Discussing the NFL

Football writing is at a crossroads.  We have gotten so used to stories written by crotchety old writers who kept getting cigar ash in their keyboards. Joe Montana is the greatest leader ever because he looked up in the stands and said "Is that John Candy?" Their victory had less to do with all the talent own the field and the play calling and was the result of Montana being the quarterback and the leader and calming his team down so they could execute and wouldn't buckle under the pressure.  These stories are still interesting and fun to read, but we have a better understanding that they don't quite explain wins and losses as well as certain statistics, but statistics are very boring.  I've even noticed this as a writer.  Its a lot more fun to make things up and dive into the complexities of line play, than to read to you the results of my statistical analysis.  Any article trying to used statistics has to use half the words recapping Statistics 101 and how the writer is using them.  I agree with you, it's boring and it sucks.  But to have a better understanding of what leads to wins and losses, you need suck it up and take your medicine.  Trust me it gets better...towards the end, we have to get through our statistics lecture first.

When Bill Barnwell started his article "Fantasy football is an unfair game ruled by randomness and luck." I was worried the professional writer was going to give out all my ideas before I could find the spare time to embark on my hobby. Mostly because he had done exactly that four days earlier.  But I got a little lucky.  They actually turn out to be a better, more in-depth discussion of a few topics I want to touch on, but didn't want to spend as much time on going into as much detail. Pythagorean winning percentage, Plexiglass principle, turnovers, randomness; thanks for saving me a lot of time Bill.

Now for the statistics lecture.  What are the odds of flipping heads on a fair coin twice in a row? Correct 25% or 1/4, however you want to express it.  50% times 50% gives us 25%. Now, having already flipped heads once, what are the odds of flipping heads again? Correct nerdy kid in the corner who has ruined my punch line, 50%. Flipping coins are independent events and the result of the first flip has no effect on the result of the second flip.  The odds of getting heads on the second flip are the same as getting heads on a single flip.  Now, let’s say there are 2 blue balls and 2 yellow balls in a sack.  The odds of pulling one yellow ball are 50%.  What are the odds of pulling two yellow balls in a row (without replacing)? 1/6 or 16.67%.  And having already pulled one yellow ball, what are the odds of pulling a second yellow ball?  1 in 3.  This is an example of a dependent event.  The odds of the second pull depend on the first event.  If you pull a blue ball on the first pull, the odds of pulling a yellow ball increase to 2/3.  Remember this distinction between independent and dependent events.

This distinction applies to all types of predictions, and for illustrative purposes we’re going to look at Rob Gronkowski’s 17 receiving touchdowns. (He also caught one backward pass that was technically a rushing touchdown.)  There have been 50 individual seasons where a player scored at least 14 touchdowns.  Of those, only 6 guys have had multiple 14 TD seasons.  Jerry Rice, Randy Moss, Mark Clayton, Art Powell, Terrell Owens, and Marvin Harrison.  And Mark Clayton and Art Powell only did it twice.  We’re going to look at the four guys who have done it three times or more.  Rice did it 6 times in 21 seasons,  Moss did it 4 times in 15 seasons, Owens did it 4 times in 15 seasons, and Harrison did it 3 times in 12 seasons.  So combined, these guys have scored at least 14 TDs in 27% of their seasons.  Back to Gronkowski…the odds of him scoring 14 or more TDs are dependent on his ability and the offense he is in, but independent of how many touchdowns he scored last year.  This is pretty much the case with every football (or any) prediction.  Past performance does not guarantee future success.  But ability, talent, scheme, etc. can be predictors of future success.  So if you think Rob Gronkowski is in the same class as Jerry Rice, Randy Moss, Terrell Owens, or Marvin Harrison (which has really, really low odds), then there is about a 25% chance of him catching more than 14 touchdowns this year.  Now, because he is a very talented individual with a Hall of Fame quarterback throwing to him, I think he has the ability to score 8-12 touchdowns, but it’s easy to predict that there will be a regression from his 17 touchdown performance last year.

This is the single biggest thing to remember in any type of prediction.  We use past performance to gauge ability and talent, but we have to remember that future success, while dependent on ability and talent, is independent of past performance.  LeSean McCoy will score fewer than 17 touchdowns.  Ray Rice will have over 1500 total yards barring injury.  Cam Newton will have fewer than 14 rushing touchdowns.  Calvin Johnson follows the exact same analysis as Rob Gronkowski above.  If you feel that he is in the same class as the Hall of Fame receivers mentioned above, which is more likely than Rob Gronkowski being in the same class, then he has about a 25% chance of scoring 14 touchdowns.

For football specific predictions, I’m going to provide a meta-analysis approach.  Well, not a true meta-analysis, because I’m not going to include all analyses.  I am going to do a lot of research and rely even more on other’s research and relay what I have been convinced of and what I have determined for myself.

The few stats and forms of analysis that I have aggregated from a variety of sources, such as Bill Barnwell at Grantland, KC Joyner at ESPN.com, Football Outsiders, Cold, Hard Football Facts, Advanced NFL Stats, and others that I can’t think of right now.  I look at Pythagorean winning percentage, but record in close games is already reflected in the difference between Pythagorean wins and actual wins.  Turnovers play a disproportionate role in the outcome of any individual game.  If you knew which team would have more turnovers in a game, you would be very successful in Vegas.  Avoiding interceptions is a quarterback skill, with some variation, but getting the defense to drop interceptions is random.  Defensive ability and scheme can also play a big difference in interceptions.  Causing and avoiding giving up fumbles can be a skill, but recovering fumbles is random.  Because it is hard to predict turnovers based on fumbles, and I’m lazier than the guys at Football Outsiders who track fumbles, but not recoveries, I focus on interceptions.  This is also reflected in passer rating differential.  In the past I have used throws per interception resulting in numbers that range from 30 to 90, with the higher number the better.  Part of Alex Smith's success last year was that he threw an interception every 90 throws, while Eli Manning threw a pick every 36 throws.  This is identical, but the inverse of an interception rate, which can range from 1% to 5% of throws, with the lower the better (for the quarterback).  Which format is easier for you to understand?