WEEK 2 · Sun, Sep 20, 12:00 PM CDT
Cincinnati Bengals @ Houston Texans
2 qualified model plays · 2 modeled lines did not qualify · 1-1-0
MODEL OVERVIEW
What the board is seeing
Projection gap is the raw difference between PDR's projection and the sportsbook line. No-vig edge is the probability advantage after removing sportsbook hold. They answer different questions.
How the model works →QUALIFIED PLAYS
Game prop board
Ranked by the frozen published model board.
WIN
OVER 1.5
fanduel -118
Projection
2.6
Proj. gap
+1.1
Model win
78.1%
No-vig market
50.6%
No-vig edge
+27.5%
Model-implied EV
+44.3%
WHY THIS PLAY
Projection 2.6 is 1.1 above the 1.5 line.
Model win probability 78.1% vs 50.6% no-vig market probability creates a +27.5% edge.
Qualified with 516 residual samples, 78.1% model win, +27.5% no-vig edge, and +44.3% model-implied EV; publication floors are 500 samples, 57.5% win, +5.0% edge, and +5.0% EV.
Historical residual sample: 516
Final: 2 passing TDs
LOSS
UNDER 240.5
fanduel -114
Projection
205.5
Proj. gap
-35.0
Model win
67.6%
No-vig market
50.0%
No-vig edge
+17.6%
Model-implied EV
+27.0%
WHY THIS PLAY
Projection 205.5 is 35.0 below the 240.5 line.
Model win probability 67.6% vs 50.0% no-vig market probability creates a +17.6% edge.
Qualified with 516 residual samples, 67.6% model win, +17.6% no-vig edge, and +27.0% model-implied EV; publication floors are 500 samples, 57.5% win, +5.0% edge, and +5.0% EV.
Historical residual sample: 516
Final: 353 passing yards
MODELED - DID NOT QUALIFY
Other lines PDR reviewed
Context only. These are not published recommendations and are not graded.
Did not qualify
Model side: UNDER 1.5
fanduel -122
Projection1.5
Proj. gap-0.0
Model win54.8%
No-vig market51.4%
No-vig edge+3.4%
Model-implied EV-0.2%
Why it did not qualify:
Model win probability did not clear the publication floor.
Historical residual sample: 516
Did not qualify
Model side: OVER 247.5
fanduel -114
Projection256.5
Proj. gap+9.0
Model win52.9%
No-vig market50.0%
No-vig edge+2.9%
Model-implied EV-0.7%
Why it did not qualify:
Model win probability did not clear the publication floor.
Historical residual sample: 516