Prediction Markets vs Sportsbooks: The House Edge Just Changed Seats
On January 17, 2026, one football game carried two prices. FanDuel hung the Bills at -100 (50%) and the Broncos at -118 (54.1%), so its two sides add up to 104.1%. Kalshi listed the same Bills–Broncos game at $0.48 and $0.52, exactly 100%, according to Lines.com.
That 4.1-point gap is the entire sales pitch for prediction markets. The gap is real, but it's only the first line on the receipt.
I spent 15 years grading lines for a Vegas book before I switched sides. Now I read closing-line value the way an accountant reads a P&L. So let's run the whole receipt on prediction markets vs. sportsbooks: the fees, the spread, and the trader on the other side of your click.
The 104.1% vs. 100% Gap: What the Vig Actually Costs You
A standard -110/-110 line implies 52.38% per side, or 104.76% in total. Effortless Math calls that extra 4.76% the overround, or hold. It's the book's cut, and it's baked into the price before you tap "place bet."
Benzinga puts it in dollars. Say one bettor puts $110 on each side. The book takes in $220, pays the winner $210 and keeps $10, about 4.5% of the money wagered. The book doesn't care who wins. It only needs balanced action and a 104.76% price.
FanDuel's Bills–Broncos market totaled 104.1%, so its margin was locked in before kickoff. Kalshi's 48¢/52¢ split totaled exactly 100%, so the quote itself carried no overround.
I won't pretend that difference is fake. It's a real number on a real screen. But the edge didn't disappear. It moved to places the quote screen doesn't show.
The book didn't vanish. It changed seats.
Where the House Edge Goes on a Prediction Market
A prediction-market contract trades between $0.01 and $0.99. It settles at $1 if the event happens and $0 if it doesn't. Per Action Network, a 65¢ price reads as roughly a 65% market-implied probability.
The plumbing differs too. Sports Illustrated notes that prediction markets are peer-to-peer exchanges regulated by the CFTC, while sportsbooks are house-banked and licensed state by state. That one structural fact drives all the math below.
On an exchange, the house edge turns up in three places.
1. Fees
Benzinga and Action Network both say the platform earns fees on executed trades, whether you win or lose. Lines.com puts Polymarket's fee at 2% on winning positions, so a $100 winning trade costs $2.
Compared with the vig, the cut is smaller and comes later. You still pay it.
2. The bid-ask spread
Action Network spells out the catch: you get your price only if another trader is willing to meet it. A wider bid-ask spread makes every trade more expensive, and market makers may post orders on both sides just to collect that spread.
On a sportsbook, the vig sits inside the odds. On an exchange, the spread does the same job. It's the gap between the price on the screen and the price you can actually get filled at.
3. The trader across from you
Benzinga says it plainly: every dollar one trader wins comes from another trader who lost, minus the platform's fees.
Benzinga also notes that sportsbooks can cap how much consistent winners may bet, because those winners cost the book money. An exchange has no reason to cap anyone, since it collects its fee either way. So the sharpest traders stay in the pool, and the weekend bettor is often trading straight into them.
At a sportsbook, you bet against a house that prices for a margin. On an exchange, you may be trading against someone whose whole job is pricing games better than you do.
Price It Like Closing-Line Value: The Ledger's Break-Even Test
Start with the 48¢ Bills contract from January 17. Here's the math I'd run before clicking buy. Kalshi's own fee schedule isn't in my sources, so I'm using Lines.com's 2% Polymarket figure as a stand-in fee:
- Cost: 100 contracts at 48¢ = $48
- Gross return if the Bills win: $100
- Fee: 2% on the winning position = $2
- Net return: $98
- Break-even: $48 / $98 ≈ 49.0%
FanDuel's -100 needed 50% to break even. On this ticket, the exchange wins by about a point. Good. Write that down.
Now add a spread (hypothetical)
Now suppose 48¢ was only the last trade, and the cheapest ask you can actually fill is 50¢. That's an illustration, not the real order book on that game.
- Cost: 100 contracts at 50¢ = $50
- Net return: $98
- Break-even: $50 / $98 ≈ 51.0%
That's now worse than FanDuel's -100. A 2¢ spread erased the entire "no vig" advantage and then some. You paid for the exchange experience and got a worse price than a regulated book was already offering.
Convert both venues before you compare
Benzinga's conversion: a -160 favorite implies about 61.5% (160/260). Buying 100 YES contracts at 62¢ costs $62 and returns $100, a $38 profit before fees.
