Player props · 2023–25 closing lines · 2026 live
Prop Lab: Do Trait Matchups Beat Vegas?
Fast receivers against slow corners, agile route runners against corners who struggle with them, big backs against light boxes: we wrote down 23 specific angles before looking at a single result, then searched 31,114 more combinations of player traits, defenders and conditions against 31,044 closing prop lines. Every idea had to work in 2023–24 and again in 2025, a season we did not touch until the list was frozen.
The short answer: nothing cleared our bar for an edge. The owner’s speed and agility ideas lost money in 2025. The best combination the search found in real data was weaker than what the same search finds in shuffled, meaningless data most of the time. A handful of angles are “promising” and tracked live from Week 6, mostly because unders were unusually strong in 2025.
Research: what we found
Our projection model, tens of thousands of trait-based angles, matchup splits, weather and injuries, all tested against real consensus prop lines. Here is what the research found, one topic at a time.
How prop markets are priced
Over/under tilts, the juice, whether a projection model can beat closing lines, and why most systems are luck.
- Verdict: Mixed7 min read
Do NFL Player Props Hit the Over or Under More Often?
Unders win more often in most NFL prop markets. We graded three seasons of consensus lines to see why, and whether betting every under actually makes money.Updated - Verdict: Myth8 min read
Can a Projection Model Beat Closing NFL Player Prop Lines?
We pointed our fantasy projection model at three seasons of NFL prop lines. The market was more accurate on every major market. One narrow split is worth watching, and one data bug is worth remembering.Updated - Verdict: Myth7 min read
Why Most Sports Betting Systems Are Luck: 31,114 Tests and Zero Survivors
Test enough ideas and some will look brilliant by chance. We searched NFL player props for trait-based angles, then measured exactly how good pure noise looks. Noise won.Updated
Player traits and matchup myths
Combine speed, WR vs CB matchups and defense vs position, tested against real lines and projections.
- Verdict: Myth7 min read
Does 40-Yard Dash Speed Predict Receiving Yards Props?
Speed sells. We wrote down the most popular combine-based prop ideas before looking at the results, then tested them on a season they had never seen. None of them held.Updated - Verdict: Mixed8 min read
WR vs CB Matchups and Player Props: Mostly Guesswork, One Angle to Watch
Receiver-versus-corner matchups drive a lot of prop talk. We checked how often anyone can even identify the right corner, then tested size, age and speed gaps against real lines.Updated - Verdict: Myth8 min read
Is Defense vs Position Useful in Fantasy Football? Mostly Not
Fantasy points allowed by position is the most-used matchup stat in fantasy football. We tested whether it repeats, whether it predicts, and whether any matchup angle beats a good projection.Updated
Weather and injury situations
Wind, calm weather, dome teams outdoors and Questionable running backs.
- Verdict: Promising8 min read
Does Wind Affect QB Passing Yards Props? Yes, With a Big Catch
Strong wind is the most physical, most believable prop angle in football. Our data backs it up. The catch is that the test used wind no bettor could know before kickoff.Updated - Verdict: Exploratory6 min read
Questionable Running Backs and Rushing Yards Props: A Real Lean or a Lucky Slice?
Running backs who play through a Questionable tag went under their rushing line far more often than the average back. We found it after a lot of looking, so here is the full picture, sample sizes included.Updated
Why digging into specifics finds fake edges
Slice prop results finely enough (this corner, against receivers this fast, in this weather) and some slice will look brilliant by luck alone. Test 31,114 slices and about 1 in 20 clears a 5% significance test with no real effect behind it: here, 716 did. Sports-betting “trends” are mostly this.
How we guard against it:
- Pre-registration. Every hypothesis, threshold and rule was committed to our methods document before any result was computed.
- Holdout. Ideas are found on 2023–24 and judged once on 2025.
- Permutation benchmark. We shuffle the over/under results within each market and week 200 times and re-run the whole search, to see how good the best “edge” looks in pure noise.
- Shrinkage for individual corners. A corner’s record is pulled toward the league average until he has enough matchups to earn his own number.
- Pregame data only. No feature uses anything learned after kickoff. Historical weather is the observed game weather, which is optimistic; live calls use forecasts.
The search funnel
31,114
combinations tested (both sides, 150+ bets each in 2023–24)
716
looked good in 2023–24 (p < 0.05, before correcting for the number of tests)
0
survived the false-discovery correction
12 of 20
top combinations held above 52.4% in 2025; 10 also beat blind betting
14
angles tracked live from 2026 Week 6 (0 edges)
The best edge you’d find in random data
Each grey bar counts shuffled-data searches by the best result they found. The best real 2023–24 combination (65.5% on 177 bets, z = 3.50) sounds impressive, but 129 of 200 searches on meaningless data did as well or better. To count as an edge, a result has to beat 4.14, the line 95% of random searches stay under.
