Every prediction market milestone seems to arrive with a number so large it functions as its own argument. March Madness 2026 produced one of the biggest yet: $530 million in prediction market handle-equivalent, a 33-fold increase over the $16 million recorded in 2025. The figure has circulated through earnings calls, regulatory filings, and operator boardrooms as evidence that prediction markets have arrived as a mainstream force in U.S. sports betting.
It hasn’t. Or rather: they have, but not at $530 million. To understand what the number actually represents—and what decisions operators should make based on it—requires decomposing four distinct inflation sources that compound on each other. Strip them out, and a genuinely significant market emerges. Leave them in, and you risk building strategy on a fiction.
The HeadlineThe $530M Number Everyone Is Citing — and Why It Misleads
The $530 million figure originates from H2 Gambling Capital, a respected gaming data firm, and represents a “handle-equivalent” estimate for prediction market activity during March Madness 2026 across all U.S. states. The qualifier is doing enormous work. Handle-equivalent is not handle. It is H2’s translation of prediction market volume—a structurally different metric—into language that sportsbook analysts are accustomed to reading. The translation introduces distortion at the first step.
The first relevant disaggregation is geographic. Prediction markets operate legally in states that have not legalized sports betting, capturing users with no sportsbook alternative. When H2 filters the figure to states where legal sportsbooks already operate—the only geography where prediction markets actually compete with regulated betting operators—the comparable number is $135–150 million. That is not a rounding difference. It is the actual competitive overlap.
Even $135–150 million, in context, is 3.5% of the projected $4 billion sportsbook handle for March Madness 2026 in those same states. That is material and growing. It is not existential. And it is not $530 million.
| Figure | Value | What It Represents |
|---|---|---|
| PM handle-equivalent, all U.S. states | $530M | H2 Gambling Capital; includes non-sportsbook states |
| PM handle-equivalent, legal betting states only | $135–150M | Actual competitive overlap with regulated sportsbooks |
| Sportsbook handle, March Madness 2026 | $4B (all-time record) | Excludes prediction market activity; U.S. sportsbooks only |
| PM March Madness baseline, 2025 | $16M | Prior year; methodology unchanged from 2026 |
The jump from $16 million to $530 million looks like 33x growth. The underlying methodology has not changed—what changed is adoption. That is a meaningful distinction when calibrating trend lines: the accelerant is real-world usage expansion, not improved measurement accuracy. Which makes it all the more important to be precise about what is actually being measured.
Methodology GapVolume and Handle Are Not the Same Thing
Sportsbook handle has a clean, universally understood definition: it counts the bettor’s stake. A $1 wager on a 10-to-1 outcome generates $1 in handle. That $1 is the amount at risk, the amount taxed, the amount regulators track, and the amount operators report to their gaming commissions. The definition is enforced by law in every regulated jurisdiction.
Prediction market volume has no equivalent regulatory definition. In most implementations, it counts both sides of each contract trade—buyer and seller. A $1 bet placed on a contract priced at 10 cents (10-to-1 equivalent odds) involves one buyer and one seller, and the total transaction is logged as $10 in volume. The same economic activity that generates $1 in sportsbook handle generates $10 in prediction market volume. The ratio is not fixed; it varies by contract price and market structure. It routinely runs 5–15x.
DraftKings made this asymmetry explicit in its Q4 2025 earnings letter. The company reported $2.5 trillion in “total potential payouts”—a figure derived from multiplying out parlay leg exposure across the book—and explicitly footnoted it as “comparable to the volume that predictions operators report.” DraftKings’ actual handle for the same period: $54 billion. The same business, framed in prediction market terminology, produces a 46x larger headline number.
There is no universal standard. Platforms mix taker-side-only, notional gross, and handle-equivalent approaches. Kalshi reported approximately $1.1 billion in monthly sports volume for early 2026. Without knowing the methodology, that figure cannot be compared to a sportsbook handle number, to Polymarket’s reported volume, or to Opinion Labs’ figures. All four use different counting conventions. Cross-platform league tables built from these numbers are, in the most precise sense of the word, meaningless.
Technical InflationPolymarket’s Double-Counting Bug: 100% Overstatement Confirmed
In December 2025, Paradigm Research published a detailed technical investigation into Polymarket’s reported volumes. The finding was unambiguous: a smart contract bug caused each trade to emit two OrderFilled events—one for the maker side, one for the taker side—and every major third-party analytics platform had been summing both. A $4.13 trade was recorded as $8.26 in volume across DefiLlama, Allium, and Dune, the three data sources that essentially every mainstream prediction market analysis cited.
