Tuesday, February 26, 2013

Breaking down FBS QBs from 2007-2012

Would you rather have a Freshman or Senior under center?  Does it matter?  Maybe, maybe not.  Play with this viz for a bit and make the call for yourself.  Most everything is interactive in one way or another.

Pick a season (2007-2012) and dig in.  Freshman QBs are in red, Sophmore QBs in orange, Junior QBs, in blue, and Senior QBs in Blue.  

Not all QBs are here.  I only went with QBs who accounted for more than 50% of the team's pass attempts for the season.  It's a good proxy for 'starting QB'.  

Thursday, February 21, 2013

How Much Does Losing an Experience QB Hurt?

An experienced QB brings much value to a team.  A lot of that value is in the intangibles, but some of it can be directly measured.

I looked at all teams from 2007-2012 who replaced a QB with 2 or more years of significant playing time, defined as greater than 100 pass attempts per year, with a QB who did not have that amount of experience.  In most, but not all, cases, this is the same as replacing a multi-year starter with a new QB.

I did not consider the school class (FR-SR) of the new QB, and because of my definition of significant experience, it does not include all possible QB replacement situations.

As measurement space I looked at the average Offensive Yards Per Game with the experienced QB and the same statistic the following year with a new QB.

Overall, what I found was that teams experienced a decline in YPG approximately 60% of the time, with an overall average decline in YPG of approximately 4%.  The amount of decline varies dramatically when I consider the average YPG of the team and the number of years of experience the QB to replaced had.

This chart breaks out the changes by years of experience.  Points on the graph further from the center line indicate a greater change in absolute YPG (not percent).  Points above the line indicate years that the YPG increased with the new QB, points below indicate a decrease. 

Of note to Nebraska fans, who will be enjoying the services of Taylor Martinez as a 4th year starter in 2013, the probability of a team which must replace a 4-year starter experiencing a decline in YPG is about 70%, with an average decline in YPG of almost 16%.  

In 2012 Nebraska had an average offensive YPG of 460 which ranked 26th in the country.  Assuming a similar number for 2013, Nebraska should expect a decline in YPG in 2014 of .112 (.70 x .16), or 51 YPG.  This year, 409 YPG was good for 55th in the country.

When I group the average YPG into bins of 50 YPG, and don't consider the number of years of experience at QB that NU has to replace, NU falls into the 450-500 YPG bin.  Looking at the data this way, NU has a 22% probability of remaining in the same YPG bin, a 28% probability of falling to 400-450 YPG, 33% probability of falling to 350-400 YPG, and an 11% probability of falling to 300-350 YPG.  All is not lost, however.  There is a 6% probability of actually improving to greater than 500 YPG.

(note that the 72% probability of decrease in this model is almost the same as the 70% probability of decrease in the first)

There are many factors that I didn't consider in this.  For instance, I didn't consider the number of years that the replacement QB was the understudy, or how much game time the replacement had if he didn't reach the 100 pass atts threshold.  All of this could impact the results.  

In NU's case, however, I believe it would make the situation bleaker.  It would be a difficult case to make that Coach Pelini has done a good job of getting his backup QBs significant experience.  So, if this were considered for other teams, NU's backups lack of experience would likely make the probabilities of an offensive decline even greater than they already are.

What can be done?  Coach Pelini must get his backups as much experience as possible.  An offensive dropoff may be inevitable, but getting the backup as much real game experience as possible may make the difference between a decline of 5% and 20%. 

Monday, February 18, 2013

Population (in millions) within 500 miles of NCAA Schools

Map point labels are the total population within 500 miles of a school. 

Map background is the relative population of states from the 2010 census...darker indicates greater population.

Charting Rivals Team Rank and Recruit Densities

Here are 4 quad charts that illustrate the relationship between Rivals Team Recruiting Rank and the density of recruits in a given mile radius around a school.

For this analysis, recruit density is the number of Rivals recruits of a given star level located within 500 miles of a given school, normalized to number of recruits per 1,000,000 population.  Density rank is a 1-122 rank of the density at each measurement level.

