For office use only Team Control Number For office use only

26636
T1 ________________ F1 ________________
T2 ________________ F2 ________________
T3 ________________ F3 ________________
Problem Chosen
T4 ________________ B F4 ________________

An evaluation model of college coaches

Summary
In working out a evaluation system that would fairly judge the best all-time college
coaches, our team decided to apply the grey correlation analysis model. To figure out
parameters in this model, we firstly refer to some expertise and then select four indi-
cators that would be applied in general across both genders and all possible sports
–NCAA Champions, Pct, wins, and Coach Awards. We use fuzzy consistent matrix to
grant each kind of indicator a proper weight in our evaluation system. Then we use
specific statistics in basketball, football and ice hockey and give out the ranks of the
coaches. To further examine our model, we compare the top 5 coaches in our lists
with those in the reports from ESPN or other research institutes.

We also find out that the competition system of different events of NCAA will inevi-
tably change over time, and that the time line horizon we choose will result in differ-
ences in our four factors. The real data prove our supposition – in both of the basket-
ball and the football fields, the “ancient” coaches enjoy a significantly higher average
of winning percentage than the present coaches, while the other three factors show no
discrimination. To enhance the accuracy of the model, we modify our original model
with a correction factor that derives from the moving weighted average method to ad-
just the winning percentage over time, which is the shining point of our theory. In ad-
dition, we find out the significant difference among coaches before 1950s and those in
later periods, which is also taken into consideration in the adjustment of our model. In
terms of gender, we reach the conclusion that it has no influence on our selection of
factors and the metric of our assessment.

We test our model by examining its stability and sensitivity. In stability test, we
change the priority of the four factors to judge if the model is still applicable. Fur-
thermore, since fuzzy consistent matrix has inborn disadvantage in subjectivity,
Shannon’s entropy method is applied to substantiate the stability. In addition, we im-
plement sensitivity test by changing the  in the grey correlation analysis model.

In the last part of this report, we state the strengths and weaknesses of our model.
Convenience and universality are the greatest advantage of our model, while the sub-
jectivity in selecting indicators may still exist.

更多数学建模资料请关注微店店铺“数学建模学习交流”
https://k.weidian.com/RHO6PSpA
 Contents
1 Introduction ............................................................... 2
1.1 Background ....................................................................................................... 2
1.2 Our Work........................................................................................................... 2

2 Model .......................................................................... 3
2.1 Introduction ....................................................................................................... 3
2.2 Notation and definitions .................................................................................... 4
2.3 Model designing................................................................................................ 5

3 Application ................................................................. 7
3.1 Introduction ....................................................................................................... 7
3.2 The selection of factors ..................................................................................... 7
3.3 Data collection .................................................................................................. 8
3.4 Calculation ........................................................................................................ 8
3.5 Application of our model to other sports ........................................................ 13

4 Comparison of the ranks from different sources... 15

5 The analysis of time line horizon ............................ 16
5.1 Introduction ..................................................................................................... 16
5.2 Procedures ....................................................................................................... 16

6 The analysis of gender ............................................. 20

7 Stability test .............................................................. 21
7.1 Testing the influence of changing the method on the determination of weights .
…………………………………………………………………………………...21
7.2 Testing the influence of changing the comparative importance among the four
factors .................................................................................................................... 24

8 Sensitivity test .......................................................... 26

9 Strengths and Weaknesses....................................... 27
9.1 Strengths ......................................................................................................... 27
9.2 Weaknesses ..................................................................................................... 28

Reference ..................................................................... 28
Appendix ..................................................................... 29
An article for Sports Illustrated ................................ 36
Team #26636 Page 2 of 38

1 Introduction
1.1 Background

College sports teams occupy a pivotal position in the sporting world. The major col-
lege sports events, such as National Collegiate Athletic Association (NCAA) champi-
onships, have long been the focus of attention of the public. Each year, not only the
top sports teams but also the head coaches grab the eye of a considerable number of
people. To promote the development of college sports, people are increasingly apt to
seek for the help of modern science and technology to evaluate the performance of
college coaches.

Designing an objective and effective system to evaluate the performance of college

coaches has many advantages. Such system will enable either college sports teams or
mess media to analyze the strengths and weaknesses of a specific coach, as well as to
reveal the keys of victory of those legendary coaches. Current approaches of evalua-
tion, however, are neither quantitative nor broadly applicable.
Most of studies use subjective and qualitative methods for assessment. The methods
they use for collecting data include: Designing questionnaires for athletes to judge
their coaches (Rushall, B. S.,et, al.,1985), interviewing coaches (Cassidy.,et,al.,2006),
observation (Nelson, Lee J.,et,al., 2006), etc. Using the above methods, the evaluation
of coaches simply relies on human judgment, which is not optimal since numerous
data are accessible on the Internet.

1.2 Our work

As modelers, we are required to build a mathematical model for Sports Illustrator to

select the best college coach or coaches (male or female, past or present) in various
kinds of sports. In addition, we are required to discuss the influence of time line hori-
zon and gender on our model.

Instead of using subjective evaluation method, we adopt a quantitative method to

measure and rank the performance of college coaches.