The owner’s hypotheses
Written down before any result was computed. “Blind” is how often the same side of that market won without any angle. Break-even at −110 is 52.4%.
| Hypothesis | Bet | Discovery 2023–24 | Holdout 2025 | Holdout 95% CI | Verdict |
|---|---|---|---|---|---|
| H1 · Fast young WR vs a corner beaten by speed, calm airowner’sWR forty ≤ 4.43, under 30, wind < 10 mph, and a likely corner in the top quartile of "beaten by fast WRs" (shrunk, n ≥ 3) · weather observed (optimistic) | Receiving yards over | 51.5% (118–111)blind 48.8% · ROI −3.7% | 45.6% (52–62)blind 46.4% · ROI −14.7% | 36.8%–54.8% | Myth-bustedholdout 45.6% ≤ 52.4% |
| H1 · Fast young WR vs a corner beaten by speed, calm airowner’sWR forty ≤ 4.43, under 30, wind < 10 mph, and a likely corner in the top quartile of "beaten by fast WRs" (shrunk, n ≥ 3) · weather observed (optimistic) | Longest reception over | no lines before 2025 | 46.4% (52–60)blind 47.6% · ROI −13.1% | 37.5%–55.6% | Myth-bustedholdout 46.4% ≤ 52.4% |
| H1b · H1 with a sub-4.30 fortyowner’sAs H1 with forty < 4.3 · weather observed (optimistic) | Receiving yards over | 62.5% (5–3)blind 48.8% · ROI +15.5% | 33.3% (2–4)blind 46.4% · ROI −38.2% | 9.7%–70.0% | Myth-bustedholdout n 6 < 60 |
| H1b · H1 with a sub-4.30 fortyowner’sAs H1 with forty < 4.3 · weather observed (optimistic) | Longest reception over | no lines before 2025 | 33.3% (2–4)blind 47.6% · ROI −38.7% | 9.7%–70.0% | Myth-bustedholdout n 6 < 60 |
| H2 · Agile WR vs a corner beaten by agile receiversowner’sWR 3-cone ≤ 6.82 or shuttle ≤ 4.15, and a likely corner in the top quartile of "beaten by agile WRs" for that stat | Receptions over | 51.1% (112–107)blind 47.6% · ROI −4.3% | 37.5% (48–80)blind 45.2% · ROI −29.8% | 29.6%–46.1% | Myth-bustedholdout 37.5% ≤ 52.4% |
| H2 · Agile WR vs a corner beaten by agile receiversowner’sWR 3-cone ≤ 6.82 or shuttle ≤ 4.15, and a likely corner in the top quartile of "beaten by agile WRs" for that stat | Receiving yards over | 52.9% (117–104)blind 48.8% · ROI −0.8% | 39.6% (42–64)blind 46.4% · ROI −25.7% | 30.8%–49.1% | Myth-bustedholdout 39.6% ≤ 52.4% |
| H3 · Breakaway WR (top-speed proxy)owner’sWR in the top quartile of 20+ yd reception rate over his previous 16 games (≥ 0.231); in-game top speed is not available legally | Receiving yards over | 48.6% (376–397)blind 48.8% · ROI −9.3% | 47.2% (158–177)blind 46.4% · ROI −11.8% | 41.9%–52.5% | Myth-bustedholdout 47.2% ≤ 52.4% |
| H3 · Breakaway WR (top-speed proxy)owner’sWR in the top quartile of 20+ yd reception rate over his previous 16 games (≥ 0.231); in-game top speed is not available legally | Longest reception over | no lines before 2025 | 47.5% (159–176)blind 47.6% · ROI −11.4% | 42.2%–52.8% | Myth-bustedholdout 47.5% ≤ 52.4% |
| H3b · Breakaway WR vs a big-play defenseowner’sH3 and the opponent's prior 20+ yd completions per attempt in its top quartile (≥ 0.101) | Receiving yards over | 48.6% (101–107)blind 48.8% · ROI −9.2% | 47.0% (39–44)blind 46.4% · ROI −12.1% | 36.6%–57.6% | Myth-bustedholdout 47.0% ≤ 52.4% |
| H3b · Breakaway WR vs a big-play defenseowner’sH3 and the opponent's prior 20+ yd completions per attempt in its top quartile (≥ 0.101) | Longest reception over | no lines before 2025 | 47.0% (39–44)blind 47.6% · ROI −12.2% | 36.6%–57.6% | Myth-bustedholdout 47.0% ≤ 52.4% |
In-game top speed (the “23.5 mph” idea) is not available legally for free, so H3 uses each receiver’s share of catches gaining 20+ yards over his previous 16 games as the closest proxy. Only 173 of 257 WRs with lines ran a combine 40; the owner’s sub-4.30 cut leaves almost no bets.
