Short answer: unders hit more. Across 1,634 NFL player props graded against real 2025 closing lines, the under hit 53.5% and the over hit 46.5%. That 7-point gap is not a hot streak. It is structural, it shows up in every major prop market, and it is the single most useful thing a new prop bettor can learn.

Here is the data, why the gap exists, and — importantly — why it does not mean you should blindly hammer unders.

The data

We captured closing lines from DraftKings and FanDuel roughly one hour before kickoff for the 2025 regular season, weeks 1 through 9, then graded every prop against the actual box score. No pseudo-lines, no estimates: the real number bettors could have taken.

| Prop market | Under hit rate | Props graded |

|---|---|---|

| Rushing yards | 53.8% | 679 |

| Receptions | 53.7% | 1,334 |

| Receiving yards | 51.0% | 1,372 |

| Passing yards | 50.2% | 255 |

| All props | 53.5% | 1,634 |

For context, at standard -110 pricing you need 52.4% to break even. A blind under across every prop just barely clears that bar. A blind over loses badly.

Why the gap exists

The reason is the shape of football production, not sportsbook malice.

Player stat lines are right-skewed. A running back's floor is roughly zero. His ceiling is a 140-yard afternoon. That long right tail pulls his *average* above his *median* — the number he beats exactly half the time.

Sportsbooks price close to the expected value, which sits above the median. So the posted line is a number the player clears slightly less than half the time. Before you handicap a single matchup, the under has a small head start.

This is the same effect that shows up in baseball strikeout props and basketball scoring props. Football just has it worse because single-game variance is enormous with only 17 games a season.

What the gap does NOT mean

Here is where most articles on this topic go wrong. A 53.5% structural tilt does not mean "bet every under."

Break-even at -110 is 52.4%. A blind under strategy nets roughly one percentage point of edge before variance, and any vig above -110 erases it. In our data, the blind under actually returned -1.4% once real prices were applied, because unders are frequently juiced to -115 or -120 precisely because books know this.

The tilt is a starting point, not a strategy. What turns it into an edge:

• -Only bet where your projection agrees with the market. In our data, props where the model's number sat within 15% of the book's line hit 57.6% out of sample. Props where the model disagreed wildly hit 45.9%. Disagreement is usually your model missing information the book has, not edge.
• -Shop the line. DraftKings and FanDuel post different numbers on the same prop 48% of the time. Taking the higher line on an under improved the hit rate from 52.4% to 53.2% across 3,640 props.
• -Respect the price. An under at -105 and the same under at -125 are completely different bets.

The overs question

If unders hit 53.5%, are overs simply unbettable? No, but they are harder, and honesty matters here.

We tested every over angle we could construct on the same closing-line data: how far the projection sat above the line, the model's raw probability, price bands, player position, line size, and how often the player had cleared that number historically. We then split every promising result in half — weeks 1 through 5 versus weeks 6 through 9 — to see if it held.

Almost nothing survived. The one pocket that did: low rushing lines (30.5 yards or fewer) where the projection sat above the line, which hit 55.3% in the first half and 54.5% in the second across 82 bets.

A rule that looks brilliant on the full sample and collapses when you split it in half was never real. That is the discipline most prop content skips.

The trap: past hit rate against today's line

A common approach: "He has gone over this number in 13 of his last 17 games, so bet the over."

We tested exactly that. For every prop, we calculated how often the player had cleared *that week's actual line* in his prior games, then checked what the over did.

| He cleared today's line in... | Over actually hit |

|---|---|

| 70%+ of past games | 55.6% (n=72) |

| 60-69% | 46.1% (n=230) |

| 50-59% | 46.2% (n=504) |

| Under 50% | 46.8% (n=724) |

The top row looks like a signal until you split it: 72% in weeks 1 through 5, 44% in weeks 6 through 9. Same rule, opposite result. That is noise.

The reason is simple. Last season's line was not this week's line. The book sets today's number knowing the opponent, the game plan, the weather, and the injury report. A player's history of clearing *old* numbers is already priced into the *new* one.

How to actually use this

See the measured number on every prop

BetBlum shows the measured hit rate of picks like yours — same prop, same side — graded against real closing prices, not a model's opinion of itself. When our number says 56, that is what picks like it actually did.

Start your free 48-hour trial and see the honest number on every NFL prop this week.