Well Played, Alex Smith

Categories: NFL, Statistics
Comments: No Comments
Published on: October 30, 2012

(*Quick note: These numbers are likely to change as more data is added throughout the season and I find a way to deal with running backs as receivers, which create a lot of noise in the model)

Alex Smith is forcing me to work a lot harder than I thought I needed to. You see, I’ve known what my first NFL post would be for some time. It was going to be about using Alex Smith as an example of regression to the mean, the tendency for extreme performances to be followed by closer to average performances.

Since he entered the league, Alex Smith’s performance has ranked in the bottom third among NFL quarterbacks. Up until 2011, Alex Smith had not had a spectacular career as a quarterback. Then comes the 2011 season. Suddenly, Alex Smith shoots up to 9th best among NFL quarterbacks.

The story basically writes itself. We know that extreme scores tend to be followed to a movement back toward the average. Alex Smith’s average ranking is somewhere in the neighborhood of 24th from the time he entered the league through 2010. So, Alex Smith should see his rankings move back toward the bottom third of the league during the 2012 season.

Now that we are halfway through the season, we finally have enough data to begin checking these predictions. A few quick calculations tell a very interesting story. The problem for me is that the story is different than I expected. Today I learned that the data don’t care about my well-crafted narratives. Alex Smith’s performance is not moving back toward the mean. Instead, he is currently ranked 4th among NFL quarterbacks. He is improving in ways we wouldn’t expect if his 2011 performance was based on random chance.

This means that I have to start checking a number of different, alternative scenarios. I’ll be looking at head coach effects, development effects, age effects, and a large number of other possibilities. The problem is that these tests take time and time is something I have in relatively short supply these days. I’m still busy collecting all the NCAA data I can is preparation for draft season.

So…well played, Mr. Smith. Your sustained ability to play the quarterback position has resulted in a lot more work for me.

The Sad Case of Kellen Moore, Part 2

Categories: NCAA FBS, Statistics
Comments: No Comments
Published on: October 27, 2012

Today we continue the discussion of Kellen Moore.  I’ve already mentioned that the physical “issues” scouts made about him are likely spurious arguments.  I’ve also waxed philosophic about sand.  And while that was fun and everything, it’s time to get to something both useful and novel.

Kellen Moore won a lot of games (50) at Boise State.  He also threw for a ton of yards (14,667 career) and was remarkably accurate (69.78% career, 74.3% senior season).  What negative could one possibly find in these outlandish numbers?

The argument seems to be that the numbers simply cannot be believed.  Yes, Kellen Moore won a large number of games, but who did he win them against?  For more than a decade, Boise State has been head and shoulders above their conference competition.  There have only been three seasons since 1999 in which they didn’t win their conference championship.

So we construct endless scenarios in our brains.  We quickly imagine Boise State being a member of the PAC-12 or Big 12.  We picture Kellen Moore lining up against the likes of Oregon or Texas.  Our brains create these pictures easily because our brains are designed for this type of creative thinking.  We are good at imagining the what-ifs.  The thing is, our imaginings are not accurate, especially when they involve statistics.  Our brains are great at thinking about Boise State being in a different conference, but terrible at being correct about what would actually happen.

To figure out what would happen, we need data.  We’ll look at completion percentage for two reasons:  1) Completion percentage is one of the few statistics that has even had the suggestion or hint that college performance predicts pro performance, and 2) It is the easiest data to restructure from publically available data into something we can use.

Kellen Moore’s Accuracy

You might believe that Kellen Moore’s stats should not be believed because he does not play against “elite” competition.  However, if you want to use that argument you have to answer a secondary question.  How much does it matter?  Any Intro Psych textbook will tell you that we can’t answer this question just by thinking about it.  We need math to answer it.

During Kellen Moore’s senior season, he completed 326 of 439 passes for a completion percentage of 74.3%, the best in FBS football.  According to our handy-dandy mathematical model, 32.5 of those 326 completed passes can be attributed to the fact that Kellen Moore plays for Boise State.  Contrast that with Robert Griffin III.  We can attribute only 3.7 of his 291 completed passes to the fact that Griffin plays for Baylor, in a much more competitive conference.