 Firstly, we determine the elements of our evaluation system. In order to make our
model applicable in various kinds of sports and both genders, we select wins (the
number of times that one specific coach has successfully led a team to win), win-
ning percentage, the number of times that they had led their teams to win national
championships, and the number of times that they have won the title of Coach of
the Year as evaluation factors.
 Secondly, we create a fuzzy consistent matrix to determine the weight of each
factor. This approach includes processes of determining fuzzy preferential rela-
Team #26636 Page 3 of 38

tion matrix and fuzzy consistent matrix, as well as the introduction of root meth-
od. We will further explain the above processes in the next section.
 Thirdly, we develop a grey correlation analysis model to evaluate and rank the
performance of college coaches. This model includes the calculation of grey rela-
tional coefficient and grey rational degree, which will be displayed in the next
section.

2 Model
2.1 Introduction
The evaluation model we adopt is based on the principle of grey correlation analysis.
Using this approach, we can calculate the grey relational degree between the coaches
on our lists and the ideal coaches (the fictitious coaches with ideal data). Then we sort
the grey relational degrees by descending order to get the rank of the coaches. Note
that the prerequisite of our model is the determination of the weights of all factors.
This prerequisite can be met by establishing a fuzzy consistent matrix.

Therefore, our evaluation system can be divided into two parts:

The first part is the determination of the weight of each factor, which is critical in our
model and will exert vital influence on our conclusion. Theoretically speaking, both
subjective methods and objective methods can be applied to the determination of
weights. Subjective methods are mainly based on human judgment. Such methods
depend on the experience of experts. Currently prevailing subjective methods include:
The Delphi method (Hwang, C.L.; et, al., 1987), analytic hierarchy process (Saaty,
T.L., 1980), the weighted least square method (Chu, A.T.W.; et, al., 1979). Objective
methods involve the use of quantitative models. Based on the data collected from da-
tabases or websites, researchers can use mathematical models to determine the
weights of factors that mostly reflect the performance of the objects they study. Cur-
rently prevailing objective methods include: Entropy method (Shannon, C.E., 1948),
principle component analytical method (Fan, Z.P., 1996), etc. Therefore, considering
the features of NCAA competitions and the interrelationship among factors, we de-
termine the weights of factors by creating fuzzy consistent matrix.

The second part of our evaluation system is the grey relational analysis model. We use
this model to evaluate and rank coaches. Because of the scarcity of data sources and
the complexity of interrelationship among factors, it is difficult for modelers to ac-
quire enough information. Therefore, it is comparatively advantageous to use grey
relational analysis to analyze such a small-sample data system. This system can elim-
inate the subjectivity to some degree and will thusly lead to an objective conclusion.
Team #26636 Page 4 of 38

2.2 Notations and definitions

2.2.1 Notations
Notations Explanations
F Fuzzy consistent matrix
A Fuzzy preferential relation matrix
w Weights of factors
k The kth factor
Standard data array (The best value of
x 0( k )
factor k among all coaches)
Correlation data array (The raw data ar-
xi( k )
ray of factor k among all coaches)
i( k ) Grey relational coefficient
 Distinguishing coefficient
E Evaluation matrix
ri Grey correlational degree
R Grey comprehensive evaluation matrix

2.2.2 Definitions
Fuzzy consistent matrix
1) For fuzzy matrix F  ( fij )mm , if fij  fji  1, then this matrix is defined as fuzzy

reciprocal matrix. For any k ( k =1,2,…m), if fij  fik  fjk  0.5 , then matrix F is

defined as fuzzy consistent matrix.

2) For fuzzy consistent matrix F  ( fij )mm , if fij  0.5 , then element i and

element j are equally important. If 0  fij  0.5 , then element j is more important than

element i . The smaller fij is, the more important element j is than element i . If

0.5  fij  1 , then element i is more important than element j . The greater fij is, the more

important element i is than element j .

Fuzzy preferential relation matrix

We use three numbers (0, 0.5, and 1) to measure the comparative importance among
the factors. For fuzzy matrix A  ( aij )mm , if aij  0.5 , then element i and element j are
Team #26636 Page 5 of 38

equally important to the upper level. If aij  0 , then element j is more important to the
upper level than element i . If aij  1, then element i is more important to the upper lev-
el than element j .

For example, suppose that we have four factors: a , b , c , and d . If the degree of im-
portance to the upper level degrades from a to d , then we create the fuzzy preferential
relation matrix A as follows:
 0.5 1 1 1 
 
 0 0.5 1 1 
A
 0 0 0.5 1 
 
 0 0 0 0.5 

2.3 Model Designing

Process I: How to determine the weight of each factor
1) We create the fuzzy preferential relation matrix A .

2) We convert the fuzzy preferential relation matrix into a fuzzy consistent matrix:

Sum column vectors of each row in the fuzzy preferential matrix:

m
fi   aik , i  1, 2,..., m
k 1

Apply the following conversion:

fi  fj
fij   0.5
2m
Thus the fuzzy consistent matrix F  ( fij )mm is created.

3) We use the root method to determine the weights of each factor. The procedures of
the root method are displayed as follows.