Paradigm tested multiple independent methodologies to determine the true figure. All valid approaches converged on approximately 50% of the headline number. Major dashboards corrected their calculations after publication. The practical implication is stark: every Polymarket-inclusive volume figure cited before December 2025—every growth rate, every market share estimate, every TAM projection—requires a 50% downward revision to be usable for strategic planning.
The global prediction market total of $63.5 billion reported for 2025 (CertiK, with approximately 400% year-over-year growth) incorporates Polymarket’s inflated figures. Adjusted for the double-count, the true 2025 global figure is approximately $40–45 billion. That is still a very large number representing genuine growth. It is also 30–35% smaller than the headline, and the distinction matters when making strategic investment or partnership decisions based on TAM.
The $5.35 billion weekly record set by Kalshi and Polymarket combined during the week of March 2–8, 2026 is the post-correction high-water mark. It remains an extraordinary figure. But it represents actual activity rather than doubled output from a software bug—a foundation that makes subsequent growth comparisons interpretable.
Artificial ActivityWash Trading, Incentive Loops, and the 45% Sports Problem
Even after correcting for the double-counting bug and the methodology gap, reported volumes contain a third inflation layer: artificial trading activity generated by the incentive structures of blockchain-native prediction markets.
A November 2025 Columbia University study examined three years of Polymarket historical data and found that wash trading—trades executed to generate volume without real economic exposure—accounted for an average of 25% of all platform volume. In sports markets specifically, the figure rose to 45%. Fourteen percent of wallets were flagged; during high-incentive periods, wash trading peaked at approximately 60% of all activity.
The mechanics enabling this are structural, not accidental. Polymarket charges zero transaction fees. Blockchain accounts are pseudonymous. Token airdrop programs reward volume generation regardless of intent—creating a direct financial incentive to churn trades with no underlying economic position. Under these conditions, a meaningful share of reported “volume” represents cost-free self-trading rather than real capital at risk.
Kalshi faced a separate allegation from a Polymarket team member: that its reported esports volume of $1.7 billion was fabricated through double-counting Counter-Strike markets under both “CS:GO” and “CS2” labels, with the true figure estimated at approximately $63 million—a 27-fold gap. Kalshi has not confirmed the methodology underlying its esports figures.
Opinion Labs, which reported $8.08 billion in January 2026 volume representing 31% of the entire industry, exhibited a different set of anomalies. Its average trade size was $2,525 versus $147–175 at comparable competitors—13–25 times higher. Just 0.7% of its transactions generated 13.2% of industry-wide volume. And per-user volume doubled as the platform grew, the reverse of normal scaling behavior where volume-per-user typically declines as a platform matures beyond its early power-user base. None of these patterns are definitively fraudulent. All are structurally inconsistent with organic trading activity.
Contract MechanicsBinary Contracts and the Open Interest Reality Check
The fourth and final inflation layer operates independently of bugs, wash trading, or methodology choices. It is intrinsic to how binary prediction market contracts work.
In a binary market, each YES or NO share is valued at $1 at settlement. Contract volume is reported in notional terms: every share, regardless of its purchase price, counts as $1. A share purchased for 1 cent still registers $1 in notional volume at settlement. A share purchased for 99 cents also registers $1. Because prediction market contracts typically trade far from the 50-cent midpoint—underdogs and heavy favorites dominate sports markets—the average purchase price is substantially below $1, and the ratio of notional volume to actual capital deployed is correspondingly high.
The most direct measure of actual capital deployed is open interest: the real money locked in live positions at any given moment. In late 2025, total prediction market open interest was approximately $700 million versus $44 billion or more in reported annual notional volume. That is a gap of approximately 63x—meaning that for every dollar of capital actually at risk in prediction markets, the notional volume metric reports $63 in activity.
| Market Type | Typical Volume-to-Handle Conversion | Example |
|---|---|---|
| Game markets (basketball, football, baseball) | 45–55% | $100M volume ≈ $45–55M real handle |
| Futures / tournament outrights | 5–20% | $100M volume ≈ $5–20M real handle |
| Golf (Masters, majors) | 5–20% | $87M Kalshi Masters volume ≈ $4–17M real risk |
Sporttrade’s analysis of Kalshi’s $87 million in reported Masters golf volume put the true handle equivalent at $4–17 million. For futures and golf markets, only 5–20% of reported volume corresponds to what a sportsbook would record as handle. Game markets run closer to 45–55%—still far below the 1:1 ratio that most press coverage implicitly assumes when comparing prediction market volume to sportsbook handle.
What Prediction Markets Are Actually Capturing
Strip out the four inflation layers and a more legible picture emerges—one that is simultaneously less alarming and more strategically interesting than the headline suggests.