The quadrants, starting in the upper left and rotating clockwise, are:

1.  Low Rivals Rank and Low Recruit Density
2.  High Rivals Rank and Low Recruit Density
3.  High Rivals Rank and High Recruit Density
4.  Low Rivals Rank and High Recruit Density.

One can think of the relationship as signifying the degree to which a school overcomes distance and population barriers to pull in talented recruits.  Starting from lower left to upper right, moving up and to the right implies that a school is doing better at overcoming these barriers.

Wednesday, February 13, 2013

Percentage of available recruits within 100 miles who signed with each school

Here's a quad-chart showing the percentage of Rival's 5-, 4-, 3-, and 2-star 2013 recruits within 100 miles of each school who eventually signed with that school.  The greater the diameter of the circle over each school city, the greater the percentage of available recruits signed by the school.

Download the underlying data here.

Location! Location! Location! Part 2

My last post illustrated that schools in the SEC footprint, along the East Coast, and California have a greater number of highly rated recruits within 100 and 250 miles than schools in the Midwest.

This post will look at whether recruits have a greater propensity to attend schools closer to home. My next post will look at whether the recruits Rivals star rating impacts that decision.

Please note that when I use the word 'random' or 'chance' I mean it in the sense that there is no decision criteria applied to the choice whatsoever.  In reality, distance from home to school is only one of many possible decision criteria.  This analysis does not yet explore the degree to which the criterion of distance from home affects the decision, only whether it appears to affect it or not.

First, I made a matrix that calculated the distance from every 2013 recruit on the Rivals website to every FBS school.  It was, to say the least, a large matrix (2677 recruits x 122 FBS school = 326,593 distance calculations).

Taking those 326,593 distances and grouping them into bins with a width of 50 miles and plotting the number in each bin on the vertical axis and the bins on the horizontal axis I get a picture of how far recruits live from schools.

The peak of the graph at around 600 miles means that the most common distance between every recruit home and every possible school destination is about 600 miles.

The first indication that recruits are more likely to sign with a school closer to home than if the decision was completely independent of the distance is apparent when I add the distances to the schools that recruits actually signed LOIs for.  The added red line clearly shows that schools closer to home were preferred by recruits.  (the counts for the red line are on the right vertical axis).

The overall average distance from every recruit home to every FBS school is 1082 miles.  The overall average distance from every recruits home to the school they chose to attend is 512 miles.  The conclusion that should be drawn from this is that the distance from home to school appears to be a factor in recruits' decisions.  The actual distance to the schools that recruits choose to attend is less than half that of what would be expected if the decision to attend a particular school were truly random.

My next post will explore whether the Rivals rating of a particular recruit affects the propensity of a recruit to sign with a school further from or closer to home.


Tuesday, February 12, 2013

Location, Location, Location!

This is a sneak peek at some work I'm doing on the impact that geography has on recruiting.
First up...6 maps that reflect how many 2013 5-, 4-, and 3- star recruits (as determined by Rivals.com) lived with 100 and 250 miles of FBS schools. The greater the diameter of the circle, the more players of that rating there are within the 100 or 250 mile radius around the school.
As always, you can download the underlying data from my public dropbox.

Wednesday, February 6, 2013

Visualizing the 2013 Big Ten Recruits

Does the SEC's oversigning make a difference? Part 2

In an earlier post I showed that there is a statistically significant difference in the average number of recruits signed by SEC teams and the average number of recruits signed by the Big12, the conference with the next highest average number of signees, and by extension, all other FBS conferences.

Answering the question of whether those extra signees translates into success on the field is an entirely different question.  And it's a question that may have to wait one more post.  This post will focus on who the recruits that make up those extra signees are.

Digging into the data, I was able to uncover one piece of information that may be important in answering this question...or at the very least helps understand this advantage/perceived advantage more fully.

This chart shows the coefficients of correlation between 4-year averages of Top-100s, 5-, 4-, and 3- Star recruits, and the average Scout.com recruiting ranking.  An negative number implies that as one goes up, the other goes down.  As the numbers expand outward from 0 towards 1 and -1 they indicate a stronger correlation between the two statistics.

