Across five NCAA Tournament games, the pricing gap between the best and worst prediction market platforms was larger than the gap between prediction markets and traditional sportsbooks. House take, the margin platforms build in by pricing both sides above true probability, ranged from 1.3% to 8.2% in our dataset. The higher the percentage, the more the platform keeps on every dollar wagered, regardless of outcome.
I tracked odds across six prediction market apps and two traditional sportsbooks during five NCAA Tournament games during March 20–22, testing everything from a 15-seed longshot to a near-coin-flip. The best platform, Polymarket US, charged a house take of 1.3–1.6% across every game — well below what traditional sportsbooks charge, and less than a third of what most competing prediction market platforms extract. The worst performer hit 8.2% on an essentially even matchup.
Pricing hasn’t settled across prediction markets yet, and the spread is wide enough to significantly affect long-term expected value. Here’s where each platform falls.
A $15 bet that said everything
Round 1 | March 20 | Furman (15) vs. UConn (2)
The experiment started as a side bet on my alma mater. I was going to wager on 15-seed Furman against UConn regardless — might as well see what the apps were offering.
The spread was jarring. Across six prediction market platforms, Furman’s effective moneyline ranged from +1,895 (Polymarket) to +1,150 (DraftKings and OG.com, a standalone predictions app from Crypto.com), a difference of 745 odds points on the same bet. On $15, that’s the difference between netting $284 and $172. Polymarket paid 65% more in profit on the same stake.
The starkest comparison, however, involved no spread at all. FanDuel Sportsbook offered +2,000, a $300 net on $15. Meanwhile, FanDuel Predicts showed the same $300 payout on screen, but only after a $6 fee on top* of the $15 wager, for a total cost of $21. That fee pushed the effective odds to +1,329. Same brand, same number on the screen, but a wildly different deal.
| Platform | FUR Eff. Odds | Cost → Payout |
|---|---|---|
| FanDuel Sportsbook | +2,000 | $15 → $315 |
| Polymarket | +1,895 | $14.99 → $299 |
| Kalshi | +1,773 | $15 → $281 |
| Gemini | +1,487 | $15 → $238 |
| FanDuel Predicts | +1,329 | $21* → $300 |
| DraftKings Predictions / OG.com | +1,150 | $14.96 → $187 |
Important note: DraftKings Predictions’ displayed price implied better odds than what actually executed, signaling a potential liquidity problem on a lopsided market. The displayed odds of +1567 should have produced a potential payout of $250.05, not $187.
For customers, it’s worth noting that the displayed price and the executed price are not always the same thing. That gap on smaller and/or newer apps and is worth watching as these platforms scale.
Game 1: Louisville vs. Michigan State
Round 2 | March 21 | Louisville (6) vs. Michigan State (3) | No sportsbook in this capture
| Platform | LOU Eff. Odds | MSU Eff. Odds | House Take |
|---|---|---|---|
| Polymarket | +202 | -217 | ~1.6% |
| Kalshi | +189 | -233 | ~4.5% |
| DraftKings Predictions / OG.com | +194 | -245 | ~5.0% |
| FanDuel Predicts | +186 | -233 | ~5.0% |
| Gemini | +180 | -256 | ~7.6% |
The hierarchy that would hold all weekend established itself immediately: Polymarket alone at the top, a cluster in the 4.5–6% range, and Gemini at the bottom. DraftKings and OG’s Louisville price (+194) was decent, but their Michigan State line (-245) was the worst favorite price apart from Gemini. That split pricing inflated the overall take despite competitive underdog odds.
Game 2: Miami FL vs. Purdue
Round 2 | March 22 | Miami FL (7) vs. Purdue (2) | Closest capture: 6 min pre-tip
| Platform | PUR Eff. Odds | MIA Eff. Odds | House Take |
|---|---|---|---|
| Polymarket | -321 | +299 | ~1.3% |
| Gemini | -345 | +296 | ~2.8% |
| Kalshi | -345 | +280 | ~3.8% |
| FanDuel Sportsbook | -317 | +260 | ~4.5% |
| FanDuel Predicts / DraftKings Predictions / OG.com | -355 | +270 | ~5.0% |
| Hard Rock Bet | -375 | +275 | ~5.6% |
The one game where Gemini, the crypto exchange that launched prediction markets in December, looked competitive with 2.8% take, second only to Polymarket. On lopsided markets, Gemini’s price compression matters less because one side is deeply discounted: its Miami price (+296) was genuinely good.
Hard Rock posted -375 payout odds on Purdue, meaning bettors backing the favorite paid more vig than on any other platform in the dataset. For Florida users, whose only licensed sportsbook alternative is Hard Rock under the Seminole gaming compact, prediction market pricing compared to the sportsbook is particularly consequential. This game illustrated both why and how much it can affect potential payouts.
FanDuel Sportsbook’s -317 on Purdue was the best favorite price across all platforms, the one game where the traditional sportsbook was the single best place to back the favorite. But it’s not available to Florida customers.
Game 3: St. John’s vs. Kansas
Round 2 | March 22 | St. John’s (5) vs. Kansas (4)
| Platform | SJU Eff. Odds | KU Eff. Odds | House Take |
|---|---|---|---|
| Polymarket | -164 | +156 | ~1.3% |
| FanDuel Sportsbook | -162 | +134 | ~4.5% |
| Kalshi | -175 | +145 | ~4.5% |
| Hard Rock Bet | -190 | +155 | ~4.7% |
| FanDuel Predicts / DraftKings Predictions / OG.com | -178 | +144 | ~5.0% |
| Gemini | -185 | +138 | ~6.9% |
Near-even games are where overround, or house take, starts to sting hardest. On a close matchup there’s no lopsided favorite to absorb the cost; both sides are priced near even money, so the full margin hits every bettor regardless of which side they’re on.
