NFL prop holds ran from 4.54% to 6.52% by book at the close this season.
01 · What hold is
Hold is the rent on every bet
Add the implied percentages of both sides of a market and you get a number above 100%. The hold turns that overlap into the share of money the book keeps: a −110 / −110 market holds 4.55%. You pay it whether you win or lose, because it lives inside the price. Taking the book's cut out of a price walks through the arithmetic.
A lower hold does two things for you. Your winners pay more, and the fair price you are trying to beat sits closer to the price you can actually bet. That is why sharp bettors care which book posts a market before they care about the market itself.
02 · By book
The same NFL props, priced by seven books
Median closing hold on NFL player props
Books with 500 or more over and under pairs, 2026-09-10 to 2026-10-26
| Book | Pairs | Median hold |
|---|---|---|
| Pinnacle | 3,856 | 4.54% |
| DraftKings | 9,506 | 5.74% |
| FanDuel | 3,722 | 5.75% |
| ESPN BET | 6,969 | 6.35% |
| Bovada | 8,088 | 6.37% |
| Hard Rock Bet | 8,014 | 6.48% |
| BetMGM | 4,859 | 6.52% |
On a coin-flip prop, the cut at the cheapest book costs about $4.54 per $100 in expectation. At the dearest it costs about $6.52. Over a season of bets, that difference is real money for no extra skill.
03 · By market
Every prop market charges about the same
| Market | Pairs | Median hold |
|---|---|---|
| Receiving yards | 9,305 | 6.29% |
| Receptions | 7,624 | 6.12% |
| Rushing yards | 4,595 | 6.19% |
| Longest reception | 3,981 | 6.48% |
| Passing yards | 2,264 | 6.31% |
| Tackles and assists | 1,945 | 6.10% |
| Rush attempts | 1,588 | 5.87% |
The spread across markets is narrow: every market with real volume sat between about 5.9% and 6.5%. The book you bet with moves the cut far more than the market you pick. Low-hold opportunities do exist, but they are individual prices, not whole markets, which is why our low-hold desk lists pairs one at a time.
Before you bet
What this means for your next bet
- Price both sides at the book you are using. If they add up to much more than −110 / −110, look elsewhere.
- Keep an account at the cheapest book that takes your bets.
- Use a low-hold book's price as your fair reference, even if you bet somewhere else.
- Treat a low hold as a better price, not as a reason to bet. You still need an edge.
Method
How we measured this
Every NFL player prop over and under posted by the same book at the same line in the closing snapshot, paired. Hold is 1 minus 1 divided by the sum of the two implied percentages. Figures are medians per book (books with 500 or more pairs) and per market (markets with 1,000 or more pairs).
- Sample
- 45,204 over and under pairs across 95 games
- Window
- Sep 10, 2026 to Oct 26, 2026
- Measured
- Oct 29, 2026
- Source
- Dynatyze closing-price archive
Questions3 questions
What does low hold mean in sports betting?
Hold is the share of money a book keeps when both sides of a market are bet evenly. A low-hold market is one where both prices together leave the book a small margin, so each price sits closer to fair.
Which sportsbook has the lowest hold on player props?
On the NFL player props we measured at the close this season, the median hold was lowest at Pinnacle (4.54%) and highest at BetMGM (6.52%) among seven books with enough volume. Holds move by sport, market and time, so check the prices in front of you.
How do you calculate hold?
Convert both prices to implied percentages and add them. Hold is 1 minus 1 divided by that total. Two sides at −110 add to 104.76%, so the hold is 1 minus 1 divided by 1.0476, or 4.55%.
Responsible gaming
Dynatyze is analytics and education for people 21 and over. It is not a sportsbook and never takes a bet. Every number on this page describes the past or the arithmetic of a price; none of it promises a result, and results vary.
Set a budget before you bet and keep to it. If gambling stops being fun, help is free and confidential at 1-800-GAMBLER, and our responsible gaming page lists more ways to get support.
Continue reading
Research and education only. Results vary, and nothing here promises a profit. Every number from our own data names its sample and its dates; every worked example is the arithmetic of the prices shown.