Prediction markets are legally available in states that have not legalized sports betting. A meaningful share of their volume—the portion outside the $135–150 million that overlaps with legal betting states—represents genuinely new bettors in new geographies with no sportsbook alternative. This is not cannibalization. It is market expansion, and it operates under different competitive dynamics than the regulated sportsbook market.
Operators focused on U.S. sportsbook operations should be tracking the $135–150 million figure for March Madness, not $530 million. The $530 million conflates two fundamentally different populations: users who chose prediction markets instead of a sportsbook they could have used, and users for whom prediction markets are the only legal option. Strategy built on the combined figure will systematically overestimate the direct competitive threat.
The American Gaming Association has framed the broader volume debate in fiscal terms, estimating that prediction markets have already cost U.S. governments more than $500 million in potential sports betting tax revenue. This framing adds a regulatory dimension that extends beyond competitive dynamics: the volume question has become a policy question, with state legislatures increasingly attentive to what activity is or is not flowing through their licensed sportsbook tax base.
As of March 2026, Kalshi holds approximately 53% of combined two-platform weekly volume across itself and Polymarket—a meaningful shift in platform concentration that changes the competitive analysis of which platforms actually set market prices and which follow them.
What to MeasureA Framework for Reading Prediction Market Numbers Without Getting Burned
The problem is not that prediction market volume is immeasurable. It is that the industry has not yet established the standardized reporting that makes strategic planning reliable. Until it does, operators and vendors need their own normalization layer.
1. Always identify the methodology first
Ask which methodology is being used: taker-side only, notional gross, or handle-equivalent. Ask whether the source pre- or post-dates Paradigm’s December 2025 correction. Any pre-correction Polymarket figure requires a 50% downward adjustment before it enters a model.
2. Use open interest, not volume, for capital deployment
Open interest—actual capital locked in live positions—is a more reliable signal than notional volume for assessing real market scale. The current industry-wide ratio of approximately 63x (notional volume to open interest) provides a calibration anchor. If a platform reports an unusually low ratio, that is a signal of more genuine trading activity. If it reports a higher ratio, volume is even more detached from capital deployment than the industry average.
3. Apply sport-type conversion rates
For direct comparison to sportsbook handle, apply Sporttrade’s published conversion factors: 45–55% for game markets (basketball, football, baseball), 5–20% for futures and golf. Kalshi’s $87 million Masters figure likely represents $4–17 million in real economic exposure—not because Kalshi is misreporting, but because the contract structure produces this ratio.
4. Segment geography before making competitive assessments
Use the $135–150 million legal-state-overlap figure, not $530 million, for competitive analysis. The remainder represents a different market with different user profiles and different strategic implications.
5. Monitor behavioral anomalies as leading indicators
Before regulatory correction arrives, platform-level anomalies provide early warning. Watch for: average trade size that significantly exceeds industry norms, transaction concentration (a small percentage of transactions generating a disproportionate share of volume), and per-user volume that increases rather than decreases as a platform scales. These were all present in Opinion Labs’ January 2026 data before independent analysts flagged the inconsistencies.
The real story is not that prediction markets are a small and ignorable force. $40–45 billion in annual volume—even after all corrections—is a significant market. Kalshi and Polymarket combined set a $5.35 billion weekly record in early March 2026. These are not rounding errors.
The story is that the industry has not yet established the measurement conventions that make this scale interpretable. Until it does, every headline number requires decomposition. Operators who build strategy on the raw figures will systematically overestimate the competitive threat in their core markets and underestimate the incremental opportunity in unserved geographies. Both errors are costly.
SourcesData Sources & Attribution
- H2 Gambling Capital / Yahoo Sports — $530M handle-equivalent estimate, March Madness 2026; $135–150M legal-state-only figure; $16M 2025 baseline
- Covers.com — $4B sportsbook handle projection, March Madness 2026
- Paradigm Research, December 2025 — Polymarket double-counting confirmation; $4.13 trade recorded as $8.26; all valid methodologies converge on ~50%
- Columbia University / Fortune, November 2025 — 25% average wash trading; 45% in sports markets; 14% of wallets flagged; 60% peak
- Closing Line / Sporttrade analysis — 5–20% volume-to-handle conversion for golf/futures; 45–55% for game markets; Kalshi $87M Masters → $4–17M real risk
- Esports.net — Kalshi $1.7B claimed esports volume vs. ~$63M actual; CS:GO/CS2 double-count allegation
- Kaiko Research — ~$700M open interest vs. $44B+ notional volume (late 2025); ~63x gap
- CertiK / Decrypt — $63.5B global reported volume 2025; ~400% YoY; adjusted to ~$40–45B post-correction
- DeFiRate — $5.35B combined Kalshi+Polymarket weekly record, March 2–8, 2026; Kalshi ~53% of two-platform volume