It' not a Regresson of Means, but it's close.....I've been doing 2 a days behind the scene.
Introduction
Heard it once; heard it a thousand times. Mark Richt is on the hot seat for not winning enough, and eventually the rhetoric, argument, comes around to CMR teams play un-sound football because his teams are un-sound, undisciplined off the field. Rather than letting that float nebulously in the blogosphere and on talk radio, I decided to test that theory.
My central question was there a link between negative
off field incidents, “distractions”, and undisciplined
performance on the field. I decided to use Fulmer Cup (FC) points as my measure for off field incidents, my dependent variable, and I decided to test many independent variables against it. Most frequently mentioned variables for poor performance on the field variables are penalties and turnovers. I decided to test, both rank and value, for: penalty yards, penalties per game, interceptions, fumbles, turnovers and turnover margin (takes into account defense playing ‘sound’ and causing turnovers). Also, this means squadoosh if it isn’t backed up with wins. So I also measured wins ranking, winning percentage and actual win number. A total of 15 independent variables were considered.
Also, I was also studying if any thing could be extrapolated to CMR lead UGA teams. One could also argue that turnover margin, both rank and value are measureable variables for “aggression”, something that critics point to as lacking from CMR teams in general and specifically after Brian Vangorder left.
Methodology
Background:
The key statistical tool used was a
Pearson r test. A Pearson r test determines if there is a relationship between variables. It doesn’t determine causal effect. An analogy might be that umbrella use is higher on rainy days than sunny days, but umbrella use didn’t cause the rain. Since the dependent variable is off the field in nature; there is no basis to say scientifically there is a causal effect to on field activity, but you can measure if there is a link and whether or not that link rises to a level of scientific significance.
A Pearson r test is an index scale. The values range from -1 to +1. The closer to 0.000 the result is means that no relation exists. “Girl, I don’t even know who you are.” A value of +/- 0.500 means that there is statistical significance. A negative value means that there is an inverse action occurring. As one variable rises the other variable is decreasing. Think of a negative value like a seesaw. On the other hand, a positive value is somewhat a misnomer than the name implies. It is more like a swing, when one variable rises the other one does as well; when one variable decreases the other variable decreases, not that all values are increasing all the time.
A marriage analogy might help explain what type of relationship a Pearson r test is measuring. A value of -1 is like
War of the Roses. The 2 variables are at constant odds with one another. A reaction by one has the opposite effect on the other variable. A value of +1 is like
On Golden Pond. The 2 variables are soul mates. Again, -0.5 means one party has a divorce lawyer, but not hostile, and +0.5 mean the 2 are officially “an item”; a significant other.
Methods:
I used the SAS Wiki page
http://www.sportsargumentwiki.com/index.php?title=Fulmer_Cup , presumably EDSBS supported page, to download my FC historical data. (
I did write to one of the EDSBS authors to try to get first hand data, but I got no response, and used what I had.) I used the NCAA website to download my football statistics.
Within each year, the main stratification was 2 groups: Teams with FC points opposed to all teams. I also studied what was happening within the FC teams. That is “Was there an effect going on to how lawless the team was?” If a team has one point compared to 20 points for example, does that matter?
To provide uniformity through out the years, I threw out W. Kentucky, comparing not only apples to apples, but Granny Smith to Granny Smith. That gave each year an “n” of 119 or N = 476. Only 2006 had less than 49 teams in the FC group – so “good” subgroup size. Those values were n 09 fc = 49, n 08 fc = 57, n 07 fc = 55 and n 06 fc =34. Each year was studied independently and then assimilated all data by each variable, aggregate, and examined that effect.
Also inherent to wins rank (vs. winning percentage and win value) and turnover margin rank (vs. turnover margin), these 2 variables are inversely related already, meaning for all other statistics the lower the rank the lower absolute value. Example interceptions, the lower the rank the lower number of passes picked. With wins rank and turnover margin rank, the lower the rank the higher the value. Example a turnover margin rank of 1 might have a value of +2.0, meaning two extra possessions per game, and of course win rank o1 likely means you have 13 or 14 wins. The Pearson r test is measuring the inverse of the inverse in these 2 situations.
Additionally just to get a macro perspective, I used averages for FC teams versus the teams without points in a given season in addition too the Pearson r test.
Findings
There were 3 independent variables which had a relationship link trending all the same way for all years and the aggregate total. Those 3 were: wins, penalty yards rank and total turnover rank. These were the most consistent variable predictors. The FC subgroup averaged at least (1) win more per season than teams that didn’t have any FC points and that trend was also true in aggregate tally as well. Penalty yards rank, for the FC subgroup only, had a negative index trend for all periods, favorable; meaning that more FC points the LESS penalty yards rank and vice versa. Also means that FC subgroup doesn’t play undisciplined football in terms of being measure by penalty yards rankings. In fact the opposite was true – more FC points the better rank. Finally, turnover rank, for both groups – with FC points and overall, had a positive link trend for all periods, unfavorable; meaning the more FC points MORE turnovers. With a fairly large N, but short sample timeframe – 4 seasons, in the end 2 variables that could be propitiously linked to FC points are wins and turnovers. All other variables revealed mix, inconsistent results.