Sibling hypotheses
Age, size and speed mismatches, rookie corners, slow safeties, big backs vs light boxes, tight ends vs linebackers, and arm strength in wind. Also pre-registered.
| Hypothesis | Bet | Discovery 2023–24 | Holdout 2025 | Holdout 95% CI | Verdict |
|---|---|---|---|---|---|
| S1 · Young WR vs ageing cornersWR aged 26 or younger and the likely corners average 30+ | Receiving yards over | 50.9% (57–55)blind 48.8% · ROI −5.1% | 50.6% (40–39)blind 46.4% · ROI −5.0% | 39.8%–61.4% | Myth-bustedholdout 50.6% ≤ 52.4% |
| S2 · Height mismatch (4+ in)WR height − mean likely-corner height ≥ 4 in | Receiving yards over | 49.3% (221–227)blind 48.8% · ROI −7.7% | 51.9% (111–103)blind 46.4% · ROI −2.5% | 45.2%–58.5% | Myth-bustedholdout 51.9% ≤ 52.4% |
| S2b · Weight mismatch (25+ lb)WR weight − mean likely-corner weight ≥ 25 lb | Receptions over | 44.1% (179–227)blind 47.6% · ROI −16.3% | 44.5% (98–122)blind 45.2% · ROI −13.7% | 38.1%–51.2% | Myth-bustedholdout 44.5% ≤ 52.4% |
| S3 · WR faster than his corners (0.08 s)WR forty ≤ mean likely-corner forty − 0.08 (both measured) | Receiving yards over | 50.0% (225–225)blind 48.8% · ROI −6.5% | 41.3% (97–138)blind 46.4% · ROI −22.7% | 35.2%–47.7% | Myth-bustedholdout 41.3% ≤ 52.4% |
| S3b · Corners faster than the WR (0.08 s)Mean likely-corner forty ≤ WR forty − 0.08 (both measured) | Receiving yards under | 53.0% (341–302)blind 51.2% · ROI −0.5% | 57.3% (199–148)blind 53.6% · ROI +7.5% | 52.1%–62.4% | Promisingdoes not beat the permutation benchmark |
| S4 · Rookie corner vs veteran WRA likely corner in his rookie season and a WR with 4+ years of experience | Receiving yards over | 47.3% (133–148)blind 48.8% · ROI −11.5% | 41.3% (38–54)blind 46.4% · ROI −22.6% | 31.8%–51.5% | Myth-bustedholdout 41.3% ≤ 52.4% |
| S5 · Fast WR vs slow safetiesWR forty ≤ 4.43 and the opponent's top-2 safeties average a 4.55+ forty | Receiving yards over | 43.6% (147–190)blind 48.8% · ROI −18.4% | 46.1% (59–69)blind 46.4% · ROI −13.8% | 37.7%–54.7% | Myth-bustedholdout 46.1% ≤ 52.4% |
| S5 · Fast WR vs slow safetiesWR forty ≤ 4.43 and the opponent's top-2 safeties average a 4.55+ forty | Longest reception over | no lines before 2025 | 50.8% (64–62)blind 47.6% · ROI −5.0% | 42.2%–59.4% | Myth-bustedholdout 50.8% ≤ 52.4% |
| S6 · Big, fast RB vs light boxesRB speed score ≥ 105.9 and the opponent's prior light-box rate in its top quartile (≥ 0.589) | Rushing yards over | 51.0% (75–72)blind 46.1% · ROI −4.8% | 46.2% (78–91)blind 47.3% · ROI −13.4% | 38.8%–53.7% | Myth-bustedholdout 46.2% ≤ 52.4% |
| S7 · Fast TE vs slow linebackersTE forty ≤ 4.63 and the opponent's top-2 linebackers average a 4.70+ forty | Receiving yards over | 47.1% (16–18)blind 48.8% · ROI −11.8% | 38.1% (8–13)blind 46.4% · ROI −28.9% | 20.8%–59.1% | Myth-bustedholdout n 21 < 60 |
| S7b · Big TE vs short linebackersTE 6'5"+ and 250+ lb, and the opponent's top-2 linebackers average 6'1" or shorter | Receptions over | 56.7% (59–45)blind 47.6% · ROI +6.9% | 45.9% (51–60)blind 45.2% · ROI −13.8% | 37.0%–55.2% | Myth-bustedholdout 45.9% ≤ 52.4% |
| S8 · Weak-armed QB in windQB trailing max air distance in the bottom tercile (≤ 57.0 yds) and wind ≥ 10 mph · weather observed (optimistic) | Passing yards under | 57.9% (44–32)blind 50.0% · ROI +8.0% | 69.6% (16–7)blind 49.7% · ROI +30.3% | 49.1%–84.4% | Myth-bustedholdout n 23 < 60 |
| S8c · Control: strong-armed QB in windQB trailing max air distance in the top tercile (≥ 60.6 yds) and wind ≥ 10 mph · weather observed (optimistic) | Passing yards under | 56.9% (37–28)blind 50.0% · ROI +6.8% | 61.8% (21–13)blind 49.7% · ROI +16.2% | 45.0%–76.1% | Myth-bustedholdout n 34 < 60 |
Follow-up: does corner age matter?
