CLEATZ / Prediction market analysis
The exchange label does not guarantee a better parlay.
A better deal comes down to the payout, every fee and the probability of winning. Here is what the loss data reveals about Kalshi, Polymarket US and DKeX, plus what Novig’s parlay pricing adds to the comparison.
Research checked October 20, 2026 · Historical snapshots and clearly labeled calculations
The reported results
What the prediction market parlay loss data shows
An InGame analysis reported substantial losses for parlay takers: the participants accepting the prices offered to them. Its figures provide a starting point for evaluating the economics of these products.
| Platform | Losses after fees | Reported loss rate | Scope |
|---|---|---|---|
| Kalshi | About $340 million | 14.4% | September 1 through the article’s snapshot |
| Polymarket US | About $69.6 million | 15.4% | Parlay launch through the snapshot |
| DKeX | About $60.8 million | 25.8%* | Published trade sample since parlay launch |
| Novig | Not established from the public data reviewed | Unavailable | Comparable parlay settlement results are missing |
*DKeX’s published loss and total-stake figures imply about 25.4%; the reported 25.8% may use a different settled-stakes denominator. That difference is unresolved. See the methodology notes below.
These are aggregate historical losses relative to reported stakes. They are not the percentage of customers who lost, a forecast of your results or a comparison of identical tickets.
A defined sportsbook benchmark
Illinois’ FY2026 report lists $4.9728 billion in parlay handle and $4.0936 billion in payouts. The difference implies approximately 17.7% realized hold. That provides context for the exchange figures, but a full fiscal year and shorter exchange samples have different sports, customers and reporting methods.
Comparing platforms requires more than a loss-rate leaderboard. Our prediction market app comparison provides broader platform context; the analysis here focuses specifically on parlays.
Where the cost comes from
Pricing, fees and the people setting the quote
There are two separate costs to examine: the price offered for the parlay and the fee charged to execute it. A low fee does not rescue an expensive quote. A competitive quote can also lose some of its advantage once fees are included.
Novig’s request-for-quote process illustrates the distinction. Registered liquidity providers compete to price a requested parlay, and the lowest contract price wins because it offers the customer the highest payout. Competition can improve a quote, but the best quote in that auction is not automatically the best available across other platforms.
A taker accepts a price; a maker supplies it. Those are trading roles. API access makes quoting less accessible to a typical app user, but it does not establish that every maker is an institution or every taker is recreational.
DraftKings has confirmed its market-making role
In its Q2 2026 shareholder letter, DraftKings said it operated a market maker integrated with DKeX and owned the brokerage, exchange and market maker. That documents the structure. It does not establish which trades its affiliate took or how much of the reported customer losses accrued to it.
Exchange fee revenue and market-maker trading gains are different money flows. For the wider distinction between activity, fees and revenue, see the CLEATZ prediction market revenue tracker.
The Novig question
Novig parlays include a fee in the quoted price
Novig’s published schedule distinguishes pregame straight trades, which have no trading fee, from parlays. Its parlay taker fee is 0.10 × P × (1 − P) per $1-payout contract, where P is the price before that fee. The fee is already included in the displayed parlay quote.
App and website users cannot quote parlays; API participants can make them. The maker side has no parlay fee under the published schedule.
A small charge per contract can still matter
Novig’s official example starts with a 10¢ price before fees and adds 0.9¢, producing a 10.9¢ displayed quote. For 100 contracts:
100 contracts at 10¢ each
0.10 × 0.10 × 0.90 × 100
$100 total payout if the parlay wins
Now assume the true chance of winning is exactly 10%. The expected payout is $10 against a $10.90 cost. Our calculation of the expected loss is:
This is a mathematical illustration under an explicit probability assumption. It is not Novig’s measured parlay loss rate. A different relationship between the underlying quote and the actual probability changes the result.
The fee is 9% of the price before fees, or 8.26% of the total paid. Those are different denominators for the same charge. Do not add the fee to an app quote that already includes it.