Adjusting for the competition, Kellen Moore has a completion percentage of 66.86%, which places him 6th in FBS football if we make the same adjustment to every quarterback.  While not the super-human 74.3% without adjusting, 66.86% completion percentage is still good enough to believe that Kellen Moore would have an effective professional career.  If Matthew Stafford ever goes down again, I think the Lions will find that they got a fantastic bargain for Kellen Moore.

A Quick Side Trip Into Sand

Categories: General Info, Statistics
Comments: 1 Comment
Published on: October 25, 2012

Last time, I offered up some quick thoughts on why Kellen Moore is better than the experts think.  Before I get to the second part of my discussion of Kellen Moore and why the arguments against him being an NFL caliber quarterback should be ignored, we need to take a little side trip.  I need to tell you about why football is unique among American sports.  And to do that, I need to tell you about my wedding ceremony.

When my wife and I got married, we decided we wanted to do something different.  The standard light-the-unity-candle thing wasn’t for us.  We wanted a different symbol to represent joining our lives together.  We found the sand ceremony.  In a sand ceremony, the bride and groom are each given a vile of sand.  My sand was grey and hers was pink.  The couple pours the two different sands into a common container.  The ceremony represents that your two lives are now intertwined and, just like the sand, can never be completely separated again.

Football is the sand ceremony of sports.  You see, most other American sports can be broken into individual efforts rather easily.  Baseball is the best example of this.  Each pitch is about one pitcher and one batter attempting to accomplish their goals.  The pitcher wants an out, the batter wants to get on base and we can easily see when each individual accomplishes that goal.

Even basketball can be broken into individual achievements rather easily.  Wages of Wins demonstrates this well.  If you are interested, they have excellent write ups about how to assign a value of wins produced to each player’s contributions.  Hockey could also be analyzed using this same type of analysis, provided the sport drastically changes the data it collects in a box score.

In all these sports, individual performance can be attributed directly or easily inferred from individual statistics.  If a baseball player has a low on-base percentage, it’s because he has trouble getting on base.  If a basketball player has a lot of offensive rebounds, it’s because he is good at going up and getting the rock.

But football is different.  It’s the only major American sport where your teammates must do their jobs before you can do your job.  A quarterback can throw the ball in the perfect spot, but if the receiver drops that perfect throw it doesn’t do anybody any good.  The performance of the quarterback and the wide receiver are permanently tied together and, just like the sand ceremony, can never be completely separated from one another.

This creates a major problem for statistically analyzing the game of football.  One of my favorite lines in any academic paper I’ve ever read is when Berri and Simmons are discussing quarterback fumbles.  They say “All of the measures of quarterback performance considered so far are infected by the performance of teammates.” [italics added for emphasis].  When I read that I imagined a spreadsheet stuck in bed, thermometer sticking out of its mouth, and a mother spreadsheet standing over it with a caring yet impatient look on her face.  If only that poor spreadsheet could shake off this infection, it could get back to doing some real, useful work :D.

But that’s never going to happen.  The nature of football is that the measures are intertwined and there’s nothing that can be done to change that.  Rather than lamenting that fact, instead we should embrace it.  We can’t ever separate the sand, but there is an important aspect of the sand ceremony I mentioned briefly.  The sands are different colors.  The patterns that the different colors of sand make in the common container aren’t trivial.  My wife and I can look at our sand ceremony container and see that there is more pink sand on the bottom and more grey sand on the top.  From that we can understand that she was pouring her sand faster at the beginning and ran out of sand before I did.  The same thing can be done for statistically analyzing football.  There are statistical techniques available to us that will “color the sand” so to speak.

And that is what I will be doing with this blog.  My model colors the sands of football so that we can recognize who is responsible for what.  We’re going to start with quarterback and wide receivers because their data is the easiest to collect.  If I ever get the time, the urge, and the ability to collect the data about other positions, I will certainly analyze them as well.  Next time, I promise we’ll get to Kellen Moore and what patterns we can see in his sand.