The owner’s refinement of S3b: a corner over 29 has lost speed, so his combine 40 is stale. Three pre-registered variants drop older corners, age-adjust their 40 (+0.02 s a year after 27, a heuristic), or keep only corners 27 and younger. They pick almost the same bets as S3b and land within a point of it. The parent S3b had already been scored on 2025 when these variants were registered, so the holdout is not blind for them; the live record is the real test.
| Hypothesis | Bet | Discovery 2023–24 | Holdout 2025 | Holdout 95% CI | Verdict |
|---|---|---|---|---|---|
| S3b · Corners faster than the WR (0.08 s)Mean likely-corner forty ≤ WR forty − 0.08 (both measured) | Receiving yards under | 53.0% (341–302)blind 51.3% · ROI −0.5% | 57.3% (199–148)blind 53.6% · ROI +7.5% | 52.1%–62.4% | Promisingdoes not beat the permutation benchmark |
| S3b-a · Faster corners, dropping corners 30+owner’sAs S3b, after dropping likely corners aged 30 or older (unknown age dropped) | Receiving yards under | 52.7% (316–284)blind 51.3% · ROI −1.2% | 57.9% (184–134)blind 53.6% · ROI +8.4% | 52.4%–63.2% | Promisingdoes not beat the permutation benchmark |
| S3b-b · Faster corners, 40 aged +0.02 s/yr after 27owner’sAs S3b with each corner's 40 + 0.02 s per year over 27 (heuristic, not a measured aging curve) | Receiving yards under | 52.6% (283–255)blind 51.3% · ROI −1.4% | 57.1% (169–127)blind 53.6% · ROI +7.0% | 51.4%–62.6% | Promisingdoes not beat the permutation benchmark |
| S3b-c · Faster corners, only corners 27 or youngerowner’sAs S3b using only likely corners aged 27 or younger | Receiving yards under | 53.0% (289–256)blind 51.3% · ROI −0.5% | 57.8% (177–129)blind 53.6% · ROI +8.3% | 52.2%–63.2% | Promisingdoes not beat the permutation benchmark |
The search’s top 20, tested once on 2025
All 20 are unders. Lines shade toward overs, and 2025 was an under-friendly year, so “held up” needs a second look: the “beat blind betting” column compares each with simply betting every line of that market the same way.
| # | Combination (frozen before 2025 was looked at) | Bet | Discovery 2023–24 | Holdout 2025 | Beat blind betting? | Verdict |
|---|---|---|---|---|---|---|
| 1 | Weight: middle third (WR 192–206, TE 246–251) · 20+ yd completions allowed: most third (WR > 9.6%, TE > 9.7%) · Temperature (observed): dome | Receptions under | 65.5% (116–61)blind 52.4% · ROI +21.4% | 55.6% (50–40)blind 54.8% · ROI +0.6% | Yes, by 0.8 pts | Promising |
| 2 | Height: shortest third (WR ≤ 72, TE ≤ 76) · Likely corners’ age: youngest third (WR ≤ 25.1) · Venue: away | Receiving yards under | 62.7% (175–104)blind 51.2% · ROI +17.7% | 58.2% (53–38)blind 53.6% · ROI +9.7% | Yes, by 4.6 pts | Promising |
| 3 | Broad jump: longest third (WR > 126, TE > 124) · Likely corners’ yds/target allowed: middle third (WR 6.91–7.40) · Roof: outdoors | Receptions under | 66.0% (103–53)blind 52.4% · ROI +24.3% | 59.6% (59–40)blind 54.8% · ROI +5.2% | Yes, by 4.8 pts | Promising |
| 4 | Height: shortest third (WR ≤ 72, TE ≤ 76) · Likely corners’ yds/target allowed: stingiest third (WR ≤ 6.91) · Temperature (observed): 40°F or warmer | Receiving yards under | 61.9% (192–118)blind 51.2% · ROI +16.1% | 55.6% (75–60)blind 53.6% · ROI +4.8% | Yes, by 2.0 pts | Promising |
| 5 | Height: shortest third (WR ≤ 72, TE ≤ 76) · WR − corner height gap: smallest gap (WR ≤ 0.0) · Temperature (observed): 40°F or warmer | Receiving yards under | 58.8% (388–272)blind 51.2% · ROI +10.3% | 54.8% (193–159)blind 53.6% · ROI +3.1% | Yes, by 1.2 pts | Promising |
