SpartanResource.com ~ BEATchUMps.com ~ DUMPthechUMps.com

Friday, July 12, 2013

The Myth Of A Widening Recruiting Gap - Part II

In Part I, it was shown, using 5-year rolling averages of ranking data obtained from Scout.com that the gap between UM and MSU recruiting rankings was actually following a closing trend. This contradicts the chUMp claims that the gap is widening.

At the end of Part I, an interesting trend was mentioned concerning the numbers of recruits in those recruiting classes. A hypotheses was proposed that the trend involving the numbers of recruits might actually prove to show the gap was closing even more than was initially calculated.

In fact, it is.

Before proceeding, we must note that after the 2006 class, Scout changed its algorithm for computing the recruiting class scores. Therefore rankings were used rather than the class scores. Using the rankings imposes a quantification effect. What this means is that the difference between numerically sequential ranks could be caused by a delta score of just 1 or of 1 million. This means that the delta scores between ranks is not uniform. That is always true with any system that ranks scores.

Not sure what that means? Let's look at a simple example.

Rank  Score  Delta (Score)
1         3401         -
2         3290     111
3         3289         1
4         3100     189
5         3005       95

The ranks are evenly spaced, but the scores associated with the ranks are not. This is a quantization effect.

The reason to mention this is not that it is significant to the analysis. It is actually useful. It allows the different scoring algorithms to be neutralized.

To analyze the hypothesis that normalizing the rankings based on the number of recruits was tested. Since the chUMps declared UM was superior the decision was made to leave the UM data as the baseline and normalize the MSU recruiting data to the number of UM recruits.

The math is simple.

MSU SCORE / # MSU RECRUITS * # UM RECRUITS = normalized MSU SCORE

The normalized scores were computed for each year and a new ranking for MSU was obtained from the Scout.com data base using those normalized scores. The Scout.com limit of 25 recruits per class was included in the analysis.

The results, again using 5-year rolling averages of the class rankings:

Years    UM    MSU  delta
02-06     8.6    49.8   41.2
03-07     6.8    53.0   46.2
04-08     6.4    43.2   36.8
05-09     8.2    47.2   39.0
06-10   10.2    42.6   32.4
07-11   14.2    38.0   23.8
08-12   13.0    32.2   19.2
09-13   12.2    25.6   13.4
10-14   10.8    20.8   10.0

Scout.com uses a scoring system that has factors that resemble exponential funcions. Mentioned above was how integer rankings quantizes scores. Scout.com has a scoring system that puts that on steroids.

Stars  Points
    5      200
    4      120
    3        40
    2        20
 NR          1
 JC           1
 PG          1

To the aggregate star score is the top 100 points. This means that the higher up in the rankings, the more points it takes to achieve the next higher ranking position.

Since UM is generally ranked higher than MSU over the 9 data sets, it seemed worthwhile to reverse the normalization and alter the UM data based on the MSU number of recruits.

The results are interesting.

Years    UM    MSU  delta
02-06     7.4    46.2   38.8
03-07     5.0    48.6   43.6
04-08     7.4    40.6   33.2
05-09     8.8    45.4   36.6
06-10   12.4    43.8   31.4
07-11   13.2    40.6   27.4
08-12   13.2    37.8   24.6
09-13   11.6    35.6   24.0
10-14   12.2    31.8   19.6

The trend is not as steep as when the MSU number of recruits is normalized.

While it is clear that the analysis supports the hypothesis that normalizing the number of recruits would show that the trend in the gap between the recruiting class rankings is narrowing faster, the algorithm Scout employs for calculating class ranking scores does not support a rigorous analysis leading to a definitive answer.

Perhaps the rolling 5-year averages of the win loss records and margins of victory over the same time frame will illuminate. MoV is MSU score minus UM score regardless of winner.

Years   WL%   MoV
02-06  0.0000  -16.4
03-07  0.0000    -8.0
04-08  0.2000    -3.8
05-09  0.4000    -1.0
06-10  0.6000   +3.0
07-11  0.8000   +9.4
08-12  0.8000   +9.8

The trends are clear. The gap is closing. The only question that remains indeterminate is: by exactly how much.

And for how long. The last time we saw these kinds of trends was in the late 1940s when Biggie Munn was transforming MSU (nee MSC) into the Nation's Football Factory.

Part I

------------- Remember to check Spartan Headline links, updated real time, at the bottom of the page. -------------- Please click on COMMENTS below the post to enter and view reader comments. ***If you are receiving this post via automated email, ***you need to go to the site to view headline links and embedded videos in this post: http://spartanresource.blogspot.com/

4 comments:

  1. Fantastic information from these stats!

    The last chart where you analyze the score differential over five-year rolling periods is especially awesome. That paints a picture of what it has felt like, and what it looked like to me when I was checking the scores recently.

    This is the type of thing that Coach Dantonio feels on a daily basis, as he has not only experienced it (with us), he has created it (for us). It's that type of daily reality that occasionally slips through with a comment like "Where's the threat"...to him, it really feels like that, and but for a field goal missed/made, it would feel like that to everybody else right now as well.

    This statistical presentation quantifies our belief that our MSU Football team will bounce back well this year after so many close losses last season. Thanks for the breakdown!

    ReplyDelete
  2. SM82 - You have a way with words. Framing it the way you did from Dantonio's perspective just grabs me as "ah-ha"! It is so sensible.

    MSU76, Congrats on you first post already rising to #7 on the all-time most viewed posts! Excellent work. We may have to hire a translator.

    A-Town's first post is about to crack the top 10 any day.

    It would be great to have an email conversation with SM82, MSU76, and ATown, then edit it down to something "post-able". I think it would be really entertaining and informative for BEATchUMps readers. Three smart guys talking Spartans sports. Just a thought...

    ReplyDelete
  3. As posted on M-Live in response to an Ann Arbor fan commenting on a post about the recent verbal commitment of Matt Morrissey. As always, the comments were not really about the posted article, but just more of the same MSU vs Ann Arbor banter.

    Dear derogatory misspelled screen name guy,

    While we can debate the semantics of the term "dominate", I thought the following factual statistical data would be enlightening and less subjective.
    Below is the Spartans winning percentage and margin of loss/victory in consecutive 5-year rolling averages vs. Ann Arbor. 5-year increments is relevant since most players, who complete eligibility, are in the program for ideally 5-seasons after signing day.

    Years W% MoV
    02-06 0.0000 -16.4 (ouch!!!)
    03-07 0.0000 -8.0
    04-08 0.2000 -3.8
    05-09 0.4000 -1.0
    06-10 0.6000 +3.0
    07-11 0.8000 +9.4
    08-12 0.8000 +9.8

    That is what you call a real trend. It's not a feeling or a hope or projection or a put down or a troll lob of a bomb. It is an actual statistically significant trend. (I know you can throw out the 134 years of data. We have all seen that a million times - I'm talking current TREND - data analysis of which you may have not been aware.) Until statistical results compile over time to reverse this trend, it IS the CURRENT TREND.

    If you like this kind of hard core data analysis, MSUSpartan has posted Part II on "The Myth of the Widening Recruiting Gap" at http://spartanresource.blogspot.com/2013/07/the-myth-of-widening-recruiting-gap_12.html

    ReplyDelete
  4. Part III is in the works, guys.

    It will be an eye opener...

    ReplyDelete

Please sign in using the method most convenient for you. We do not receive your login information. This function is provided by Blogger.