clock menu more-arrow no yes mobile

Filed under:

FTRS Reader MillsGT49 goes number crunching! Expected Win Probability

FTRS reader MillsGT49 broke down some GT football stats that get pretty interesting. I can't FP it and make it the most recent article, so here's a straight up copy and paste from a loyal reader.

           I recently read the article from the Football Sabermetrics Fanshot about Expected Win Probability (EWP) developed for baseball being applied to football. I personally love in depth statistical analysis and looking back at the past using some new methods. For those of you who did not read the article, which I recommend you do, EWP is the square of the number of runs (points) scored divided by the sum of the square of runs scored and the square of runs allowed. 

Image and video hosting by TinyPic

      While I love stats I do not pretend to know the ins and outs of a stat but a thorough explanation is given HERE. Anyway this stat gives a baseline for how many wins a team is expected to earn during the season (EWP*# of games=Expected # of wins, for last season our EW was .65*14 = 9.1) any variance is attributed to "luck" or some skill that helps win close games (great kicker, clutch QB, CPJ winning game on 4th down, etc…). What I wanted to do with this stat was try and apply it to individual players. I wanted to determine exactly how many wins Nesbitt and Dwyer contributed this season. This is an impossible question to answer. I will explain why pretty much all of my assumptions are wrong as I work through the post but the number one problem with individual player football stats is that it is near impossible to completely separate one player’s contributions from the 10 other players on the field. BUT for this post lets assume we can.  The first question I had to answer is how do I separate the points that Nesbitt contributed from the rest of the team. What I did is calculate our EWP with Nesbitt’s points removed (from both the numerator and denominator) and then subtract that EWP without Nesbitt (.525) from our original EWP (.65) to get an EWP+(.125) and multiplied that number by 12(A historical comparison which I do later requires me to consider a 12 game season). That gives us Nesbitt’s Expected Wins Added (1.5). According to EWP Josh Nesbitt’s point contributions last season gave us 1.5 extra wins. I also calculated all of this data for each of Coach Paul Johnson’s QBs since 2004.


Image and video hosting by TinyPic

 Of course this is not very fair. Without Nesbitt another QB would have stepped up and scored SOME points, and maybe all of Nesbitt’s goal line dives would go to Dwyer instead, who knows. What this does do is show how important Nesbitt was to our team this year compared to other years as he had the second highest EWA of any CPJ QB since 2004. But this still did not quite do it for me in regards to the previous statement. How much better did Nesbitt perform than another CPJ QB? What would have happened if another QB stepped in. I wanted to compare Nesbitt to a "base" level of CPJ QB performance (This is done a lot in baseball). I took the average points scored from each CPJ QB since 2004, all the way back to Navy, and calculated the average. I then subtracted each QBs points from that years total and replaced it with this average scoring contribution. This gave me an EWP from a "base" level quarterback, which I call EWPavg. This is the expected win percentage of each team with the average performance from a QB. This will change because each team’s EWP is still very dependent on the amount of points given up and points scored by other players. EWP is a situational stat, EWP is less influenced by points scored when the point differential is already fairly large than when points allowed and points scored are equal. So a big QB contribution will matter less when the defense was already great, or the B-back is a monster, etc… Sorry, random tangent, anyway once I had this EWPavg (for 2009 it was .621) I subtracted the EWP with Nesbitt and multiplied this difference by 12 games to get a value for Wins Above Replacement.


Image and video hosting by TinyPic

       Josh Nesbitt contributed .35 extra wins than the average CPJ QB would have. This was the largest WAR value of any CPJ QB for the data I have (2004-2009). This was a HUGE improvement over last year when he "cost" us .59 wins compared to an average QB. This can be attributed a little to the fact that he missed some games in 2008, but he should naturally struggle compared to other CPJ QBs considering it was his first year in the offense, not just first year starting.

        Of course this needs to be taken with a grain of salt. I am not nor pretend to be a master of statistics and am sure I violated some rules in my assumptions. And anyone who saw the Jackets play this season knows Nesbitt contributed more than just .352 wins above an average QB. His EWA also seems low, I know he won us more games than that. But of course this analysis is not perfect and wasn't meant to be. This was more a look at how Nesbitt has ranked in terms of QBs Coach Paul Johnson has had. I won’t do much concluding, I will let you guys do that. What do you think of the post? What do you take from the data and if you have any comments or criticisms or ways to improve please post to let me know.