| 6 | Likely corners’ height: shortest third (WR ≤ 71.5) · Game total: 48 or higher | Receptions under | 62.0% (170–104)blind 52.4% · ROI +14.8% | 55.7% (108–86)blind 54.8% · ROI +2.7% | Yes, by 0.9 pts | Promising |
| 7 | Short shuttle: slowest third (WR > 4.29, TE > 4.37) · Likely corners’ age: middle third (WR 25.1–27.0) | Receiving yards under | 64.2% (115–64)blind 51.2% · ROI +20.5% | 56.3% (40–31)blind 53.6% · ROI +6.0% | Yes, by 2.7 pts | Promising |
| 8 | 3-cone: slowest third (WR > 7.01, TE > 7.12) · Rookie likely corner: no · Temperature (observed): 40°F or warmer | Receptions under | 63.1% (135–79)blind 52.4% · ROI +16.3% | 51.7% (45–42)blind 54.8% · ROI −4.1% | No | Myth-busted |
| 9 | Age: 25–28 · WR − corner height gap: smallest gap (WR ≤ 0.0) · Temperature (observed): 40°F or warmer | Receiving yards under | 62.1% (162–99)blind 51.2% · ROI +16.3% | 51.7% (93–87)blind 53.6% · ROI −2.8% | No | Myth-busted |
| 10 | 40-yard dash: fastest third (WR ≤ 4.42, TE ≤ 4.62) · Likely corners’ height: shortest third (WR ≤ 71.5) · Venue: home | Receiving yards under | 63.3% (126–73)blind 51.2% · ROI +18.8% | 45.1% (46–56)blind 53.6% · ROI −14.8% | No | Myth-busted |
| 11 | Age: under 25 · Likely corners’ age: youngest third (WR ≤ 25.1) · Venue: away | Receiving yards under | 62.2% (150–91)blind 51.2% · ROI +17.0% | 53.1% (43–38)blind 53.6% · ROI +0.3% | No | Promising |
| 12 | Role (pregame): outside_x · Game total: 48 or higher | Receptions under | 62.4% (138–83)blind 52.4% · ROI +15.2% | 58.0% (76–55)blind 54.8% · ROI +7.0% | Yes, by 3.2 pts | Promising |
| 13 | Weight: middle third (210–218) · Temperature (observed): 40°F or warmer | Rushing yards under | 61.3% (171–108)blind 53.9% · ROI +14.9% | 47.7% (73–80)blind 52.8% · ROI −10.0% | No | Myth-busted |
| 14 | Weight: middle third (210–218) · Team spread: underdog by 3+ | Rushing yards under | 62.9% (124–73)blind 53.9% · ROI +17.8% | 49.1% (55–57)blind 52.8% · ROI −7.8% | No | Myth-busted |
| 15 | NGS cushion: tightest third (WR ≤ 5.97, TE ≤ 5.87) · WR − corner height gap: smallest gap (WR ≤ 0.0) · Roof: outdoors | Receiving yards under | 61.8% (144–89)blind 51.2% · ROI +16.0% | 56.3% (67–52)blind 53.6% · ROI +6.1% | Yes, by 2.7 pts | Promising |
| 16 | 40-yard dash: fastest third (WR ≤ 4.42, TE ≤ 4.61) · Likely corners’ height: shortest third (WR ≤ 71.5) · Venue: home | Receptions under | 62.7% (121–72)blind 52.4% · ROI +16.2% | 51.5% (50–47)blind 54.8% · ROI −6.9% | No | Myth-busted |
| 17 | Weight: middle third (WR 192–206, TE 246–251) · Rookie likely corner: no · Team spread: favourite by 3+ | Receptions under | 59.9% (218–146)blind 52.4% · ROI +12.4% | 52.9% (101–90)blind 54.8% · ROI −2.1% | No | Promising |
| 18 | Likely corners’ age: youngest third (WR ≤ 25.1) · Venue: away | Receiving yards under | 58.5% (325–231)blind 51.2% · ROI +9.7% | 60.0% (132–88)blind 53.6% · ROI +13.2% | Yes, by 6.4 pts | Promising |
| 19 | Linebackers’ forty: middle third (4.54–4.63) · Team spread: underdog by 3+ | Rushing yards under | 61.8% (139–86)blind 53.9% · ROI +15.9% | 52.3% (57–52)blind 52.8% · ROI −1.8% | No | Myth-busted |
| 20 | 20+ yd catch rate: lowest third (WR ≤ 14.4%, TE ≤ 9.5%) · Game total: 48 or higher | Receptions under | 61.9% (135–83)blind 52.4% · ROI +13.1% | 47.4% (102–113)blind 54.8% · ROI −11.2% | No | Myth-busted |
Promising angles, tracked live
Tracked, not recommended. Calls lock about an hour before kickoff and can never be edited. None is an edge: each failed the permutation benchmark or the 2025 confidence bar.
