaming interests, are actively litigating whether sports event contracts are unlicensed betting. Kalshi has won important early rounds on CFTC preemption, including preliminary injunctions in Nevada and New Jersey and a November 2025 win against a California tribal challenge, but several cases remain live and a CFTC rule change is possible. If sports contracts are curtailed, roughly 80% of current volume is at risk.
• Market integrity. Insider trading on event contracts is not hypothetical: a US Army master sergeant was charged with using classified information to profit on a contract about the capture of Venezuela's Maduro, the first US case of its kind, and Kalshi has disclosed opening around 200 investigations. This is both a reputational risk and a live adverse-selection cost for market makers.
• Durable asset class or hype cycle? The bull case: real hedging use cases, genuine information aggregation, and serious institutional infrastructure commitments (ICE's investment of up to $2 billion in Polymarket, CME and FanDuel launching, DraftKings buying Railbird). The bear case: volume is dominated by sports and concentrated in a small active cohort, around 75% of Kalshi users reportedly never trade, notional volume overstates real capital, and the markets are still too thin to absorb institutional size. The honest read is that market making and relative value are already viable at modest scale, but the "serious institutional asset class" thesis is not yet proven.

Build the general quant foundation first: probability, statistics and market making are the gate, and they transfer regardless of what happens to prediction markets. Then add the niche fluency, because firms explicitly value it and it is the cheapest differentiator available:

1. Trade small on Kalshi or Polymarket to understand fills, fees and resolution mechanics first-hand.
2. Build a calibration tracker. Log your own forecasts, compute your Brier score, and compare it against the market's. This is exactly the feedback loop the desks run.
3. Build a small cross-venue scraper and fair-value model. Even a simple one that flags sum-to-one violations demonstrates you understand the actual edge sources.
4. Put the project on your CV. Akuna and DRW both explicitly ask for personal prediction markets experience.

The honest downside: the skill transfer is narrower than a mainstream seat if you over-specialise, the desks are small and could be scaled back if regulation turns, the gambling stigma is real, and total compensation for the niche is undocumented and probably below top mainstream seats today. The upside: it is a young, inefficient market where a junior can have outsized impact, the intellectual content is rich, and the core skills transfer if you keep them general.

A sensible framing is to target a firm with both a mainstream book and a prediction markets desk, so you are not betting your whole career on one asset class. Susquehanna, Jump, DRW and Akuna all fit that description, and in the interview, talk about prediction markets the way the desks do: arbitrage, microstructure, calibration and risk, not predicting outcomes.

Are prediction markets just gambling?

Legally, in the US, regulated event contracts sit under the CFTC as derivatives, and courts have so far upheld that framing against state gaming challenges for election contracts and in early sports rounds. Economically, quant firms treat them as binary options with an information-aggregation function. The gambling classification fight is precisely the live regulatory risk.

Do I need to know sports or politics to work on these desks?

For sports contract roles, yes: a genuine domain interest is asked for explicitly. For arbitrage, microstructure and engineering roles, the quant skills matter more than domain knowledge.

Do prediction markets desks hire new graduates?

Some do. Akuna's junior sports researcher roles say prior trading-firm experience is valued but not necessary, and DRW's trader role weights demonstrated passion and personal trading. Engineering roles hire grads routinely.

How is pricing a binary contract different from pricing a stock?

The payoff is bounded and settles at 0 or 1, price equals implied probability, spreads and risk live in probability space, quoting near the extremes is dangerous, and there is often no continuous instrument to hedge against.

Is prediction markets trading a safe career bet?

It is a promising but narrower and riskier bet than a mainstream quant seat, mainly because of regulatory risk and desk immaturity. Keep your core skills general and target firms that also run mainstream books.

Everything these desks test is trainable. Make probability and expected value automatic with the probability brainteasers and the Probability Trainer, practise quoting two-sided prices out loud in the market games, sharpen your estimation and calibration with the Fermi questions, and keep your arithmetic fast with the Math Trainer. When you know where you are applying, work through everything we have for that firm, and read the Ultimate Guide to Quant Trading Interviews before your first stage.