Normalize each column of matrix F :

fij
wij  n

f
i 1
ij

Multiply column vectors of wij by row and extract the nth root:
Team #26636 Page 6 of 38

n
wi  (  wij )1 / n
i 1

Normalize wi :
wi
wi  n

w i 1
i

Therefore, the approximate feature vector is:

w  ( w1, w 2 ,..., wn)T

Process II: Using grey correlation analysis to rank coaches

1) We determine the factors used for evaluation and collect the data of college coaches.
Assume that the number of coaches on the list is m , and the number of factors used
is n . The standard data array x 0( k ) is the best values of factor k among all coaches.

The correlation data array xi( k ) is a raw data array of factor k among coaches on the

list.

Therefore, for each factor k , the standard data array is:

x 0  { x 0( k )| k  1, 2,..., n }

And the correlation data array is:

xi  { xi( k )| k  1, 2,..., m }

Then we normalize the standard data array and the correlation data array by scaling
between 0 and 1.

Next, we calculate the grey relational coefficient as follows:

min min x0 ( k )  xi ( k )   max max x0 ( k )  xi ( k )
i(k )  i k i k

x0 ( k )  xi ( k )   max max x0 ( k )  xi ( k )
i k

Where  i ( k ) is the grey relational coefficient that measures the relationship between

the standard data array x 0( k ) and the correlation data array xi( k ) , when the k th factor

is the object of study. And  is named as distinguishing coefficient, the values of

which range from 0 to 1. The greater  is, the greater the distinguishing rate will be,

and vice versa. In our model, we choose 0.5 as the value of  . Using  i ( k ) , we can
Team #26636 Page 7 of 38

get the evaluation matrix E  (  i( k ))n  m .

2) Then we calculate the grey relational degree as follows:

n
ri   wi  i ( k )
k 1

Where ri is the grey relational degree of the standard data array to the correlation data
array.

3) Using ri and w , we get the grey comprehensive evaluation model:

R  w  E T  ( r1, r 2 ,...rn )
By sorting the listed coaches by grey relational degree ri in descending order, we get
the rank of all coaches, which enable us to select the top coaches.

3 Application
3.1 Introduction
In this section, we apply our evaluation system to the selection of top 5 all-time col-
lege men’s basketball coaches.

3.2 The selection of factors

We elect two types of factors to represent the accomplishments the coaches have
achieved: The awards they have received as excellent coaches and the performance of
the teams they coach. This thusly leads to the choice of four factors: Wins, Pct.,
NCAAB1 Champions, and Coach Awards. Table 1 shows the definition of each fac-
tor.

Table 1 Factors and Explanations

Factors Explanations
The number of times that one specific
Wins
coach has successfully led a team to win
Pct. The winning percentage
The number of times of winning NCAAB
NCAAB Champions
championships
The number of times of winning the fol-
lowing awards: NABC Coach of the Year,
Coach Awards Basketball Times National Coach of the
year, National Coach of the Year, AP na-
tional coach, and Henry iba awards.
1
NCAAB: National College Athletic Association Basketball
Team #26636 Page 8 of 38

3.3 Data Collection

We obtain the data of college men’s basketball coaches, who meet two criteria as fol-
lows:
 The coaches selected should be the head coaches of NCAA Division I teams be-
cause Division I teams represent the highest level of intercollegiate teams.
 The wins of coaches, which denote the times that the coaches have led their teams
to win, should reach a certain number.

The process of data collection consists of the following steps:

 We obtain a list of college men’s basketball coaches with more than 600 wins on
Wikipedia.2
 Having obtained the name list of coaches, we refer to the NCAA Basketball
Coaching Records3 to select the coaches of Division I teams.
 Next, we searched for the data of the Division I coaches with more than 600 wins
in the NCAA database4 to get the data of the four factors we need (wins, winning
percentage, times of winning NCAAB championships, and times of winning the
title of Coach of the Year).

3.4 Calculation
Process I: The determination of weights
Firstly, we sort the importance of the four factors and create a fuzzy consistent matrix
to determine the weight of each factor. This step can be divided into two parts as fol-
lows:

1) Determine the fuzzy preferential relation matrix

Among the four factors, we assume that the times of winning NCAAB championships
has greater influence on the evaluation of college coaches’ performances than other
factors, while the times of winning the title of Coach of the Year has the least influ-
ence. The winning percentage and wins rank second and third respectively. Table 2
summarizes the interrelationship among the four factors.

2
Data source: List of college men’s basketball coaches with 600 wins,
http://en.wikipedia.org/wiki/List_of_college_men%27s_basketball_coaches_with_600_wins
3
Data source: "NCAA Basketball Coaching Records". NCAA. Retrieved June 14, 2012.
http://web1.ncaa.org/web_files/stats/m_basketball_RB/2011/Coaching.pdf
4
Data source: “NCAA Career Statistics Database”. NCAA. Retrieved June 14, 2012. (The NCAA Career Statis-
tics database allows the viewer to obtain coaching records for all NCAA coaches by inputting the individual’s
name in the linked window.)
http://web1.ncaa.org/stats/StatsSrv/careersearch
Team #26636 Page 9 of 38

Table 2.Interrelationship among the four factors5

Wins Pct. NCAAB champions Coach Awards
Wins 0.5 0 0 1
Pct. 1 0.5 0 1
NCAAB champions 1 1 0.5 1
Coach Awards 0 0 0 0.5

Therefore, the fuzzy preferential relation matrix is:

 0.5 0 0 1 
 
 1 0.5 0 1 
A
 1 1 0.5 1 
 
 0 0 0 0.5 

2) Convert the fuzzy preferential relation matrix to a fuzzy consistent matrix.

