Predicting Peyton Manning’s 2015 season with SAP Predictive Analytics 2.0


Peyton Manning is an amazing commercial actor, but did you know that he also moonlights as quarterback on the weekends? All kidding aside, Peyton Manning will go down as a Top 5 quarterback in NFL history. In November 2015 he broke the all-time passing yards record held by Brett Favre. It was also during that game that he had his single worst outing as a professional quarterback, scoring a 0 QB rating. I repeat a 0 QB rating.He was later pulled out of the game by the head coach and has not played a down since then after it appears he suffered a partial tear of the plantar fascia. I don’t know what a plantar fascia is but if it makes you throw a 0 QB, I can only hope it is contagious for a certain New England quarterback. Just kidding. Sort of.

A passer rating or quarterback rating (QBR) is a calculation used by the NFL to measure the efficiency and effectiveness of a quarterback. It factors in completions, yards, interceptions, and touchdowns. The higher the score the better the results. The baseline between a good and bad score depends on who you ask and the era that the quarterback played in. However, most current analysts today would say anything above a 75 is acceptable and anything above 100 is excellent.

Ironically, the same year he broke Favre’s record has also been his most disappointing statistically. Before this year Peyton Manning had been averaging a QBR (Quarterback Rating) of 97.6 for his career. That includes his rookie season where he threw 28 interceptions and had a QBR of 71.2. While the Broncos are still having a great season, Peyton Manning was averaging a QBR of 67.6 for the 9 games he played this season. This is lower than his rookie season. Ever since his potential career-ending neck surgery in 2011, many talking heads on the four-letter network have been predicting his demise. They were wrong in 2012, 2013, and 2014.

Rather than going based on the ramblings of those who have too much free time, I thought it would be fun to see if we could forecast Manning’s 2015 season using SAP Predictive Analytics 2.0 time series analysis models. Could we predict his current season knowing how his previous 16 seasons panned out? Let us find out.

Below is a time series plot of the QBR by week and year for Peyton Manning’s entire career (including the 9 games this season).


image 1








The data for the QBR will be broken up into two datasets:

1998-2014: Training Dataset

2015: Testing Dataset

The training dataset will be used to formulate a triple exponential smoothing time series forecast for the 9 weeks (periods) in 2015. Those 9 periods will equate to the 9 weeks of data we have for 2015. We can then compare our new forecasted 2015 data with the 2015 actual test data and see how close we came. The data for Manning’s career was pulled from the

Step 1

Pull Data into Predictive Analytics 2.0 – Expert Analytics

Step 2

Connect Data into Triple Exponential Smoothing Time Series Algorithm


image 2









Step 3

Configure Time Series Forecast

Since a Time Series is just a series of points over a sequential period of time, we can easily forecast 9 periods based on the previous 200+ periods available before 2015.


image 3








Our target variable is the QBR. We have a custom time period of 16 weeks of football per year. We are forecasting for 9 period and the starting year is 1998, Manning’s rookie season.

Additionally, we will not change the advance properties for the triple smoothing model and keep the Alpha, Beta, and Gamma values at 0.3, 0.1, and 0.1


image 4








Not to get too much into the weeds, but exponential smoothing is used to smooth out time series data to get rid of some seasonality that may exist as noise within the data. Since games are played at approximately the same time each year, there may exist seasonality in the data that can affect the outcome. So, these factors (alpha, beta, and gamma) can assist in smoothing out the seasonal and trending affects of the data.

Once that is complete we can go ahead and run our model.

The last 9 rows in the model show us our new predicted values for 2015.


image 5











We can now go ahead and visualize our actual 2015 QBR values with the predicted 2015 values and see how close we are by week.


image 6














While on a week-to-week basis there is some fluctuation between the actual and predicted score for QBR, overall the trend is consistent with what actually happened. Manning’s 2015 QBR average was 67.61 and the predicted score average was 53.5. Neither score is great and is predicting a sharp decline from his 2013 and 2014 QBR, which were 115 and 101 respectively.

Additionally, and most ironically the lowest score predicted came on the same day that Peyton Manning had the worst game of his entire career. While a 35.98 is better than a 0, you can probably put me in the game and I could finagle my way to something slightly better than a 0 score. Maybe.

So in conclusion, based purely on the data and nothing else, we could have predicted using SAP Predictive Analytics 2.0, this would be Peyton Manning’s worst season statistically as well as predicting Week 9 would be his worst statistical performance.



Passer Rating Definition from Wikipedia

Quarterback Passer Rating Calculator

Peyton Manning Statistics by Year

Learn more about SAP Predictive Analytics 2.0

One thought on “Predicting Peyton Manning’s 2015 season with SAP Predictive Analytics 2.0

Leave a Reply