| Angle | Bet | Holdout 2025 | Live 2026 (from Week 6) |
|---|---|---|---|
| Pre-registered: S3b · Corners faster than the WR (0.08 s) | Receiving yards under | 57.3% (199–148)blind 53.6% · ROI +7.5% | no graded calls yet |
| Search: Weight: middle third (WR 192–206, TE 246–251) · 20+ yd completions allowed: most third (WR > 9.6%, TE > 9.7%) · Temperature (observed): dome | Receptions under | 55.6% (50–40)blind 54.8% · ROI +0.6% | no graded calls yet |
| Search: Height: shortest third (WR ≤ 72, TE ≤ 76) · Likely corners’ age: youngest third (WR ≤ 25.1) · Venue: away | Receiving yards under | 58.2% (53–38)blind 53.6% · ROI +9.7% | no graded calls yet |
| Search: Broad jump: longest third (WR > 126, TE > 124) · Likely corners’ yds/target allowed: middle third (WR 6.91–7.40) · Roof: outdoors | Receptions under | 59.6% (59–40)blind 54.8% · ROI +5.2% | no graded calls yet |
| Search: Height: shortest third (WR ≤ 72, TE ≤ 76) · Likely corners’ yds/target allowed: stingiest third (WR ≤ 6.91) · Temperature (observed): 40°F or warmer | Receiving yards under | 55.6% (75–60)blind 53.6% · ROI +4.8% | no graded calls yet |
| Search: Height: shortest third (WR ≤ 72, TE ≤ 76) · WR − corner height gap: smallest gap (WR ≤ 0.0) · Temperature (observed): 40°F or warmer | Receiving yards under | 54.8% (193–159)blind 53.6% · ROI +3.1% | no graded calls yet |
| Search: Likely corners’ height: shortest third (WR ≤ 71.5) · Game total: 48 or higher | Receptions under | 55.7% (108–86)blind 54.8% · ROI +2.7% | no graded calls yet |
| Search: Short shuttle: slowest third (WR > 4.29, TE > 4.37) · Likely corners’ age: middle third (WR 25.1–27.0) | Receiving yards under | 56.3% (40–31)blind 53.6% · ROI +6.0% | no graded calls yet |
| Search: Age: under 25 · Likely corners’ age: youngest third (WR ≤ 25.1) · Venue: away | Receiving yards under | 53.1% (43–38)blind 53.6% · ROI +0.3% | no graded calls yet |
| Search: Role (pregame): outside_x · Game total: 48 or higher | Receptions under | 58.0% (76–55)blind 54.8% · ROI +7.0% | no graded calls yet |
| Search: NGS cushion: tightest third (WR ≤ 5.97, TE ≤ 5.87) · WR − corner height gap: smallest gap (WR ≤ 0.0) · Roof: outdoors | Receiving yards under | 56.3% (67–52)blind 53.6% · ROI +6.1% | no graded calls yet |
| Search: Weight: middle third (WR 192–206, TE 246–251) · Rookie likely corner: no · Team spread: favourite by 3+ | Receptions under | 52.9% (101–90)blind 54.8% · ROI −2.1% | no graded calls yet |
| Search: Likely corners’ age: youngest third (WR ≤ 25.1) · Venue: away | Receiving yards under | 60.0% (132–88)blind 53.6% · ROI +13.2% | no graded calls yet |
Exploratory angles: found after looking, proving live
Five angles our team found by digging through the same 2023–25 prop lines, after about 100 looks. Their history is in-sample, including 2025, so it is not evidence: the best of 100 slices always looks good (see the random-data chart above). We wrote the exact rules down on 2026-10-07 and track them live from Week 6, using forecasts and pregame injury reports only.
Model 20+ pts below the market on rushing exploratory — found after looking; proving live
Market no-vig P(over) − our model P(over) ≥ 0.20 (live projection at call time); any position with a line.
History (in-sample): 2023 60.8% (n 209, ROI +13.8%); 2024 57.8% (n 173, +9.2%); 2025 55.2% (n 172, +4.0%). The 5–10 and 10–20 point buckets are weaker or inconsistent.
Live 2026: no graded calls yet
Questionable RB exploratory — found after looking; proving live
RB listed Questionable on the latest pregame report at call time, with a line; rush attempts logged as a sibling call. Tags: script7 (|spread| ≥ 7), practice:<last practice>.
History (in-sample): 2023–24 59.7% (n 134); 2025 62.7% (n 51). Spread ≥ 7 either way 34 of 46; Limited 61.9% (n 113), DNP 77.8% (n 18), Full 54.0%.
Live 2026: no graded calls yet
Forecast wind 16+ mph exploratory — found after looking; proving live
Outdoor or open-roof game with forecast sustained wind ≥ 16 mph at kickoff → QB passing yards under.
History (in-sample): Observed wind: 16–18 mph 61.9% / 68.4%; 19+ mph 71.4% / 90.0%; 40 of 57 overall (2023–24 / 2025).
Live 2026: no graded calls yet
Forecast calm (≤ 5 mph) exploratory — found after looking; proving live
Outdoor or open-roof game with forecast wind ≤ 5 mph → QB passing yards over.