Using the calculation we have previously explained. The fuzzy consistent matrix is:
 0.5000
0.3750 0.2500 0.6250 
 
0.6250
0.5000 0.3750 0.7500 
F 
 0.7500
0.6250 0.5000 0.8750 
 
 0.3750
0.2500 0.1250 0.5000 
Therefore, the weights of the four factors are:
w   0.2169 0.2854 0.3527 0.1450 
T

Where 0.2169 is the weight of the times of winning NCAAB championships; 0.2854
is the weight of winning percentage; 0.3527 is the weight of wins; and 0.1450 is the
weight of times of winning the title of Coach of the Year.

Process II: Using grey correlation analysis to rank coaches

1) After preprocessing the data (see Appendix), we get the correlation data array and
standard data array by normalizing them by the scaling between 0 and 1. Table 3 dis-
plays the correlation data array and the standard data array.

Table 3.Correlation data array and standard data array

NCAAB Best
Name Wins Pct Coach award
Champions Coach
Mike Krzyzewski 1 0.725118 0.4 0.714285714 1
Jim Boeheim 0.89939 0.658768 0.1 0.285714286 1
Bob Knight 0.77743 0.464455 0.3 0.142857143 1

5
The metric of measuring the interrelationship among the factor has already be explained when we define the
fuzzy preferential relation matrix.
Team #26636 Page 10 of 38

9
0.70731
Dean Smith 0.781991 0.2 0.571428571 1
7
0.69817
Adolph Rupp 1 0.4 0.714285714 1
1
0.68902
Jim Calhoun 0.436019 0.3 0 1
4
0.55792
Jim Phelan 0.009479 0.1 0.142857143 1
7
0.47865
Eddie Sutton 0.469194 0 0.285714286 1
9
Lefty Driesell 0.42378 0.260664 0.1 0 1
0.40548
Lute Olson 0.592417 0.1 0.142857143 1
8
0.40243
Lou Henson 0.203791 0 0 1
9
0.35670
Henry Iba 0.388626 0.2 0 1
7
0.34146
Edgar Diddle 0.492891 0 0 1
3
0.30182
Phog Allen 0.606635 0.1 0 1
9
Jerry Tarkanian 0.25 0.819905 0.1 0 1
0.24695
Norm Stewart 0.236967 0 0 1
1
0.23475
Ray Meyer 0.2891 0.1 0.428571429 1
6
0.21951
Don Haskins 0.28436 0.1 0 1
2
0.19207
Bob Huggins 0.549763 0 0 1
3
0.19512
Roy Williams 0.85782 0.2 0.428571429 1
2
0.08536
Denny Crum 0.402844 0.2 0.428571429 1
6
0.06402
Gary Williams 0.123223 0.1 0 1
4
0.05182
John Wooden 0.914692 1 1 1
9
0.05182
Rick Pitino 0.587678 0.2 0.285714286 1
9
0.03048
Ralph Miller 0.099526 0 0.285714286 1
8
Team #26636 Page 11 of 38

0.02743
Mike Montgomery 0.345972 0 0.285714286 1
9
Cliff Ellis 0 0 0 0 1

2) Calculate the grey relational coefficient and the grey relational degree.

Table 4. Grey relational coefficient and Grey relational degree matrix

NCAAB Cham- Coach Grey relational
Name Wins Pct.
pions Awards degree
Mike 0.6452 0.63636363
1 0.454545455 0.653648096
Krzyzewski 6 6
0.8324 0.5943 0.41176470
Jim Boeheim 0.357142857 0.535868778
87 66 6
0.6919 0.4828 0.36842105
Bob Knight 0.416666667 0.488272356
83 38 3
0.6307 0.6963 0.53846153
Dean Smith 0.384615385 0.54928851
69 7 8
0.6235 0.63636363
Adolph Rupp 1 0.454545455 0.673244141
74 6
0.6165 0.4699 0.33333333
Jim Calhoun 0.416666667 0.463138417
41 33 3
0.5307 0.3354 0.36842105
Jim Phelan 0.357142857 0.3902421
44 53 3
0.4895 0.4850 0.41176470
Eddie Sutton 0.333333333 0.421891832
52 57 6
0.4645 0.4034 0.33333333
Lefty Driesell 0.357142857 0.39020928
89 42 3
0.4568 0.5509 0.36842105
Lute Olson 0.357142857 0.435701385
25 14 3
0.4555 0.3857 0.33333333
Lou Henson 0.333333333 0.374800311
56 4 3
0.4373 0.4498 0.33333333
Henry Iba 0.384615385 0.407244353
33 93 3
0.4315 0.4964 0.33333333
Edgar Diddle 0.333333333 0.40120218
79 71 3
0.4173 0.5596 0.33333333
Phog Allen 0.357142857 0.424543753
03 82 3
0.7351 0.33333333
Jerry Tarkanian 0.4 0.357142857 0.470881312
92 3
0.3990 0.3958 0.33333333
Norm Stewart 0.333333333 0.365430894
27 72 3
0.3951 0.4129 0.46666666
Ray Meyer 0.357142857 0.397191835
81 16 7
Don Haskins 0.3904 0.4113 0.357142857 0.33333333 0.376378649
Team #26636 Page 12 of 38

