Institutional entry into prediction markets is real, named, and accelerating — but later and narrower than folklore suggests. Susquehanna (SIG) became Kalshi's first dedicated institutional market maker in April 2024 (not 2021, as commonly repeated), and by mid-2026 DRW, Wintermute, and IMC are all hiring dedicated prediction-market traders explicitly targeting Polymarket/Kalshi market-making and cross-venue arbitrage. Where institutions patrol, the edge is provably dead: single-market NBA arbitrage on Polymarket collapsed to 7 executable episodes in a month (median lifespan 3.6 seconds). But two independent academic studies converge on a documented capacity floor — Kalshi's top-decile markets average only ~$526k lifetime volume and Polymarket sports arb caps out at ~15-share fills — meaning residual inefficiencies (favorite-longshot bias, combinatorial/cross-market arb at ~100bps, thin long-tail markets, under-arbitraged sports rebalancing) are structurally confined to retail-scale capital and cannot absorb institutional size. For a $10k–$200k solo operator the pond in mid-2026 is crowded at the top-of-book of large liquid markets but demonstrably open below the institutional minimum-viable-market-size line, with the caveat that the DRW/Wintermute/IMC hiring wave is the clearest edge-death early-warning signal now flashing.
TIMELINE CORRECTION (folklore vs. fact): SIG's dedicated institutional market-making on Kalshi dates to April 3, 2024 — not 2021. Kalshi's own press release, Bloomberg's same-day coverage, and Kalshi CEO Tarek Mansour all identify Susquehanna Government Products, LLLP (SIG subsidiary) as the FIRST dedicated institutional market maker on the exchange, announced April 3, 2024, described as the first tier-one market maker aligning exclusively with an event-contract exchange. Adversarial searches found zero credible sources placing SIG in an MM role in 2021; the '2021 entry' appears to be folklore conflating Kalshi's July 2021 launch with SIG's entry. SIG separately self-reports establishing its dedicated prediction-markets trading desk in 2023 ('first quant trading firm' to do so, per its own site) — so the internal desk (2023) preceded the public Kalshi MM commitment (April 2024).
high
Press release: 'Serving as Kalshi's first dedicated institutional market maker to provide liquidity to the exchange... First instance of a tier-one market maker aligning exclusively with an event contract trading exchange.' Bloomberg (Apr 3, 2024): 'Susquehanna Starts Trading Desk for Event Contracts on Kalshi.' SIG site: 'In 2023, we became the first quant trading firm to establish a dedicated pr
WHAT SIG DOES AND WHY: SIG operates as Kalshi's flagship market maker, self-describing 24/7 two-sided liquidity provision across finance, crypto, elections, culture, weather, and sports — institutional spread/depth provision is category-broad, not confined to elections or finance. Bloomberg documented the role as explicit market-making (facilitating transactions for institutional investors including hedge funds), and the stated strategic rationale was an institutional land-grab: attracting Wall Street to a new asset class, with Kalshi simultaneously raising trading limits to accommodate hedge funds and large participants.
high
SIG site (live July 2026): 'a leading market maker across regulated venues, including serving as the flagship market maker on Kalshi'; 'deep, two-sided liquidity across markets spanning finance, crypto, elections, culture, weather, sports, and more'; 'Our team works 24/7.' Bloomberg: 'Susquehanna will function as a market maker on Kalshi's platform, facilitating transactions for institutional inve
THE $40M BASELINE: The peer-reviewed AFT 2025 study (Saguillo et al., IMDEA) measured ~$39.6M ($39,587,585) in realized arbitrage profit extracted on Polymarket from on-chain order book data across markets resolving April 1, 2024 – April 1, 2025, spanning market-rebalancing (within-market, genuinely riskless) and combinatorial (cross-market) arbitrage — this is the documented source of the '~$40M' figure. Extraction was concentrated among bot-like accounts: the top account made ~$2.0M across 4,049 transactions, and the top 10 made $380k–$2M each (together ~21% of the total) — quasi-professional actors already worked Polymarket arb in 2024–25, well before the named institutional desks arrived.
high
Paper: 'The combined amount extracted from all strategies presented totals $39,587,585.02'; 'a realized estimate of 40 million USD of profit extracted'; top user '$2,009,631.76'; Table 1 top-10 range $383,569.94–$2,009,631.76. Peer-reviewed (LIPIcs AFT 2025, DOI 10.4230/LIPIcs.AFT.2025.27). Caveats: election-cycle window (not a run-rate), heuristic epsilon=$1-per-trade estimate, and 'dominated' ov
INEFFICIENCY OUTPACED EXTRACTION THROUGH APRIL 2025: Despite $40M extracted, Polymarket remained grossly inefficient in the study window — 7,051 of 17.2K conditions had at least one arbitrage opportunity (≥$0.05/dollar), median sum-of-conditions price was ~$0.60 (vs. $1.00 efficient), and only ~1% of estimated U.S. election opportunities were actually exploited. Sports was the documented under-arbitraged pocket: most within-market opportunities (~100 per NegRisk market on average, sports the outlier-heavy category) yet largely absent from rebalancing extraction — the authors flag it as 'a less explored venue for arbitrageurs' as of April 2025.
high
Verbatim: 'the median price of the sum of conditions (i.e., profit per Dollar) is around $0.60 for all topics of markets, showing remarkable market inefficiency'; 'About 1% of our estimated U.S. election opportunities were exploited by users'; 'Surprisingly, Sports are largely absent from the plots – maybe a less explored venue for arbitrageurs.' Caveats: opportunity counts are VWAP-reconstructed
KALSHI'S PERSISTENT RETAIL-SIDE MISPRICING: Kalshi prices exhibited a favorite-longshot bias from 2021 through April 2025 that was not competed away — contracts under 10c lose >60% of invested money (fee-inclusive), and average pre-fee equal-weighted return per contract is about -20% (GWU/UCD working paper, Bürgi, Deng & Whelan, first systematic academic Kalshi study). The authors' first explanation is a capacity floor: even top-decile Kalshi markets averaged only $526,245 final volume, with order-book depth at any moment far smaller — too small for professional capital, leaving the bias to be harvested (by shorting longshots / passively making on favorites) only at sizes institutions won't bother with.
