Original research · August 24, 2026
Are Unders Better Than Overs on Player Props?
By Matt Downs
Bettors have argued about this for as long as player props have existed. The claim is that unders are the sharper side, because the public likes backing a player to do the thing and books shade the over to match.
It is a claim that can be counted rather than argued about. Our prop optimizer picks legs off the DFS board every night by one criterion, whether the devigged price says the row is worth taking, and it has no opinion about which arrow it ends up with. So the arrows it produces are a measurement.
The count: 64.3% of picked legs were unders
Across 882 legs the optimizer selected between July 29 and August 23 2026, 567 were unders and 315 were overs. That is 64.3%.
It is not one lopsided night carrying the average either. 21 of the 25 days landed above an even split, and the median day ran 67.7% unders. The four days that did not clear an even split came in at 46.7%, 47.2%, 48.8% and a dead 50.0%, so even the exceptions sat close to even rather than reversed.
A second and independent signal points the same way. Among legs that qualified on tracked whale money, meaning real size had already gone through at that price or better, 88 were unders and 42 were overs. That is 66%, from a different detector than the one that produced the main count.
The unders were not better bets
Here is the finding that changes what you should do with the count. Each picked leg carries the tool's devigged read on how often it should land. Averaged across the whole sample, the unders came out at 54.79% and the overs at 54.86%.
Those are the same number to within noise, and the overs are the side fractionally ahead. The lean is in how many rows clear, and it is absent from how much they clear by. A picked over is worth what a picked under is worth.
So the practical use of the count is narrow. It tells you which side of a board tends to hold more of the rows worth checking. It does not license taking an under because it is an under.
It is a market-level effect, not a board-level rule
Broken out by prop market, the tidy version falls apart. Batter singles ran 100% unders on 24 legs and pitcher hits allowed ran 96% on 27. Batter hits plus runs plus RBIs ran 86% on 142, and assists 77% on 57.
Then rebounds came in at 42% on 80 legs, batter total bases at 47% on 56, and pitcher strikeouts at 55% on 106. Those are the same nights and the same board.
By league it splits again. MLB legs ran 70.4% unders across 560 picks while WNBA legs ran 55.3% across 300. One sport is carrying most of the lean.
Under share by market, picked legs, 25 days
| Batter singles (n=26) | 100% under | |
|---|---|---|
| Pitcher hits allowed (n=27) | 96% under | |
| Batter hits + runs + RBIs (n=148) | 86% under | |
| Assists (n=57) | 77% under | |
| Pitcher outs recorded (n=55) | 71% under | |
| Points (n=78) | 58% under | |
| Pitcher strikeouts (n=111) | 55% under | |
| Batter total bases (n=57) | 47% under | |
| Rebounds (n=80) | 42% under |
The check that replaces picking a side
DFS pick'em apps price every leg the same. PrizePicks charges a flat -119 whichever side you take, and -119 needs 54.34% to break even, which you can confirm on the implied odds calculator. A six leg flex needs roughly 54.2% per leg.
So the question on any row is whether the devigged market price puts it above that line. On the board captured at 13:00 UTC on August 23 2026, 49 rows were eligible and five cleared 54.2%. All five were unders. The optimizer's six leg slip took those five plus a WNBA over reading 54.17%, just under the line, and the slip still priced at +1.91%.
The best row on that board read 54.73% against the 54.34% a flat -119 leg needs on its own. That is the size of the edge on a leg worth taking, and the arrow it happened to point was a consequence rather than a reason.
What this measurement cannot tell you
Three limits belong with the number. First, these are the legs the tool picked, so the sample is what cleared a price threshold rather than a census of every prop posted. A full board census would be a different and larger study, and the capture table does not store the rows that were rejected.
Second, it counts prices and not outcomes. Nothing here is a win rate, and no claim is made that these legs won.
Third, the per-leg percentages are the tool's own devigged consensus figures. The captures store one side of each book's quote, so those figures were not independently re-derived here, and no single book's price is presented as an edge on its own.
Frequently asked questions
Are unders better than overs on player props?
Not in edge. Across 882 optimizer-picked legs measured over 25 days, unders averaged 54.79% to hit and overs averaged 54.86%, which is the same number to within noise. Unders were more numerous among rows worth taking, at 64.3%, but a picked over was worth as much as a picked under.
Why do more unders show up as good bets?
The usual explanation is that recreational money prefers backing a player to exceed a number, so the over gets shaded and the under is left alone. That mechanism is not proven by this data, which measures where the prices ended up rather than why.
Does the under lean apply to every prop market?
No. Over the same 25 days, batter singles ran 100% unders and pitcher hits allowed ran 96%, while rebounds ran 42% and batter total bases ran 47%. The effect is market-specific, and MLB carries far more of it than the WNBA.
What number does a PrizePicks leg have to beat?
PrizePicks prices every leg at a flat -119, which implies 54.34% break-even. A six leg flex needs about 54.2% per leg. Any row whose devigged market probability sits above that line is worth taking regardless of which side it is.