76 06 3
0.3822 0.5261 0.33333333
Bob Huggins 0.333333333 0.39899055
84 85 3
0.3831 0.7785 0.46666666
Roy Williams 0.384615385 0.508643536
78 98 7
0.3534 0.4557 0.46666666
Denny Crum 0.384615385 0.410046943
48 24 7
0.3481 0.3631 0.33333333
Gary Williams 0.357142857 0.353469034
95 67 3
0.3452 0.8542
John Wooden 1 1 0.816390818
63 51
0.3452 0.5480 0.41176470
Rick Pitino 0.384615385 0.426661333
63 52 6
0.3402 0.3570 0.41176470
Ralph Miller 0.333333333 0.352966627
49 22 6
Mike Mont- 0.3395 0.4332 0.41176470
0.333333333 0.374573553
gomery 45 65 6
0.3333 0.3333 0.33333333
Cliff Ellis 0.333333333 0.333333333
33 33 3

3) Sort the listed coaches by grey relational degree in descending order to select the
top coaches.

Table 5. Rank of college men’s basketball coaches

Win NCAAB Champi- Coach Grey relational de-
Name Pct
s ons Awards gree
0.80
John Wooden 664 10 7 0.816421266
4
0.82
Adolph Rupp 876 4 5 0.673237909
2
Mike 0.76
975 4 5 0.653625969
Krzyzewski 4
0.77
Dean Smith 879 2 4 0.549279075
6
Jim Boeheim 942 0.75 1 2 0.535847239
0.79
Roy Williams 711 2 3 0.508643731
2
0.70
Bob Knight 902 3 1 0.488258607
9
0.78
Jerry Tarkanian 729 1 0 0.470877508
4
0.70
Jim Calhoun 873 3 0 0.463127488
3
Lute Olson 780 0.73 1 1 0.435696335
Team #26636 Page 13 of 38

6
0.73
Rick Pitino 664 2 2 0.426663053
5
0.73
Phog Allen 746 1 0 0.424539802
9
Eddie Sutton 804 0.71 0 2 0.421885645
0.69
Denny Crum 675 2 3 0.410049682
6
0.69
Henry Iba 764 2 0 0.407240719
3
0.71
Edgar Diddle 759 0 0 0.401197092
5
0.72
Bob Huggins 710 0 0 0.398987605
7
0.67
Ray Meyer 724 1 3 0.397191997
2
0.61
Jim Phelan 830 1 1 0.390234465
3
0.66
Lefty Driesell 786 1 0 0.390203744
6
0.67
Don Haskins 719 1 0 0.37637647
1
0.65
Lou Henson 779 0 0 0.374794534
4
Mike Montgom- 0.68
656 0 2 0.374574409
ery 4
0.66
Norm Stewart 728 0 0 0.365427663
1
0.63
Gary Williams 668 1 0 0.353468964
7
0.63
Ralph Miller 657 0 2 0.352967731
2
0.61
Cliff Ellis 647 0 0 0.333333333
1
Coaches marked in grey are the top 5 basketball coaches of all time.

3.5 Application of our model to other sports

In our model, the four factors we choose can be broadly applied to various kinds of
college sports. When we apply our model to other fields, what we should to is just to
choose the projects which can represent NCAA champions and Coach Awards in the
specific field. So it is convenient for us to apply our model to other fields. The factors
and explanations are shown in Table 6.1 and Table 6.2. The top 5 of both sports are
shown in Table 7.1 and Table 7.2 respectively.
Team #26636 Page 14 of 38

Table 6.1. Factors and Explanations in football

Factors Explanations
The number of times that one specific coach has success-
Wins
fully led a team to win
Pct. The winning percentage
NCAA National Cham-
The amount of Bowl
pionship
Bobby Dodd Coach of the Year Award
Walter Camp Coach of the Year Award
Amos Alonzo Stagg Award
AFCA Coach of the Year
Paul 'Bear' Bryant Award
Coach Awards George Munger Award
Home Depot Coach of the Year Award
Sporting News College Football Coach of the Year
Associated Press Coach of the Year
Joseph V. Paterno Award
National Coach of the Year

Table 6.2. Factors and Explanations in ice-hockey

Factors Explanations
The number of times that one specific coach has successfully
Wins
led a team to win
Pct. The winning percentage
NCCA Champions NCAA National Championship
Spencer Penrose national Coach of the Year(only for division
one)
Coach Awards Lester Patrick Trophy (start from 1966)
Jack Adams Award
NCAA Coach of the Year

Table 7.1.Top 5college men’s football coaches

National cham- Coach Grey relational
Name Wins Pct.
pions Awards degree
Bear Bryant 323 0.78 6 3 0.744053013
Jim Tressel 229 0.74 5 12 0.605457523
Tom Osborne 255 0.84 3 2 0.605088237
Bobby Bowden 377 0.74 2 3 0.592330117
Woody Hayes 238 0.76 5 4 0.588162393

Table 7.2.Top 5 college men’s ice hockey coaches

Name Wins Pct. NCAA champions Coach Grey relation-
Team #26636 Page 15 of 38

Awards al degree
Jerry York 935 0.614 5 2 0.694032022
Vic Heyliger 352 0.685 6 1 0.684648603
Jack Parker 897 0.643 3 3 0.609648054
Ron Mason 924 0.696 0 1 0.598317509
Ned Harkness 380 0.74 3 1 0.590984292

4 Comparison of the ranks from different sources

We compare the top 5 coaches in our lists with those in the reports from ESPN or oth-
er research institutes. We find that the results are mostly identical.