medium
Verbatim: 'investors who buy contracts costing less than 10c lose over 60 percent of their money'; 'the average pre-fee return on a Kalshi contract in our data is -20%'; 'Even its largest-volume markets are small relative to the kinds of markets that professional investors will typically be willing to participate in. Table 3 showed average final trading volume in the top decile of Kalshi markets w
WHERE THE GUARDS ALREADY PATROL — SINGLE-MARKET ARB IS DEAD IN LIQUID SPORTS: By Feb–Mar 2026, single-market arbitrage in Polymarket NBA markets is nearly extinct. Across 75M order-book snapshots covering 3,042 markets (173 games), a UCLA study found only 7 valid executable in-game arbitrage episodes, corrected by automated participants with a median lifespan of 3.6 seconds. This is direct before/after evidence that the automated/institutional tier has compressed the fastest, most legible edge class to zero for anyone without co-located bots.
medium
Verbatim: 'We identify only 7 valid, exploitable in-game arbitrage episodes across 3,042 markets... corrected swiftly by automated participants, persisting for a median duration of just 3.6 seconds.' Rated medium: single non-peer-reviewed preprint (though methodologically careful: L1 executability checks, $10 liquidity floor, mirrored-book deduplication), one month, NBA-only; 3.6–5.5s polling mean
THE CAPACITY FLOOR — WHAT PROVABLY STAYS OPEN TO SOLO OPERATORS: The same NBA study finds combinatorial (Moneyline-vs-Spread) arbitrage still exists at meaningful percentage returns (290 executable episodes, median yield 101bps) but is execution-bottlenecked by shallow depth: in 76.9% of episodes even a $100 budget could not be fully deployed (average executable size ~14.8 shares), capping monthly extractable profit at $559.59 vs. a theoretical $2,032.75. The paper's core conclusion, converging with the Kalshi study's independent capacity-floor argument: residual inefficiencies 'are not accessible at institutional scale, they are structurally confined to the retail tier' — sports contracts resolve within hours, so liquidity never deepens enough for large risk-free extraction. The structural moat for a solo operator is not skill but size: markets and edge classes below the institutional minimum-viable-market-size (long-tail events, thin books, sub-$1M-volume markets, longshot-shorting on Kalshi) remain economically inaccessible to SIG/DRW-scale desks.
high
NBA paper verbatim: 'In 76.9% of all episodes, the $100 budget could not be fully deployed, with the average executable size restricted to approximately 14.79 synthetic shares'; 'structurally confined to the retail tier, offering diminishing returns to any participant seeking to deploy meaningful capital'; 'in sports markets, where contract lifespans are measured in hours rather than weeks, this l
THE 2026 INSTITUTIONAL WAVE — EDGE-DEATH CLOCK RUNNING: As of mid-2026, DRW is building a dedicated prediction-markets desk explicitly targeting Polymarket and Kalshi (live job listing, $175k–$200k base, requiring Polymarket CLOB/Gamma/Subgraph and Kalshi FIX/WebSocket/REST API experience), with a stated strategy set spanning market making, microstructure exploitation, cross-platform arbitrage, event-driven momentum, and statistical models. Wintermute is hiring algorithmic traders with prediction-market experience and IMC is hiring quant traders for binary event contracts (CoinDesk, June 2026). This confirms named prop-firm entry into both venues and direct intent to compete in spread-capture and cross-venue arb — the classes closest to what a technical solo operator would run. Job postings evidence desk buildout and intent, not live P&L or market share.
high
DRW listing verbatim (live, fetched July 2026): 'building out a dedicated prediction markets desk focused on platforms such as Polymarket and Kalshi'; 'consistent positive expectancy through market making, microstructure exploitation, cross-platform arbitrage, event-driven momentum, and statistical models.' CoinDesk: DRW candidates must 'monitor prices in real time across both platforms simultaneo
(1) Timeline folklore corrected: the research question's premise of a 2021 SIG market-making entry on Kalshi is refuted by every source found — the documented first institutional MM commitment is April 3, 2024; treat any '2021 SIG MM' reference as folklore. The origin story ('earlier SIG activity was trading participation') is itself undocumented speculation. (2) Coverage gaps: no surviving verified claims on Jane Street, Jump, Flow Traders, or GSR on Polymarket; on ICE's reported $2B Polymarket investment; on Polymarket's QCEX/CFTC re-entry path; on documented hedge-fund election-market participation in 2024; on the French whale; or on direct before/after spread-compression measurements as MMs entered — these parts of the question remain unanswered by verified evidence, not disproven. (3) Time-sensitivity: the $40M Polymarket study covers Apr 2024–Apr 2025 (election-cycle peak, not a run-rate); the Kalshi bias study ends April 2025, before Kalshi's maker-fee introduction and the late-2025/2026 sports volume surge, so both the bias magnitude and the $526k top-decile volume figure may be stale; the NBA arb study is one month (Feb–Mar 2026), NBA-only, non-peer-reviewed preprint. (4) Source quality: SIG's category-breadth and 24/7 claims are the firm's own marketing copy; job postings prove hiring intent, not deployed capital or P&L; Kalshi's press release is promotional (though Bloomberg-corroborated). (5) Three claims were refuted in verification and excluded: Kalshi Makers' +