The Sad Case of Kellen Moore, Part 1

Categories: NCAA FBS, NFL, Statistics
Comments: 2 Comments
Published on: October 23, 2012

I think the 2012 quarterback draft class will be remembered fondly.  It seems like one of those classes where you’ll be watching ESPN in a couple years and they’ll pop up some feature story about the “Magnificent ’12’s” or some such nonsense.  What makes this quarterback draft class so different from so many others is that so many of the top prospects are getting playing time (top prospects according to my model, which I’ll tell you about in good time :D).  Luck, Griffin, Wilson, and Weeden are already starters.  If Vick’s popularity keeps falling in Philly, you’ll probably see Nick Foles get into the action before too long.  However, there is one top prospect player that isn’t getting the love I believe he deserves, Kellen Moore.

Kellen has an interesting story.  He has more wins at the college level than any other quarterback.  In four years at Boise State he amassed gaudy looking statistics.  Yet on draft day, he went undrafted and is currently a third-stringer with the Lions.  Why?

Looking back at blogs and ratings of Kellen, there seems to be two major problems.  The first problem is physical.  He’s short for a “prototypical” NFL quarterback and he, apparently, doesn’t have NFL arm strength.  The second problem is that he plays for Boise State or, more specifically, that Boise State doesn’t play against “elite” competition.  We’ll tackle the first problem now and save the second one for next time.

Kellen is short

Kellen Moore is slightly less than 6’ tall.  The question is, should that impact our evaluation?  Obviously we can all think of single cases in which height doesn’t matter.  Brock Osweiler at 6’7” was decidedly average at Arizona State and Russel Wilson is currently doing quite well for the Seahawks at less than 6’.  But can we get a more systematic analysis?

Berri & Simmons conducted just such a study.  They examined data from 1970 – 2007 and analyzed whether height predicted better NFL performance.  They find that, while height does predict draft position, it doesn’t predict how well quarterbacks do once they get into the NFL.  Draft decision makers seem to take height into account when drafting.  Unfortunately, they aren’t getting any value for choosing based on height.  The argument that Kellen Moore couldn’t play in the NFL because he’s short seems to go out the window.

 Kellen doesn’t have a strong arm

This one I don’t know how to address.  Specifically because I don’t know how we measure arm strength.  Do we measure the speed the ball comes out of the quarterback’s hand?  How far he can throw the ball?  How do we know Kellen doesn’t have a strong arm?  And, more to the point, does it matter?  Can Kellen develop a stronger arm?  I don’t know that the data exists out there to answer these questions, yet the answers are vitally important before we use a concept like “arm strength” to draft quarterbacks.

That’s all for this time.  Next time we tackle the question of competition level in FBS College Football.

Welcome

Categories: General Info
Tags: No Tags
Comments: No Comments
Published on: October 21, 2012

Welcome to Football Figures Online.  Here you will find advanced statistical analyses that can be used to evaluate individual football players.  To start, our focus will be mostly on quarterbacks and wide receivers as these are the positions that our model best predicts.  We’ll look at both NFL and NCAA players and also make predictions of which college players will make the best pros.  All of the statements you will read here will be based entirely on statistical analyses and publicly available data.
The full details of the model will be available just as soon as I can get it academically published.  In case you’re curious, this the list of Top 6 quarterbacks available in the 2012 draft class according to the model.  I included 6 because 1 & 2 are essentially a tie (also because including 6 means that 2/3rds of the list are current starting QBs in the NFL :D).
1.  Robert Griffin III (Baylor)
2.  Andrew Luck  (Stanford)
3.  Russel Wilson (Wisconsin)
4.  Kellen Moore (Boise State)
5.  Nick Foles (Arizona)
6.  Brandon Weeden (Oklahoma State)

page 1 of 1
Welcome , today is Wednesday, August 16, 2017