Table 8.Comparison of the rank of top 5 college men’s basketball coaches

ESPN(include wom- Yahoo
RANK The Model Bleacher Report
an) Sports
John
1 John Wooden John wooden John Wooden
Wooden
Mike
2 Adolph Rupp Bob knight Dean Smith Krzyzew
sk
Dean
3 Mike Krzyzewski Mike Krzyzewski Mike Krzyzewski
Smith
Bob
4 Dean Smith Adolph Rupp Pat Summitt
Knight
Jerry Tar-
5 Jim Boeheim Dean Smith John Mclendon Tar-
kanian

Table 9.Comparison of the rank of top 5 college men’s football coaches

RANK The Model Yahoo Sports Statesman Orlando Sentinel
1 Bobby Bowden Bear Bryant Bear Bryant Bear Bryant
2 Bear Bryant Knute Rockne Knute Rockne Nick Saban
3 Glenn Scobey Warner Joe Paterno Glenn Pop Warner Bobby Bowden
4 Amos Alonzo Stagg Tom Osborne Fielding Yost Eddie Robinson
5 Joe Paterno Woody Hayes Joe Paterno Woody Hayes

Table 10. Comparison of the rank of top 5 college men’s ice-hockey coaches

RANK The Model Inside College Hockey

1 Jerry York Bob Johnson
2 Vic Heyliger Herb Brooks
3 Jack Parker Murray Armstrong
4 Ron Mason Jack Parker
Team #26636 Page 16 of 38

5 Ned Harkness Ned Harkness

5 The analysis of time line horizon

5.1 Introduction
The NCAA Championships have a history of approximately seventy years. During
such a long span of time, the competition system has experienced a series of reforms.
In early years, the number of sports events was much less than it is today, thus the
competition among sports teams might be less intense. Therefore, there might be a
difference among the average winning percentages in different time line horizon. If so,
we can no longer justify our previous model, which ignores the influence of time line
horizon and equally treats coaches from different periods in terms of winning per-
centage.

Therefore, we intend to explore whether there is a connection, to some degree, be-

tween the factors we choose and the time line horizon. Note that it is difficult to ac-
quire the data of the times of winning coach of the year within limited amount of time.
So we decide to use the rest of factors (Wins, Pct., and times of winning national
championships) to examine if those factors change during different time line horizons.
If the time line horizon does have significant influence on the rest of our factors, we
need to modify our model. Now we process the data of college men’s basketball
coaches to see if there is a need for us to modify our model.

5.2 Procedures
Step 1.Firstly, we obtain the data of college men’s basketball coaches who have
coached their teams in more than 400 games. We select the coaches from all three di-
visions because the reforms of competition system cover all the divisions. The raw
data are attached in appendix.

Step 2.Secondly, we sort the coaches by the year they start their coaching career and
divide them into five groups as follows:
 Coaches who started their career before 1930s (group 1).
 Coaches who started their career during 1930s and 1940s (group 2).
 Coaches who started their career during 1950s and 1960s (group 3).
 Coaches who started their career during 1970s and 1980s (group 4).
 Coaches who started their career during 1990s and 2000s (group 5).

Then we calculate the mean and variance of each factor in each group. Figure 1 shows
the means of winning percentages in all groups. Figure 2 shows the means of wins in
all groups. Figure 3 shows the mean values of the times of winning NCAAB champi-
onships in all groups.
Team #26636 Page 17 of 38

Figure 1.The mean values of winning percentages

Figure 2.The mean values of wins

Team #26636 Page 18 of 38

Figure 3.The mean values of NCAAB Champions6

Step 3. We find that the line in Figure 1 smoothly goes down, which means that it is
easier for good college basketball coaches in early years to get a high winning
percentage. This is reasonable because the competition in not so intense in previous
years as it is today. Therefore, with the same weights, using winning percentage to
evaluate performance is not fair for younger coaches. Thus we should devise a
correction factor to modify it.

Although this trend is clearly shown in Figure 1. It is still necessary for us to use
one-way analysis of variance to test whether there is a significant difference among
the variances of different groups. If the difference among variances is acceptable, we
need to make adjustment to our model according to the tendency.

Table 11.
Source of
SS Df MS F P-value F crit
Difference
Among the 1.988653
0.026814016 2 0.013407 0.138645 3.025253
five groups 5
Within each
2.062976049 306 0.006742
group
Sum 2.08979006 308

Table 12.
Source of
SS Df MS F P-value F crit
Difference
Among the 3.3817 0.009770 2.3950026
0.114121057 4 0.02853026
five groups 495 28 16
Within each
3.264940845 387 0.00843654
group
sum 3.3790619 391

Table 11 demonstrates that the P-value is greater than 0.05, which means there is not a
significant difference of variances among group 3, group 4 and group 5. Table 12
shows that there is a significant difference of variances among all five groups. So
there exists some extreme values (the data of some outstanding coaches) that raises
the variances of group 1 and group 2.

