By Alex Koenig, Tyler Gamble and Michael Owen
Twenty-two years ago, legendary Wolverines football coach turned Athletic Director Bo Schembechler gave a press conference that created one of most overused and ambiguous terms in sports: the concept of the “Michigan Man.”
Since then, dozens of coaches have been hired, fired, considered and dismissed based upon something that would appear to be much more relevant for a number of other career paths, outside of athletics: their alma mater.
The logic follows that someone coaching at a university where they have previously played, coached or attended will not only be more popular amongst fans, but also more successful – after all, popularity and success seem to have a pretty high correlation coefficient.
When Schembechler bellowed “A Michigan Man will coach Michigan!” he was talking about the 1989 Wolverine’s basketball team, that went on to win the National Championship under Illinois State grad Steve Fisher after Michigan Man Bill Frieder left for Arizona State before the NCAA tournament.
However, the notion that coaches are better suited to coaching at their alma maters still holds, and so we looked at the careers of 629 Division I Basketball coaches who had been active in the last 10 years to see if a) Coaches performed better at their alma mater and b) if not, were there any other factors that did appear to have a significant effect on coaching success.
For the purposes of this study, success is defined as the coaches Simple Rating System (SRS) with a given team (Click here for more on how SRS is calculated). We then compiled a list of 7 different variables that we thought might have an effect on coaching success (represented by SRS). These variables included whether or not a coach is coaching at his alma mater, years of experience as an assistant coach, years of experience as a head coach, age, whether or not a coach is a head coach at a school where he previously served as an assistant, number of head coaching jobs held, and whether or not the coach is employed at a “power 6” or BCS conference.
The alma mater variable was chosen to test whether coaches who work at a school that they attended (and thus a school where they have a detailed working knowledge of the system, relationships with members of the program, and a deep emotional attachment too) will be more successful than coaches brought in from the outside. The years as an assistant and head coach variables were chosen (individually) to test the effect that experience being around and coaching basketball or (in the case of head coach) actively running a program has on coaching success. The variable describing whether or not a coach is a head coach at a school where he previously served as an assistant, will tell us something along the lines of the alma mater variable, whether or not familiarity with a program yields success. The Number of head coaching jobs held will tell us whether program builders (who stay at a program for an extended period of time) or program hoppers (who climb the career ladder) are more successful as a whole. And the Big Six variable will describe whether success is easier to achieve at larger programs, like those found in the Power 6 conferences.
These variables should paint a largely accurate picture of what goes into being a successful head coach, but obviously, they are only part of the larger mural. Recruiting is an enormous factor in college basketball success, and it is one that is incredibly difficult to analyze as only the very top recruits in the country are ranked, and thus the vast majority of head coaches have very little or no data available on their recruiting classes.
The stepwise regression yielded a model with an R-squared of 0.2467 and a Root MSE of 7.999 and indicated only three statistically significant variables in predicting a coach’s SRS: The number of years as a head coach prior to being hired, the number of previous head coaching jobs and coaching at a Big Six school. Notable absent are the alma mater and assistant coaching dummy variables. They clocked in with p-values of 0.312 and 0.101. In order to be considered significant, the p-value must be below 0.05
The predicted value equation is given below:
SRS-predicted = -6.262 + 9.175*(Big Six) + 3.113*(#of previous head coaching jobs) – .2614*(years as a head coach prior to being hired)
Looking at the significant variables defined by the stepwise regression model it can be seen that the number of head coaching stops that a coach has experienced has a coefficient of 3.114. This means that for every additional head coaching job a coach has had prior to his current position, his future SRS can be expected to raise by 3.114. Also, the number of years of head coaching experience that a coach has carries with it a coefficient of -.261. This means that head coaching experience is negatively associated with SRS. For every additional year of head coaching experience, a coaches SRS at a future position can be expected to fall by .261. The dummy variable describing whether or not a head coach has a position at a power 6 conference school carries with it a coefficient of 9.175. This strong positive effect tells us that if someone is employed at a power 6 school, they can be expected to have an SRS 9.175 points higher than their counterparts at a mid-major institution. These three variables had the following t-values, indicating their significance level:
Big Six = 11.52
# of previous head coaching jobs = 4.78
Yrs as head coach = -2.22
Looking at this data, some interesting conclusions can be reached. Coaches who are switching jobs often, presumably moving up the career ladder, seem to coach teams that perform better and thus have a higher SRS value. On the other hand, coaches with more experience entering a new job tend to perform worse.
These apparently contradictory points can be reconciled by considering that coaches who are talented are going to find a good job (one where they are successful and thus remain employed) after a certain amount of years. Coaches who are at one place for a long time and then fired and/or switching jobs can often be less talented. Perhaps their longevity was due to an Athletic Director with low standards or an apathetic administration. It can be assumed that a certain amount of years of head coaching experience are needed to assure that a coach has talent, but after a coach exceeds that – admittedly undefined – number, a coach’s talent is clear and he will most likely cease to develop more.
Thus, great coaches generally have found their final destination after a certain amount of years, and further research should be done into determining what the optimum amount of years of experience for a potential employee to have is. The lesson for Athletic Department’s here is that they should look to hire coaches who have been successful at multiple stops, hot coaches climbing the ladder, all the while avoiding coaches that have lingered at positions for a long time (maybe accumulating wins) but underperforming. The positive correlation between coaching in a big six conference and having a high SRS value can be explained by the fact that, for the most part, coaches at bigger schools have more resources at their disposal and a more rich basketball history. In short, Success is easier attainable at these schools. Thus, when hiring a coach, athletic departments should take into account that mediocre coaches can perform better at a BCS school than good coaches at a mid-major school. Though they may still be less talented coaches, the difference in their performances can be accounted for by the program at which they were coaching.
One potential weakness in this study is the fact that there is a high correlation (.8344) between years as a head coach prior to being hired and number of previous head coaching jobs held. The reason these are kept as two separate variables instead of combining them into one is because we felt that there was the possibility that the variables had different effects on head coaching success, which was shown by the opposite signs of the coefficients of the respective variables.
Another weakness in the study’s results is the lack of accounting for the impact of player talent on coaching success. For example, if there are two teams with identical schedules and one team is comprised of All-Americans that are the best players in college basketball and the other comprised of average basketball players the team with much more talent on its roster will perform much better than the average team. The head coach of a team cannot affect the level of talent that he has to work with – beyond recruiting and better drills – he can only attempt to maximize the abilities of the team he has. The coach of the team with superior talent needs only to arrive at practice and roll out the balls for his superiorly talented players and then he will appear to be incredibly successful in comparison to other coaches. This will lead to the perception that he is a more talented coach when, in fact, it is impossible to tell how much the coach impacts his team’s success and how much impact the players have on team success. It would have been incredibly difficult for this study to create variables that accounted for the talent level of the players on the roster.
Also, there is very little data out there that allows for successful quantification of recruiting or player development ability directly, two of the most important facets of coaching. The best that can be done is to evaluate a team’s final results and hope to glean some useful information for them. This study sheds light on only part of the picture of coaching success. However, the fact that it seems to dispel the theory of alumni having more success (i.e. the myth of the “Michigan Man”) and that it does reveal certain significant factors in predicting coaching success can be relevant both to fans and administrators alike.
Winthrop Intelligence contributed to the data set used for this study