Note:  I'm only looking at teams who finished in the AP Top 25.  Doing this is an attempt to reduce the signal to noise ratio...which is very high in this case.

Normally, I would not key in on a coefficient of correlation (CC) of .25 as anything noteworthy.  But for this particular analysis, that number may indicate something worth examining.

I say that because when I look at the other CCs, I see that there is a very weak negative correlation between average numbers of signees and Top-100 and average number of 5-Star players.  There is a weak positive correlation between average number of 4-star players.  

And then, with a CC of .25...virtually towering over its neighbors, is the average number of 3-star players.

But what does it mean?  To me, this may be evidence of who those extra signees are.  The SEC doesn't appear to be adding addtional 5- and 4- stars with their oversigning, but may be signing more than their 'fair share' of 3-stars.

This next chart breaks the relationship between the 4-year average number of signees and the 4-year average number of 3-stars into the SEC, B12, and the rest of the NCAA (not including SEC or B12).

Other than the clear advantage illustrated by the SEC's higher trendline, meaning an greater average number of 3-stars for the same number of signees, the slope of the SEC line (shown in the purple equation box on the lower right) is steeper.  The implication of this is that as the average number of signees increases, the average number of 3-stars in the SEC increases more quickly than for the rest of the NCAA, and slightly more than the average number of 3-stars in the B12.
 In other posts I've illustrated the dominance of the SEC in terms of numbers.  Considering this, and the SEC's greater number of 3-stars per number of signees, the implication is that oversigning by the SEC further reduces the number of 3-stars available to the rest of the NCAA...teams that find their core talent mostly in 3-star players.

To me, it's hard to believe that this is good for the college football world.  It's probably a good thing for SEC fans, but the trend indicates that unless oversigning is abolished, which would free up a significant number of 3-stars for other teams, the SEC's domination will remain largely unchanged.

You can download the data for this analysis from my public dropbox.

Tuesday, February 5, 2013

Where are the prized recruits playing?

Here are two similar infographics.  Each portrays 5-, 4-, and 3- star recruits, by conference, from 2005 to 2012.  The numbers of recruits are all 4-year moving averages, meaning the data reported for 2007 is an average of 2004-2007.

You see 40-star recruits on the X-axis, and 5-star recruits on the Y-axis.  The color is determined by the number of 3-star recruits, dark red (less) to dark green (more).  Finally, the size of each circle is the average final AP ranking of the teams in the conference.  Unranked teams have a ranking of zero, so they are filtered out.

It's not surprise that the SEC dominates the recruiting world, but this chart, which show the total numbers of highly prized recruits, illustrates that very clearly.  The SEC's positions in the upper right corner indicates that as a conference, the teams in the SEC which finished ranked in the AP Top-25 signed nearly twice as many 4-star recruits as any other conference and 30% more 5-star recruits.  The dark green color of the SEC circle also shows that they landed the most 3-star recruits. Finally, the size of the SEC circle indicates the number of teams which finished ranked in the AP Top 25 was better than any other conference.

This is a pretty clear picture of where prized recruits have gone for the last decade or so. But it doesn't tell the whole story. The next infographic, which presents the same data as an average for each conference team, seems to indicate that the per team average for the SEC isn't as well distributed as it is for some other conferences. One possible reason for the difference in appearance between the graphs is that the talent, while significant, in the SEC is more concentrated on a few elite teams, while in the Pac 10/12 it is more spread around. I haven't fully investigated this yet, but I will.

Another thing that jumps out at me is the wide disparity in average number of 4-star and 3-star recruits between the Big Four conferences (SEC, Pac10/12, Big12, and B1G) and the rest of college football. There's simply no middle ground on 4-stars. And the dark greens of those Big Four conferences contrast well with the other teams, indicating an advantage in the number of three stars as well.

 The data for this analysis is available for download from my public dropbox account.

Sunday, February 3, 2013

Where Do Colleges Find Players?

Here's a look at where ~21,500 FBS players came from. Not surprisingly, the SEC footprint, Texas, and California dominate.