Therefore, only the increasing tendency of group 3, group 4, and group 5’s winning
percentages can be used to set a correction factor. Under two circumstances as follows,
we should adopt different approaches to set correction factors.

6
The sharp change of the line in 1930s-1940s is because that John wooden had won about 10 champions, if we
exclude him from the sample, the line will become flat, so the time line horizon makes no difference on the
NCAAB champion.
Team #26636 Page 19 of 38

 The selection of the best all-time coaches who start coaching after 1950s
 The selection of the best all-time coaches who start coaching after 1900s

Step 4. Now we set a correction factor to reduce the winning percentages of coaches
in previous years and increase the winning percentages of present coaches. We adopt
the weighted moving average method to calculate the correction factor.

1) To select the best all-time coaches who started their coaching careers after 1950s,
we need to modify our model. The correction factor  i is:
1
i 
Si
n

Y Z i i

Where Si  i 1
n
, Zi is the number of samples within group i , Yi is the mean val-
Z
i 1
i

ue of the winning percentage of group i .

Table 13 shows the winning percentages and grey relational degree of top 10 coaches
before adjustment, while Table 14 shows the winning percentages and grey relational
degree of top 10 coaches after adjustment.

Table 13.The rank of Top 10 college men’s basketball coaches before adjustment
Name Pct. Grey relational degree
John Wooden 0.804 0.816421
Adolph Rupp 0.822 0.673238
Mike Krzyzewski 0.764 0.653626
Dean Smith 0.776 0.549279
Jim Boeheim 0.75 0.535847
Roy Williams 0.792 0.508644
Bob Knight 0.709 0.488259
Jerry Tarkanian 0.784 0.470878
Jim Calhoun 0.703 0.463127
Lute Olson 0.736 0.435696

Table 147.The rank of Top 10 college men’s basketball coaches after adjustment
Name Correction factor Pct. Grey relational degree
Mike Krzyzewski 1.733399695 1.324317367 0.9204074
Roy Williams 1.733399695 1.372852558 0.586667007
Dean Smith 1.695298373 1.315551537 0.574608911
Jim Boeheim 1.733399695 1.300049771 0.53458653

7
Because we select the best all time coaches who start coaching after 1950s, John Wooden and Adolph Rupp who
start their coaching careers in 1930s should not have been on the list.
Team #26636 Page 20 of 38

Bob Knight 1.695298373 1.201966546 0.525580488

Jim Calhoun 1.733399695 1.218579985 0.508524212
Jerry Tarkanian 1.733399695 1.35898536 0.496549075
Rick Pitino 1.733399695 1.274048775 0.438797739
Lute Olson 1.733399695 1.275782175 0.419173266
Denny Crum 1.733399695 1.206446187 0.417960291

2) For coaches who started their coaching careers between 1900s and 1950s, we find
out that group 1 and group 2 contains extreme values. The two outstanding coaches,
those data are probably the extreme values, are John Wooden and Adolph Rupp. After
eliminating their data, we repeat our one-way data analysis of variance and find that
the P-value among five groups increases to 0.08, which is greater than 0.05. Thus we
can know for sure that the data of John Wooden and Adolph Rupp are extreme values.

Although John Wooden and Adolph Rupp are among the greatest coaches in NCAAB
in early years, we cannot assume that they would be as successful today. Therefore,
when we select the best all-time coaches who started their coaching careers after
1900s, we should set a correction factor to reduce their grey relational degrees.
However, because of the scarcity of data, we cannot use a precise formulation for
calculation.

Although we only explain the principle for modifying our model based on the data of
college men’s basketball coaches in this section, this approach is applicable to other
sports (such as football and ice-hockey) because of the generality of the factors we
select. We also find that the winning percentage we use to select the best all-time
coaches in NCAA football has the same tendency as it is when we analyze the
relationship between winning percentage and the time line horizon. Using the
one-way analysis of variance, we find that its P-value is 0.07 (greater than 0.05),
which means that we can still use the moving weighted average method to get our
correction factor in the evaluation of best all-time college men’s football coaches.

6 The analysis of gender

By analyzing the data of college women’s basketball coaches, we find that the four
factors we select have no relation to gender and that the weight of each indicator
doesn’t change. Table 15 displays the top 5 of college women’s basketball coaches.

Table 15 The top 5 of college women’s basketball coaches

Grey
NCAA Coach
Name Wins Pct. relational
Champions Awards
degree
Pat Summitt 1098 0.841 8 8 0.989807143
C. Vivian
916 0.732 0 0 0.60757619
Stringer
Team #26636 Page 21 of 38

Tara
910 0.817 2 4 0.54686006
VanDerveer
Sylvia
908 0.739 1 6 0.481908161
Hatchell
Barbara Ste-
900 0.785 0 0 0.476794451
vens

7 Stability test
7.1 Testing the influence of changing the method on the de-
termination of weights
7.1.1 Introduction
In this section, we use Shannon’s entropy method to determine the weight of each
factor. Then we use grey correlation analysis, which has identical processes as above,
to obtain the rank of coaches to see if there is any difference between the two methods
(the fuzzy consistent matrix method and the entropy method) for determining weights.