For more platform detail, visit our Novig review. The official fee schedule linked below is the source for the example on this page.
Why Novig has no loss-rate entry above
Predict Charts’ Novig outcomes analysis excludes parlays because the public feed does not expose their settlement results. Its straight-trade results cannot stand in for parlay results. Missing data is not evidence that Novig customers lose less or more.
An illustrative comparison
Could an exchange parlay still be cheaper?
Yes. Better pricing can reduce the disadvantage even when a bet still has negative expected value. The calculation below shows why the actual quote matters.
Same 50/50 selections. Different pricing assumptions.
Every leg is assumed to be independent and truly 50/50. The sportsbook model multiplies −110 odds for each leg. The Novig model starts with the mathematically fair combined price, then applies its published parlay fee.
These are hypothetical calculations, not observed quotes or customer results. The Novig model assumes no additional maker margin. Both models exclude promotions, rounding and execution constraints.
| Legs | −110 for every leg | Fair price + Novig fee |
|---|---|---|
| 2 |
8.88%
|
6.98%
|
| 3 |
13.03%
|
8.05%
|
| 4 |
16.98%
|
8.57%
|
| 5 |
20.75%
|
8.83%
|
| 6 |
24.36%
|
8.96%
|
For n legs, the −110 model’s expected loss is 1 − (21/22)n. For the fee model, P = 0.5n and expected loss is 0.10(1 − P) ÷ [1 + 0.10(1 − P)]. These calculations use total cash spent as the denominator.
The five-leg example produces an expected loss of 20.75% when −110 prices compound, versus 8.83% when a fair combined price receives Novig’s fee. That does not prove Novig offers the assumed fair quote. It explains why comparing the actual payout is necessary.
A fair five-leg 50/50 parlay would pay 32 times the stake on a win. More legs make winning less likely; an insufficient payout relative to that probability creates the disadvantage.
Same-game parlays need a separate analysis. Related selections are not independent, so multiplying their individual probabilities can produce the wrong answer.
A practical price check
How to compare the same parlay across platforms
Compare the total payout for the same total cash commitment. A useful comparison matches:
- The exact selections. Match every team, player, market and threshold.
- The time. Collect quotes close together and keep pregame and live comparisons separate.
- The full cost. Include fees added at checkout and recognize fees already inside a quote.
- The available stake. Confirm the quoted price supports the amount you intend to spend.
- The settlement rules. Compare voids, pushes and injury treatment.
- The promotion. Evaluate standard prices separately from boosts, credits and signup offers.
Compare the dollars returned
If identical tickets cost $10 including every fee, a $75 total winning payout is 7.1% higher than a $70 payout. That is a price improvement, not a 7.1-percentage-point increase in the chance of winning.
A fair comparison can include Novig even without a public history of its parlay outcomes. It would measure executable payouts on matching tickets. The historical results and calculations on this page do not establish a winner on that test.
Questions bettors ask
Prediction market parlay FAQs
Does a 15% loss rate mean 15% of bettors lost?
No. A loss rate based on stakes describes aggregate dollars lost relative to dollars staked. It does not tell us the share of customers who were profitable.
Does a contract price equal its true probability?
A price implies a break-even threshold, but it is not proof of the true probability. Fees, available liquidity and the prices participants are willing to accept can all affect the amount paid.
Is the 8.26% Novig example its actual hold?
No. It is an illustrative expected loss calculated from a stated fee example while assuming a true 10% chance of winning. A comparable realized Novig parlay loss rate is not established by the public data reviewed.
Can we rank every platform using these figures?
The reported samples show historical outcomes under different conditions. Ranking value on a specific ticket requires matching selections, timing, stake, fees and settlement terms.
More CLEATZ prediction market coverage
Explore the platforms, their published costs and the broader activity behind the market.
Jason Ziernicki is the founder of CLEATZ and has worked in the online sports betting industry since 2012. At CLEATZ, he focuses on public betting data, sportsbook line movement, prediction markets and data-driven sports betting tools.
