Week 6 QB Rankings – A Statistical Approach

These rankings are done using a statistical approach adapted from an article written by Danny Tuccitto at Footballguys.com.  As a brief side note, Footballguys.com has a yearly subscription that’s the same price as ESPN Insider and offers SO much more information.  I switched this year and it was 100% worth it.  Moving on..

What he did was take data from the past seasons and spent a ton of time looking at each QB’s score from week to week while analyzing the conditions at the time of each game to determine what information was important when evaluating how a QB would perform and what information was not.  His results revealed 3 factors worth considering.

1. Average going into the week – This seems pretty straightforward, but QBs who have a higher average score going into the week score higher, on average, than those with lower averages.  One thing that is helpful to getting better results is to choose a source you trust as reliable for projections.  If you look at how their pre-season projections for how many PPG each QB would score and count that as 1 game’s score, it balances the average.  For instance, if one site projectsed Matt Ryan to score 19 PPG, add his first 5 totals to 19 and divide by six.  This helps balance outliers and also becomes less of a factor as the season goes on, when we should start seeing a player’s numbers balance out.

2. Spread of the Game  – So many people try to factor how each team matches up against another when determining if a QB will put up big points.  Luckily, Vegas does this for us, and they’re pretty good at it.  On average, QBs score more points when favored to win and it increases by the amount favored but once the spread gets very high, the numbers tend to go back down a bit (likely explained by teams running the ball more when up big).

However, each QB responds a bit differently to the spread, which is where a lot of my work came in. Some QBs perform better when underdogs, likely because their team is a running team and only leans heavily on the QB when behind.  Some aren’t really affected by it at all, possibly because the coach sticks to the same gameplan regardless of if they are supposed to win or lose.  Most do follow the pattern I mentioned though.

3. Over Under – Similar to the spread, Vegas is good about predicting how high scoring a game will be.  Needless to say, the higher the over/under, the better QBs perform.  It works out to being around 1 fantasy point better for every 3 points of over under.  Again, this solves the problem of predicting how a QB will matchup against a defense.  Over/under may just project how high a game will score in general but when used with the Spread, it gives a picture of how Vegas expects the game to go, and based on research is actually a reliable way of projecting fantasy points for QBs.

What to expect

No system is perfect and it’s very easy to have some big outliers each week but according to Mr. Tuccitto, his system results in projections that are within 6 points of the actual score on average and has projected whether the QB will do better or worse than his current average around 70% of the time.  I’m doing a poor man’s version of this since I don’t have all the formulas which means I doubt I’ll be that successful but as a numbers guy, I prefer a statistical approach which uses known important variables to a gut feeling approach.

What I did

With this information in mind, I went back and looked at the spread, average at the time of the game, and over under for each QB since 2008.  This has given me a picture of how every QB performed based on the spread compared to how they should have been expected to produce.  In doing this, it does seem like the majority of QBs score higher when favored in the spread and in higher over/under games.  I have been working to determine how each individual QB responds to the spread based on the 5 years of data.

Flaws in the System

The first flaw is simply that going through this much data takes a TON of time.  Remember, I’m looking at each individual game for 30+ QBs over a 5 year period.  So far, I’ve gotten through about half of the QBs so the flaw is that QBs I have not analyzed yet (they are italicized so you know who they are when reading this) have projections generalized to overall trends, and do not have specific trends for how they personally have responded to the spread.  This will hopefully be complete by week 7 or 8.

Secondly, the data is not good at responding to changes in situations.  The best example of this is that this season’s average for Tom Brady is based on Brady not having Gronk.  This week, Gronk gets back and there’s not an easy way to adjust the projection for his return.  The one thing that does help is keeping Brady’s pre-season projections as part of his average since those will boost his numbers.  This requires some gut feeling approach by raising Brady up the rankings a bit despite his lower projection.  Also, the spread and O/U could help remedy this by increasing when a big player comes back (and decreasing when a player is injured).

Finally, rookies, young QBs, and promoted QBs don’t have much data to go on.  This means that someone like Geno Smith who has played 5 games will not have the same reliability as Drew Brees who has 85 games worth of data since 2008.  It’s especially challenging for a promoted QB who has never played an NFL game before.  In those instances it’s no more than a guess (although how many people are starting Mike Glennon).

How to Read the Rankings

The projected number is based on the math using the three variables I mentioned.  However, you’ll notice that the list is not just in order of projections.  In most cases, I’m trusting the numbers but sometimes if a QB has had a much higher or lower average than projected, I’ll adjust expecting a regression to the mean.   So, if I’m making a decision who to start, I would go by the ranks, not projections.

Bold is actual numbers

1. Peyton Manning vs. JAC – 31.9 points

2. Tony Romo vs. WAS – 24.0 points - 8 points

3. Aaron Rodgers @ BAL – 20.5 points 16 points

4. Drew Brees @ NE – 19.5 points

5. Philip Rivers vs. IND – 21.0 points

6. Andrew Luck @ SD – 19.9 points

7. Robert Griffin III @ DAL – 20.4 points – 12 points

8. Jay Cutler vs. NYG – 17.8 points 20 points

9. Matthew Stafford @ CLE – 19.1 points 23 points

10. Russell Wilson vs. TEN – 16.3 points

11. Tom Brady vs. NO – 14 points

12. Terrelle Pryor @ KC – 15.6 points 12 points

13. Cam Newton @ MIN – 14.8 points – 30 points

14. Alex Smith vs. OAK – 15.0 points – 7 points

15. Andy Dalton @ BUF – 15.1 points – 24 points

16. Geno Smith vs. PIT – 14.9 points – 4 points

17. Nick Foles @ TB – 14.2 points – 29 points

18. Sam Bradford @ HOU – 14.3 points – 16 points

19. Colin Kaepernick vs. ARI – 13.4 points

20. Eli Manning @ CHI – 13.8 points – 7 points

21. Joe Flacco vs. GB – 13.2 points – 19 points

22. Matt Schaub vs. STL – 13.0 points – 7 points

23. Matt Cassel vs. CAR – 14.3 points – 9 points

24. Ben Roethlisberger @ NYJ – 12.3 points – 15 points

25. Brandon Weedeon vs. MIA – 10.8 points – 16 points

26. Chad Henne @ DEN – 11.6 points

27. Ryan Fitzpatrick @ SEA – 7.6 points

28. Thaddeus Lewis vs. CIN – 9.4 points – 21 points

29. Mike Glennon vs. PHI – 4.8 points – 18 points

30. Carson Palmer @ SF – 2.6 points

2 thoughts on “Week 6 QB Rankings – A Statistical Approach

  1. Pingback: Week 7 QB Rankings | The Waiver Wire

  2. Pingback: Week 7 RB Rankings | The Waiver Wire

Comments are closed.