None of the variables rose to the level of being scientifically significant, +/- 0.5, for any year, any variable, nor for any stratification. However, anecdotally the eventually BCS National Champion has had off field issues in the off season and was on the FC watch list each year that the FC award has been in place.
Each year that UGA had double digits in FC points produced a 10 win season. Is that a good omen with 17 points going into this season? Also of note the year in which CMR team had no FC points UGA’s win tally was 8 compared to his historical 10 win season and the lowest in CMR era. Unfavorably, UGA is within 2 points of over taking the lead from Florida for All Time FC points in the SEC, and 6 points for All Time Overall lead. Currently, UGA is tied with Minnesota this year for first. The AD situation could be a tiebreaker that leads to the hardware.
Also for UGA, separating them out as an independent sample group, they average 7.25 points worth of infractions per season; 3 DUI’s and a speeding ticket worth. Also after doing a Pearson r test on all variables (UGA alone), all came back as statistically significant, except penalty yards and penalties per game, both rank and value (11 of 15). BUT this dangerous interpretation given the sample size of only 4 values, you need more samples to be valid. Nevertheless, Winning percentage had a Pearson r value of 0.820, and Turnover margin had a value of 0.595 for example. Again, meaning that as FC points went up so did their values. Since BVG left after the 2006, it is difficulty to judge his impact alone on the FC point standings. As far as more generalized ‘aggression’ since 2007 UGA Turnover margin has gone from 0.69 to -0.23 to -1.23 in raw score. Certainly some evidence of playing a “softer defense”, but how much could have JC’s interceptions also played into that value?
There is 5 years worth of data for FC and 4 years worth of data for seasons played. The 5th year of FC data is still incomplete and there is still a season of football to go. There have been 8 teams to make the FC list, each and every year, up until this year. ‘Bama is in danger of falling off the consecutive year list. Up until now half of the 8 have been SEC schools: FL, ‘Bama, Arkansas, and S. Carolina. Three of past 4 years NC have came from that consecutive years list – thanks to the first (2) schools from that list. ‘Bama won the FC Trophy the year before they won the NC.
Teams on the All Time list by conference: SEC – all teams (Vandy 3 of 5), Big 10 – all teams (3 of Top 10 all time; NW 3 of 5 years, as well as incoming Nebraska and that would make 4 of Top 10), Pac 10 – all except AZ St (surprise somewhat) and Stanford, ACC – all except Wake Forest and Maryland (again surprise and Miami only had 4 points aggregately), Big 12 – all except Baylor and Big East – all teams. The SEC does have a 50 point lead over 2nd place Big 10. Winners of the FC Trophy have come from 3 of the Big 6 conferences: Big 10 (Illinois), SEC (‘Bama) and Big East (S. Fla). CUSA is also on the board (We are the
Thundering Terds Marshall). The Big 6 conferences have a total of 5 teams that are Lilly White all 5 years.
Speaking of being FC list virgins – never a point, collectively they win 43% of their games, a median season of 5-7 aggregately; worse than the 6 win season if failed to make the FC list in a given year. Outside the above mentioned teams, another mild surprise was the amount of Texas teams on that list: Rice, Baylor, N. TX, SMU and Houston. Given the historical significance, context, of off field trouble in the state that list was surprising large (~22% of the remaining teams from 1 state). Only 3 teams have managed be above .500 each year while maintaining a pristine off field record: C. Mich., Navy and the above mention Houston (Kevin Sumlin era). There were additional handful of teams that managed a .500 records aggregately, headlined by Wake Forest and ASU. There were 2 teams that were de-flowered this past off season Buffalo and Tulsa.
Conclusions
Given that NC have FC points, if I am a coach I’d rather be on the watch list than off it. Given the fact the FC point team win more, I’d rather be on it. The axiom about you don’t want a DE in the church choir holds some merit – Reggie White excluded.
Some questions that didn’t get answered would be “Given that fact that most Big 6 conference schools are on the list, would the results be identical if one stratified the teams that way versus by FC point teams or would the results be similar based on budget size of the program?” Will these trends continue? Will Kevin Sumlin’s name appear more often for elite jobs as a guy who wins and maintains a “clean” program, and oh by way runs a “sexy offense” that’s attractive for skilled, speed kids and who is black?
Besides X and O, besides recruiting, besides booster relations, a good coach will know how to manage off field distractions to yield positive results, because generally good teams face each year.
…. Good news for Oklahoma (on list this year) and bad news for ‘Bama (off, but still time), and UGA can look forward to another 10 win season and possible NC next year (like ‘Bama) if history repeats itself.
PS If you would like the excel files of my research, email me jimfrankli@gmail.com. Also for those that blog, any tips on displaying excel sheets and or graphs would be appreciated......my head hurts from banging it trying to display some of the results.