NFL Week 15 Picks

We selected 12/16 match-ups last week, and the model correlated at 60.3% to outcomes (NOT point total/ATS, that hit 37%).

There are still a few key injury adjustments to be made to this week’s final selections (DET and NYG to name a few), but about 90% of the alterations have been accounted for/adjusted in the model. This week we included a few more of the variables we consider subjectively, yet are not accounted for objectively in the model:

Key (Included in Model):

-DVOA – Var: Overall team performance based on Football Outsiders DVOA, relative to their consistency (Variance) throughout the year
-Coach: Coach Adjustment based on Edj Rankings (based on net game winning percentage points coach decisions have led to)
-Opp TO Li: Opponent’s Turnover Likelihood, see Week 15 Turnover Likelihood
-Inj Adj: Injury Adjustment, based on the SIS On/Off Report’s net EPA when an injured player is off the field relative to his sub’s net EPA when on the field

Key (NOT Included in Model):

Any and all of these items are variables that season long metrics will not capture relative to the match-up. For example, Team A’s Defense may be better than Team B’s Offense, yet Team A’s defense performs far better vs 11 personnel, and Team B’s Offense runs a lot of 11 (relative to other sets)…this is something we want to know. These types of “before the fact ANSWERS” are to the “after the face QUESTIONS” of how could this team have won, when all the season-long data pointed otherwise.

-O Personnel: What personnel does this team’s Offense tend to use more than any other (only teams included with relatively higher percentages of abc set vs. other sets)
-D Good At: What offensive personnel does this team’s defense out-perform their season-long average performance metric
-D Bad At: Same as above, but opposite
-QB w/Blitz Adj: Takes QB vs Cov (how well a QB performs by EPA% vs a certain coverage, compared to how often this week’s opponent deploys said coverage, see more at Week 15 QB vs. Coverage) and adjusts for the “Goff/Mayfield check”. That is, how much does pressure rate (here defined as defensive blitz rate, NOT pressure rate on purpose) impact a QB’s performance (here as net Positive EPA% rate)

These are net values between one team’s offense and their opponent’s defense.
-nOffAly: Based on FO’s ALY, which measures how well an OL creates yards for its RBs vs DL’s stop RBs (many times this is written as “netALI” because either I am an idiot, and or I tend to Freudian slip that term with my 5th grade girlfriend I have still not let go of…likley a little of both)
-nOff Power: Based on FO, measures efficiency in short yardage situations
-nOFFpsrh: Based on PFF’s grade for pass pro, vs opponent’s defense pass rush
-nWRvCOV: Based on PFF’s grade for WR’s in pass, vs opponent’s coverage grade
-Run Net: Aggregate of all run nets
-Pass Net: Same as above with pass
-NON QB Match-up Net: aggregate all (we like using this for under/overs but have NOT tested it for that objectively yet, stand by)

As a general note, although some variables are not included in our empirical model, they still help make our decisions by either giving us pause, or confirming choices. They help us “zoom-in” from overall numbers and identify glaring match-up issues that may determine outcomes of the games.

Based on how well the model has been performing we recommend the following exposure PER POSITION, per straight-up bet. Mind you, if you mix in with parlays the numbers change:

ScoreConfidenceMinimum payout to justify play
Use the score given from the model to determine first column, only play if right-most column is achieved, and use confidence rates in middle column to determine proper exposure (ideally based on your purse relative to Kelly Criterion…if you don’t know what that is, you should not be betting)

If we help make you money this weekend, please subscribe (“Follow” button on the top right). If NOT, please (as you should with any other “prognosticator”) ignore us going forward…we know our ONLY value is in being right, not fancy, not aesthetically pleasing, just right. With that being said, we note the following games, that the model picked, as either “go against” or “no play”:

-We are going with ATL (model says TB), NO PLAY (there are too many match-up advantages on the ATL side to go with TB)
-MIA/NE: NO PLAY, there are far too many STRONG corollaries on BOTH sides, this one is too hot to risk investment
-SEA. We like WAS ATS, but are not willing to back WAS SU.

Leave a Reply