A couple of weeks back we took a look at predicting Peyton Manning’s 2015 season using SAP Predictive Analytics 2.0. It wasn’t a good time for Peyton Manning. He had just gotten pulled out of the worst statistical game of his career with a QB passer rating of 0 (which we accurately predicted would be his worst game as a professional). Additionally, it looked like he may have played his last game as he suffered a foot injury that at the time looked to sideline him for the rest of the season, if not career.
But rumors of his demise have been a bit exaggerated as he is now set to start for the Denver Broncos in Super Bowl 50. Peyton Manning may very well go down as the greatest regular season QB in history. I stress ‘REGULAR SEASON’. But Playoffs! You kidding me? Playoffs??!
He has played in 26 postseason games and up until this upcoming Sunday, his record is an average 13-13. Compare that with Tom Brady who has 4 rings and a record of 22-9 in the playoffs.
Is it fair to compare him to Tom Brady? Some talking heads say that Tom Brady is a product of a better system that produces championships. This may be true.
My goal here is to create a model that will tell us how much of Manning’s postseason record can be predicted by his own stats versus other influences. Manning is only on the field with the offense, so is it fair to cast all the blame on his postseason record on his play? Let’s not forget that the matchup this upcoming Sunday in the Super Bowl will feature the #1 Defense in terms of Yards Allowed (Denver Broncos) vs. the #1 Offense in terms of Points Scored per Game (Carolina Panthers). There are many who have already said that if Denver is to win the game, it will be primarily due to how their defense can stop Cam Newton and the Carolina Panthers offense.
So, as mentioned above, we have 26 playoff games that Peyton Manning has participated in during his career. His first playoff game was in 1999 against the Tennessee Titans (L) and his last playoff game was against Tom Brady and the New England Patriots (W) just two weeks ago.
His QBR (Quarterback rating) has fluctuated quite a bit during that time as can be seen in the chart below.
Not to get too much into the weeds, but most NFL experts will say that any QBR above a 75 is acceptable and anything over 100 is excellent. He’s had a few games in the 30’s (very bad) and 8 games under 75 (not that much better). He’s also had 5 games over 100 (excellent!!!). In all 3 of his previous Super Bowl appearances he never had a score lower than 73, the lowest coming against the Seattle Seahawks two years ago, which was a blowout loss 43-8.
Our predictive model for Peyton’s postseason will consist of the following typical NFL QB stats: Completion, Attempts, Yards, Touchdowns, Interceptions, and most importantly (Wins/Losses). We will build a logistic regression model to determine a Yes/No outcome based on his throwing stats. We will then use the model to see how close those numbers actually predict an outcome of a win (1) vs. a loss (0).
It is important to note that SAP Predictive Analytics does not have a Logistic Regression model out of the box but I did find a custom R component that I imported into the dataset from the SCN courtesy of Andreas Forster.
Once the model was imported, I connected it to Peyton Manning’s playoff dataset.
Once the model is connected the next step is to identify the predictor variables (Completions, TD’s, INT’, etc..) and the classifier or response variable (Score = Wins or Losses).
The threshold to determine the line between a 0 and 1 was set to 0.5. I know this is not an aggressive model, but the guy is almost 40 years old, so I’m trying to give him all the help he can get.
After running the model we can now visualize our actual score against our predicted score of Wins and Losses.
Based purely on Manning’s stats for his postseason, the model predicted that he would have one extra loss and one fewer win than he currently has (12-14 instead of 13-13). The model seems to think that his stats are pretty consistent with the record that he currently has (+/- 1 game).
What does this all mean? Manning is his own worst enemy on the field. While there may be a few games that the model predicted he should win, but ultimately lost and vice versa, for most of his career Peyton Manning has been in control of his own record based on his performance. You could even make the case that he has overachieved by 1 game.
There is a lot riding on this game for Peyton Manning beyond the Super Bowl trophy. If he wins he will have as many Super Bowl rings as his younger brother, Eli Manning, has with the NY Giants. It’s tough to be called one of the greatest NFL quarterbacks of all time, if you don’t even has as many championships as your younger brother who many believe is an inferior QB.
Most importantly, Manning would in all likelihood retire with a winning playoff record (14-13) and prove his skeptics wrong about not being able to win the big game, especially late in his career.