History (in-sample): Observed wind: over 56.5% in 2023–24 (n 216), 54.2% in 2025 (n 118).
Live 2026: no graded calls yet
Dome team away in bad weather exploratory — found after looking; proving live
Dome team (>50% of its home games under a dome or closed roof that season) playing away at an outdoor or open game with forecast wind ≥ 12 mph or temp ≤ 40°F → receptions under.
History (in-sample): 2023–24 57.0% (n 179); 2025 54.5% (n 77).
Live 2026: no graded calls yet
Myth-busted
- Myth-bustedFast young WR vs a corner beaten by speed, calm air → receiving yards over (2025: 45.6% on 114 bets)
- Myth-bustedFast young WR vs a corner beaten by speed, calm air → longest reception over (2025: 46.4% on 112 bets)
- Myth-bustedH1 with a sub-4.30 forty → receiving yards over (too few bets: 6 in 2025, 60 needed)
- Myth-bustedH1 with a sub-4.30 forty → longest reception over (too few bets: 6 in 2025, 60 needed)
- Myth-bustedAgile WR vs a corner beaten by agile receivers → receptions over (2025: 37.5% on 128 bets)
- Myth-bustedAgile WR vs a corner beaten by agile receivers → receiving yards over (2025: 39.6% on 106 bets)
- Myth-bustedBreakaway WR (top-speed proxy) → receiving yards over (2025: 47.2% on 335 bets)
- Myth-bustedBreakaway WR (top-speed proxy) → longest reception over (2025: 47.5% on 335 bets)
- Myth-bustedBreakaway WR vs a big-play defense → receiving yards over (2025: 47.0% on 83 bets)
- Myth-bustedBreakaway WR vs a big-play defense → longest reception over (2025: 47.0% on 83 bets)
- Myth-bustedYoung WR vs ageing corners → receiving yards over (2025: 50.6% on 79 bets)
- Myth-bustedHeight mismatch (4+ in) → receiving yards over (2025: 51.9% on 214 bets)
- Myth-bustedWeight mismatch (25+ lb) → receptions over (2025: 44.5% on 220 bets)
- Myth-bustedWR faster than his corners (0.08 s) → receiving yards over (2025: 41.3% on 235 bets)
- Myth-bustedRookie corner vs veteran WR → receiving yards over (2025: 41.3% on 92 bets)
- Myth-bustedFast WR vs slow safeties → receiving yards over (2025: 46.1% on 128 bets)
- Myth-bustedFast WR vs slow safeties → longest reception over (2025: 50.8% on 126 bets)
- Myth-bustedBig, fast RB vs light boxes → rushing yards over (2025: 46.2% on 169 bets)
- Myth-bustedFast TE vs slow linebackers → receiving yards over (too few bets: 21 in 2025, 60 needed)
- Myth-bustedBig TE vs short linebackers → receptions over (2025: 45.9% on 111 bets)
- Myth-bustedWeak-armed QB in wind → passing yards under (too few bets: 23 in 2025, 60 needed)
- Myth-bustedControl: strong-armed QB in wind → passing yards under (too few bets: 34 in 2025, 60 needed)
- Myth-busted3-cone: slowest third (WR > 7.01, TE > 7.12) · Rookie likely corner: no · Temperature (observed): 40°F or warmer → receptions under (2023–24: 63.1%; 2025: 51.7%)
- Myth-bustedAge: 25–28 · WR − corner height gap: smallest gap (WR ≤ 0.0) · Temperature (observed): 40°F or warmer → receiving yards under (2023–24: 62.1%; 2025: 51.7%)
- Myth-busted40-yard dash: fastest third (WR ≤ 4.42, TE ≤ 4.62) · Likely corners’ height: shortest third (WR ≤ 71.5) · Venue: home → receiving yards under (2023–24: 63.3%; 2025: 45.1%)
- Myth-bustedWeight: middle third (210–218) · Temperature (observed): 40°F or warmer → rushing yards under (2023–24: 61.3%; 2025: 47.7%)
- Myth-bustedWeight: middle third (210–218) · Team spread: underdog by 3+ → rushing yards under (2023–24: 62.9%; 2025: 49.1%)
- Myth-busted40-yard dash: fastest third (WR ≤ 4.42, TE ≤ 4.61) · Likely corners’ height: shortest third (WR ≤ 71.5) · Venue: home → receptions under (2023–24: 62.7%; 2025: 51.5%)
- Myth-bustedLinebackers’ forty: middle third (4.54–4.63) · Team spread: underdog by 3+ → rushing yards under (2023–24: 61.8%; 2025: 52.3%)
- Myth-busted20+ yd catch rate: lowest third (WR ≤ 14.4%, TE ≤ 9.5%) · Game total: 48 or higher → receptions under (2023–24: 61.9%; 2025: 47.4%)
All the traits at once vs the closing line
One regularised model per market, offered every pregame trait on this page and starting from the market’s own probability. Trained on 2023–24, scored once on 2025.The penalty kept no trait features in any market: adding traits made 2024 validation predictions worse in every market. The small log-loss gain comes from one constant “shade toward the under” term, and its rushing-yard bets won 52.8%, no better than betting every under.
