P2PY Game Predictions For Week 3

Florida State Schedule Breakdown 2017
ORLANDO, FL - SEPTEMBER 05: Deondre Francois #12 of the Florida State Seminoles looks to pass the ball in the second half against the Mississippi Rebels during the Camping World Kickoff at Camping World Stadium on September 5, 2016 in Orlando, Florida. (Photo by Streeter Lecka/Getty Images)

Let’s make some picks! But let’s make some smart, numbers-based picks…


Previously, I introduced P2PY methodology to correlate game outcomes (HERE). I went through the 2015 bowl season and using year-long P2PY retroactively picked the winner in 28 out of 40 games correctly. This work is currently unpublished but may be subject to a future post. Going forward every week in 2016, I will pick 10 games with a focus on the bigger games for any given week, i.e. those involving ranked teams. For Week 3, I’ve thrown out games against FCS teams (NDSU versus Iowa) or those where a team has only played one game so far (Stanford, Michigan State).


As described earlier, P2PY is meant to be a simple method to estimate a team’s productivity and efficiency. Now, we’re taking it forward to project a team’s score given their opponents defensive performance. The calculation for that is:


The P2PY is adjusted for schedule. To do this, I used the conference comparisons from colleyrankings.com (HERE) that provides each conference with a rating. For example, this week, the Big Ten has the highest rating (0.670349) and the MAC is 0.383017. The data is normalized so that the Big Ten was 1 and all others were a percentage of that. A team’s actual P2PY for a week was then multiplied by their opponent’s conference rating to get the adjusted P2PY.

One note, FCS had a rating of 0.130749 which leads to a huge penalty for teams that played FCS opponents. To lessen this penalty, I adjusted this to 0.25, normalized to 0.373. This means that if an offense had an actual P2PY of 3.42 (42 points on 511 yards) but played an FCS team, this gets adjusted to a P2PY of 1.29. This example is for Houston playing Lamar in week 2.

In future weeks, I will likely adjust the defensive yards against in a similar fashion but for week 3, yards against are simply the average of a team’s previous games.

For all calculations, Weeks 1 and 2 are receiving equal weight.

P2PY Game Predictions — Week 3

A summary of all of the picks – straight up and against the spread are below.

Houston @ Cincinnati (+7)
Prediction: Houston 26 to Cincinnati 20
ATS Winner: Cincinnati

Florida St. @ Louisville (+2.5)
Prediction: Florida St. 30 to Louisville 38
ATS Winner: Louisville

Miami (FL) @ Appalachian St. (+3)
Prediction: Miami (FL) 29 to App. St. 14
ATS Winner: Miami (FL)

Alabama @ Mississippi (+9.5)
Prediction: Alabama 39 to Ole Miss 21
ATS Winner: Alabama

Oregon @ Nebraska (-3)
Prediction: Oregon 27 to Nebraska 37
ATS Winner: Nebraska

Texas A&M @ Auburn (-4)
Prediction: Texas A&M 28 to Auburn 22
ATS Winner: Texas A&M

Georgia @ Missouri (+6.5)
Prediction: Georgia 26 to Missouri 22
ATS Winner: Missouri

Ohio St. @ Oklahoma (+2.5) 42 22 Ohio St. Ohio St.
Prediction: Ohio St. 42 to Oklahoma 22
ATS Winner: Ohio State

Temple @ Penn State (-9) 20 25 Penn State Temple
Prediction: Temple 20 to Penn State 25
ATS Winner: Temple

Baylor @ Rice (+30) 36 11 Baylor Rice
Prediction: Baylor 36 to Rice 11
ATS Winner: Rice

Prediction Notes Including NGFT* Analysis

Houston and Cincinnati seems too close for me. I actually like the Cougars a lot more in this game than the prediction would indicate.

Louisville is as attractive to the eye as it is to the calculator. They’ve been absolutely unstoppable so far this year. Whether or not they can maintain this pace (the probably can’t) will be exciting to watch.

Alabama is going to crush Ole Miss.

Obviously, the predicted 20 point prediction for Ohio State over Oklahoma jumps out as improbable. But again, the Buckeyes have been on a tear to open the season and J.T. Barrett is pretty experienced.

Penn State will win by more than 5 points on Saturday.

* – NGF is Nate’s Gut Feel Theory

Main Photo:



Please enter your comment!
Please enter your name here

This site uses Akismet to reduce spam. Learn how your comment data is processed.