PROBLAY • RESULTS CENTER
2026 NFL Week 4 Results
The published Week 4 record for PROBLAY's stored Final Pregame-labeled Over/Under research, with the deduplicated view, the predicted-versus-observed difference, and the limitations that apply to the sample.
- Published
- Period covered
- 2026 NFL Week 4
This report uses frozen pre-kickoff data: the prices and model probabilities stored at the Final Pregame checkpoint before each game started. Nothing was added, removed, or re-priced after kickoff, and the report is not updated after publication.
37 wins, 6 losses, 86.0% observed rate
Population measured
Stored Final Pregame-Labeled Model Qualified Over/Under Research
Everything below measures exactly that population for the 2026 NFL regular season, Week 4: the propositions stored with a Final Pregame checkpoint label that met PROBLAY's model qualification rules for Over/Under player markets before kickoff.
Raw record
Raw record: 37 wins, 6 losses, 0 pushes
- Raw propositions
- 44
- Graded propositions
- 43 of 44 propositions graded
- Observed rate
- 86.0%
- Average model probability
- 83.8%
- Distinct players
- 38
- Games represented
- 13 of 16 games represented
- Voids / unresolved
- 1 void, 0 unresolved
Every stored proposition in the population is counted once, including several alternate lines for the same player and market. Voids are excluded from the win-loss record and from the observed rate, which is wins divided by wins plus losses (37 of 43). The average model probability is taken over the same 43 graded propositions.
Deduplicated record
Deduplicated record: 32 wins, 5 losses
- Player-market propositions
- 38
- Observed rate
- 86.5%
- Void
- 1
- Lowest-line sensitivity result
- 33 wins, 4 losses
One proposition per player, game, market and direction. The representative line is selected by implied probability closest to 50%, without using outcomes. The lowest-line sensitivity result beside it applies that same one-per-player-market rule with the lowest stored line selected instead, so the deduplicated record can be read against a different line-selection rule.
Predicted vs observed
Predicted versus observed: plus 2.2 percentage points
- Average model probability
- 83.8%
- Observed rate
- 86.0%
The observed rate came in 2.2 percentage points above the average model probability, both measured over the 43 graded propositions. That describes the recorded Week 4 sample only: it is not a correction factor, not a forecast, and not an expected future rate.
43 of the 44 qualified Week 4 O/U propositions came from alternate-line markets, so the raw propositions are not independent. This is a single recorded week and an early sample; it does not establish a durable edge or an expected future rate.
Methodology
What the population is
A proposition is in this report only if it was stored with a Final Pregame checkpoint label and met PROBLAY's model qualification rules for an Over/Under player market before kickoff. The stored line, direction and model probability are the ones captured at that checkpoint.
How wins and losses are counted
Each proposition is resolved against the stored line and direction using the final game result. Week 4 produced 0 pushes and 0 unresolved propositions. One proposition is void because no official box-score row exists for that player in that game, so 43 of the 44 raw propositions are graded. The observed rate and the average model probability are both measured over those 43 graded propositions.
Why a deduplicated view is shown
Alternate-line markets mean one player and market can appear several times in the raw population. The deduplicated view keeps one proposition per player, game, market and direction, selecting the representative line by implied probability closest to 50% and never by outcome.
How weeks relate to each other
Each week is reported on its own. PROBLAY does not create a cumulative performance claim by simply adding the headline records from separate weekly reports. Alternate-line propositions within each week are correlated rather than independent, and a valid season-to-date evaluation requires a predefined population, unit of analysis, deduplication methodology and treatment of any dependence across weeks.
Limitations
- 43 of the 44 qualified Week 4 O/U propositions came from alternate-line markets, so the raw propositions are not independent.
- One proposition came from a standard market that FanDuel offered on both sides, and it is priced no-vig. The remaining propositions come from alternate-line ladders, which the book offers Over-only by construction (an opposing side at the same threshold was never offered), so they are priced raw implied.
- One proposition is void because no official box-score row exists for that player in that game. Voids are excluded from the win-loss record.
- A line carries the Final Pregame label only when its price was new or changed at that checkpoint. Lines offered unchanged at Final Pregame keep an earlier label and are not in this population.
- This is a recorded sample, not the complete Final Pregame market.
- 13 of 16 games are represented, but not evenly, so game coverage should not be read as uniform.
- Anytime Touchdown research is evaluated separately and must not be blended into this record.
- This is a single recorded week and an early sample; it does not establish a durable edge or an expected future rate.
About this report
← Back to the Results CenterResearch and analysis only. PROBLAY is not a sportsbook and does not accept wagers.