| Market | Traits kept | Log-loss 2025: market / model | Brier 2025: market / model | Model bets 2025 |
|---|---|---|---|---|
| Receiving yards | 0 of 63 | 0.6928 / 0.6913 | 0.2499 / 0.2491 | none (never 3+ pts from the market) |
| Receptions | 0 of 63 | 0.6809 / 0.6786 | 0.2439 / 0.2428 | none (never 3+ pts from the market) |
| Rushing yards | 0 of 65 | 0.6926 / 0.6915 | 0.2496 / 0.2490 | 52.8% (861–771)blind 52.8% · ROI −1.2% |
| Passing yards | 0 of 30 | 0.6930 / 0.6930 | 0.2499 / 0.2499 | none (never 3+ pts from the market) |
What the data can and can’t tell us
- Combine coverage: WR 67% have a 40, 37% a 3-cone; TE 61% have a 40, 42% a 3-cone; RB 70% have a 40, 24% a 3-cone; QB 59% have a 40, 51% a 3-cone (players with lines, 2023–25). Many skip drills, and pro-day times are not in the data, so traits are missing, never guessed.
- Top speed: no legal free per-game source. The free nflverse NGS files have no max-speed column, and the NGS site’s terms forbid scraping. Proxies: 20+ yd catch rate, NGS separation and cushion.
- Who covers whom: unknown. The proxy corner actually played in 87.1% of games, was targeted at all in 84.2% (any corner who played: 86.7%), and was the most-targeted corner in 27.7% (random: 23.2%). PFR does not record which receiver a corner covered, so even this check is weak.
- Individual corners: the median corner has 12 matchups against fast receivers since 2022, and many of those pairings are guesses. Single-corner splits are mostly noise, which is why they are shrunk hard.
- Weather: 2023–25 tests use observed game weather (optimistic). Forecasts exist from 2026-10-07, so weather angles are capped at “promising”.
How this works
Lines are the consensus across major sportsbooks about an hour before kickoff. A bet wins when the stat lands on its side. ROI uses the consensus price, and break-even at −110 is 52.4%. Graduation, fixed in advance: promising needs a 2025 win rate above 52.4% on 60+ bets, with 2023–24 pointing the same way. An edge also needs a 2025 95% interval above 50% and a 2023–24 result beyond the permutation benchmark.
See also the weekly prop lines and the matchup-edges lab. Nothing here is betting advice. Live calls are recorded before kickoff and graded win or lose.
Sources: nflverse (combine, play-by-play, snap counts, NGS), Pro Football Reference coverage stats via nflverse, FTN Data via nflverse (box counts), The Odds API (closing prop lines)Updated 2026-10-07
Frequently asked questions
Do fast receivers beat their receiving-yards props?
Not in our test. Young WRs with a 4.43 or faster 40 facing corners who had been beaten by speed, in calm air, won 45.6% of overs in the 2025 holdout, below the 52.4% break-even.
Can I see a receiver’s in-game top speed?
Not legally for free. NFL Next Gen Stats publishes top speeds on its own site, but its terms forbid scraping or storing them, and the free nflverse NGS files have no max-speed column. We use a 20+ yard catch rate as the closest legal proxy.
Do we know which cornerback covers which receiver?
No. We use a "likely corner" proxy from snap counts and the receiver’s pregame role. The proxy corner was the defense’s most-targeted corner in only 27.7% of games, vs 23.2% for a random corner who played, so many "CB vs WR" pairs are guesses.
Why not just bet every under?
Overs hit 45–49% in most markets, so the under wins about 52–54% blindly, which is close to break-even after the vig. Several angles that "held up" in 2025 mostly rode that under tilt; the table shows whether each beat blind betting.
What is the NFL Prop Lab?
It is our public test bench for NFL player prop ideas. Each angle is written down before results are computed, tested on a holdout season, checked against shuffled random data, and then tracked live with calls that cannot be edited after kickoff.
Has Edgehalla found a proven NFL player prop edge?
No. Nothing has passed our full bar yet. A few angles are labelled promising or exploratory and are tracking live from Week 6 of the 2026 season, with no results yet.
Where do the prop lines in this research come from?
From The Odds API: a consensus of the major U.S. sportsbooks, using the median line and price, snapshotted 60 minutes before each kickoff for every game from 2023 through 2025.
Why do you test ideas on a holdout season?
Because the best results of any search are partly luck. Testing once on a season the search never saw shows how much of a result is real. When we did this, 19 of our top 20 angles got worse.