Polymarket’s favorite price (-164) was nearly identical to FanDuel Sportsbook (-162), showing alignment on the favorite pricing. But Polymarket’s Kansas underdog odds (+156) were 22 points better than FanDuel Sportsbook’s (+134). Close games reveal a consistent pattern: Polymarket is competitive on favorites and superior on underdogs.
Game 4: Texas Tech vs. Alabama
Round 2 | March 22 | Texas Tech (5) vs. Alabama (4)
| Platform | TTU Eff. Odds | ALA Eff. Odds | House Take |
|---|---|---|---|
| Polymarket | -109 | +104 | ~1.3% |
| Kalshi | -116 | -103 | ~4.5% |
| FanDuel Sportsbook | -113 | -106 | ~4.5% |
| Hard Rock Bet | -105 | -115 | ~4.7% |
| FanDuel Predicts | -117 | -104 | ~5.0% |
| DraftKings Predictions / OG.com | -113 | -108 | ~5.0% |
| Gemini | -120 | -115 | ~8.2% |
On a coin flip, Gemini’s 8.2% take isn’t just expensive — it’s mathematically punishing. Both sides at minus money means a bettor loses money in the long run even picking the winner half the time. Hard Rock was the only platform that priced Alabama as the favorite, which produced the best Texas Tech price in the dataset (-105) and the worst Alabama line (-115).
The general trends of Polymarket as the clear winner and most of the others grouping between 4.5-5% holds up across the set. The wide variety in specific team pricing is stark here, a good reminder of the benefits of cross-platform comparison depending which side you’re taking.
What the data shows about prediction market pricing
The platform hierarchy was strikingly stable across four games, different odds environments, and different days. Polymarket led every game at 1.3–1.6%. Kalshi was a consistent second usually around 4.5%, better than the field but 2.9–3.2 percentage points behind Polymarket.
The most telling pattern: FanDuel Predicts, DraftKings Predictions and OG.com landed at exactly 5.0% in every single game, despite operating completely different fee structures, with FanDuel itemizing fees explicitly on top of the wager and the others baking them into the price. Identical overround across four games and two different pricing mechanisms points to deliberate margin-targeting, not coincidence. Rather than competing on price, these platforms appear to be coordinating around a number. FanDuel and DraftKings also both route markets from CME Group, though DraftKings also offers markets from Crypto.com, the parent exchange of OG. So there are some connections amongst this group.
Gemini ranged from competitive on blowouts to punishing on close games. Its 2.8% on Purdue/Miami and 8.2% on Texas Tech/Alabama represent the widest performance swing in the dataset. Given that its a new product within its crypto trading platform, and not sharing liquidity with another DCM, low liquidity is still likely an issue.
One of the more surprising findings was FanDuel Sportsbook posted 4.5% across every game where I tracked it, which is in line with Kalshi and better than four of the six prediction market platforms in the set. While the industry has claimed prediction markets beat sportsbooks on pricing, this data snapshot suggests that it depends on the exchange (and the sportsbook). Our findings clearly show that the best prediction market price beats everything, and the worst ones are no better than some sportsbook pricing.
For anyone placing meaningful volume, the choice of platform is a major difference-maker to your bottom line. At 8.2% versus 1.3% on an even game, you’re giving back roughly $70 on every $1,000 wagered just in structural margin, before a single bad pick. Compounded across every trade, every game, and every season, one can imagine the hit to a bettor’s long-term P&L statement.
Methodology notes
Effective odds were calculated from actual transaction screens — cost to payout — rather than displayed prices. The formula: net profit ÷ actual cost × 100, expressed in American odds format. For traditional sportsbooks, effective odds equal the posted moneyline.
House take (also called overround or vig) measures how much of each dollar wagered the platform keeps, on average, across both sides of a market. A fair, zero-margin market sums to exactly 100% implied probability. The excess over 100% is the house take.
Fee handling: FanDuel Predicts explicitly itemizes fees on top of the stated wager; I used total cost (wager + fee) as the denominator for all calculations. All other prediction market platforms bake fees into the price.
Platforms tested: Polymarket, Kalshi, FanDuel Predicts, OG.com, Gemini, and DraftKings Predictions (Round 1 only), plus Hard Rock Bet and FanDuel Sportsbook as traditional benchmarks. Games were selected to cover heavy favorites, moderate favorites, and near-even matchups.
Location: All screenshots in this piece were captured from within Florida. Unlike most legal betting states where multiple licensed sportsbooks compete for your business, Florida bettors have essentially one traditional option: Hard Rock Bet, operating under the Seminole Tribe’s gaming compact. Prediction market platforms, regulated at the federal level by the CFTC, are accessible statewide.
All odds were captured within the same platform session, in advance of tipoff. Capture windows ranged from roughly 6 to 37 minutes before tip, and all screenshots for a given game were captured within a span of 20 minutes. The Furman/UConn game was a single-sided trial and is not included in cross-platform house take comparisons.
Data is for informational and observational purposes only and is not betting advice.
More on DraftKings Predictions