Put every price on either venue through that conversion first. If you compare American odds to cents in your head, it's easy to talk yourself into the worse number. If the formula is new to you, I break down the conversions in these odds-calculation hacks for escaping the 52.4% trap.
The Chiefs example, and the rule that comes out of it
Effortless Math walks through a comparison in which sportsbooks priced the Chiefs at about 58% and prediction markets at about 55–56%. A bettor whose model said 57% had a small edge buying YES at 56¢ on Kalshi. The same bettor had a negative edge paying 57.4% implied at DraftKings, and the vig came on top of that.
The same opinion produced a positive edge on one screen and a negative edge on the other. Only the price was different.
The rule: the exchange wins only when your number beats the contract price after fees and spread. If you don't have your own number, there's no edge to measure, only a feeling.
What You're Really Leaving on the Table
Here's the uncomfortable part. Most weekend bettors don't have a 57% model. They have an opinion and a team hat.
If you don't have a number that beats the contract after costs, the trade is a donation to whoever priced it better. On an exchange, that's a trader, and traders don't send you a birthday bonus.
To be fair, exchanges do offer something books don't. Action Network notes that you can sell a prediction-market position at any time, and its value moves with the contract price. A sportsbook's odds lock once it accepts your bet. That's real flexibility: you can bank a profit at halftime or cut a loser in the third quarter. Flexibility isn't an edge, though. Every exit runs you through the spread a second time.
Here's where I plant my flag: for 90% of bettors, line shopping earns more than handicapping. Look at Bills–Broncos again. On January 17, 2026, the same game showed a 50% Bills price at FanDuel and 48% at Kalshi. Prices differ between shops like that all the time, and you can collect the gap without out-trading anyone.
That's the payoff you're skipping each weekend you bet out of a single app. You don't need a model, a spreadsheet or a Discord full of sharps. You need a few minutes per bet and the willingness to open a second app. Every point of implied probability you save on your side is a point the book or the trader across from you doesn't get.
FanDuel — No Sweat First Bet up to $1,000 (code PBFD) FanDuel hung the Bills–Broncos line priced above, so it's a natural second screen to shop against your current app before this weekend's card. 21+ only, T&Cs apply. Bet Now
The Smarter Move This Weekend: Shop Lines, Then Check the Playthrough
Three steps, in order. Do them before your first bet on Saturday, not after your third loss on Sunday.
1. Convert, then shop. Before any bet, turn every price into implied probability using the -160 → 61.5% method. Then take the lowest implied probability for your side across regulated books. A point saved on every bet compounds over a season the way the hold compounds against you.
2. Rate the bonus by its math, not its headline. A bonus is only worth its playthrough math. Read the terms and pull two numbers: the bonus amount and the playthrough multiple. Multiply them to get the total you must wager before you can withdraw. Then multiply that total by the house's cut. Benzinga's -110 example works out to about 4.5% of handle, and that's your expected cost to clear the bonus. The bonus is worth roughly the headline amount minus that cost. A big headline with heavy playthrough can be worth less than a modest one with light playthrough. Run the numbers and the decision makes itself.
3. If you still want to trade, run the break-even test. Take the price you'll actually fill at after the spread, not the last trade, and divide it by your net payout after fees. Compare the result with your own number. If you can't name your number, take the single at the best regulated line. Parlays are a tax on hope, and so is a trade with no price behind it.
Pick one book you don't use yet, compare it with your current app on this weekend's card, and check its welcome offer against step 2. If the playthrough math holds up, that offer is the easiest point of value you'll find all season.
BetRivers — 2nd Chance Bet up to $500 (code PB500) If BetRivers isn't on your phone yet, it's a candidate for that new book — read the 2nd Chance terms and run the step-2 playthrough math before you deposit. 21+ only, T&Cs apply. Bet Now
This content is for adults 21+. Gambling is entertainment, not income, and no system beats the book. Set a bankroll before kickoff and keep to it. If betting stops being fun, stop.
Bottom line
Prediction markets don't remove the house edge. They reprice it. The vig becomes fees and spread, and the book is replaced by a sharper trader.
Kalshi's 100% quote on Bills–Broncos looked better than FanDuel's 104.1%. Then a 2% exchange-style fee and a hypothetical 2¢ spread pushed break-even from 49.0% to 51.0%, worse than FanDuel's -100.
Unless your own number beats the contract after costs, you'll get more by shopping lines at regulated books and judging every bonus by its playthrough. Discipline about price and bankroll is the only edge you control.