7.1.2 Notation
Notation Explanation
C Raw data matrix

C' Modified raw data matrix

Q Decision matrix
hi Entropy
W Weights of factors under the entropy
method

7.1.3 Using Shannon’s entropy method to determine the weight of

each factor
Shannon’s entropy method is a well-known method for determining the weights for
multiple attribute decision making (MADM) process (Lotfi., et, al.,2010). Entropy is a
metric for measuring the degree of disorder in a system. For a specific factor in the
system, when there is a large difference among its values, the entropy value is com-
paratively large, while a factor with less discrepancy among its values has smaller en-
tropy value.

Using this approach, we can determine the weight of each factor in our coach evalua-
tion system. The factor with larger entropy is less importance in our system, and thus
Team #26636 Page 22 of 38

has smaller weight. The factor with smaller entropy is more important in our system,
and thus has larger weight.

Therefore, the procedures for determining weights can be expressed as follows:

1) If the i th factor has some extreme values, we should modify the raw data ma-
trix C  ( cij )n  m to create a new matrix C '  ( cij )n  m and thusly eliminate the negative
influence of extreme values on our model.

The formula for creating C ' is:

cij  min( cij )

cij'  u  v
max( cij )  min( cij )

m m m
Where u   cij /  ( cij  cij )2 ,v  1 / ( c ij  cij )2 .
i 1 i 1 i 1

2) Normalize the raw data matrix C '  ( cij )n  m to create matrix Q  ( qij )n  m .
The factors of this model are positive factors, which mean that the factors with greater
values are better. Therefore, the formula for creating Q is:

cij  min{ cij }

qij  1
j

max{ cij }  min{ cij }

j j

By normalizing the modified raw data matrix, we can eliminate the anomalies resulted
from different dimensions and metrics for measuring the data.

3) Based on the decision matrix Q , we calculate the entropy of each factor:

m
hi  k  dij ln dij
j 1

m
Where hi is the i th factor’s entropy. dij  qij /  qij . k=1/ ln m . ln dij is defined as 0
i 1

if dij  0 .

4) Now we can calculate the weight of each factor. Suppose that the number of objects
is m and the number of factors is n .

By using the following formula, we can calculate the weight of each factor in a sys-
Team #26636 Page 23 of 38

tem.
1  hi
Wi  n
n   hi
i=1

n
Where 0  Wi  1, Wi  1 .
i 1

In our model, we calculate the weights of the four factors as follows:

W   0.2493 0.2492 0.2508 0.2507 
T

Where 0.2493 is the weight of the times of winning NCAAB championships; 0.2492
is the weight of winning percentage; 0.2508 is the weight of wins; and 0.2507 is the
weight of times of winning the title of Coach of the Year.

Table 16 shows the rank of college coaches based on Shannon’s entropy method and
grey correlation analysis.

Table 16.Results based on Entropy Method

NCAAB Coach Grey relational
NAME WINS Pct.
Champions Awards degree
John Wooden 664 0.804 10 7 0.80048
Mike Krzyzewski 975 0.764 4 5 0.68361
Adolph Rupp 876 0.822 4 5 0.678202
Dean Smith 879 0.776 2 4 0.562236
Jim Boeheim 942 0.75 1 2 0.548441
Roy Williams 711 0.792 2 3 0.503022
Bob Knight 902 0.709 3 1 0.489688
Jim Calhoun 873 0.703 3 0 0.458873
Jerry Tarkanian 729 0.784 1 0 0.456083
Lute Olson 780 0.736 1 1 0.433112
Eddie Sutton 804 0.71 0 2 0.429747
Rick Pitino 664 0.735 2 2 0.422348
Phog Allen 746 0.739 1 0 0.416651
Denny Crum 675 0.696 2 3 0.415138
Ray Meyer 724 0.672 1 3 0.407979
Henry Iba 764 0.693 2 0 0.401171
Edgar Diddle 759 0.715 0 0 0.398482
Jim Phelan 830 0.613 1 1 0.397834
Bob Huggins 710 0.727 0 0 0.393601
Lefty Driesell 786 0.666 1 0 0.389495
Mike Montgomery 656 0.684 0 2 0.379449
Team #26636 Page 24 of 38

Lou Henson 779 0.654 0 0 0.376859

Don Haskins 719 0.671 1 0 0.372983
Norm Stewart 728 0.661 0 0 0.365295
Ralph Miller 657 0.632 0 2 0.360621
Gary Williams 668 0.637 1 0 0.350446
Cliff Ellis 647 0.611 0 0 0.333333
Coaches marked in grey are the top 5 basketball coaches of all time.

7.1.4 Comparison of the outcomes

In order to test the stability of our model, we now compare the outcome from using
the fuzzy consistent matrix to determine weights with the outcome from using the en-
tropy method.

Table 17.Comparison between Fuzzy Consistent Matrix Method and Entropy Method
RANK Fuzzy Consistent Matrix Entropy Method
1 John Wooden John Wooden
2 Adolph Rupp Mike Krzyzewski
3 Mike Krzyzewski Adolph Rupp
4 Dean Smith Dean Smith
5 Jim Boeheim Jim Boeheim
6 Roy Williams Roy Williams
7 Bob Knight Bob Knight
8 Jerry Tarkanian Jim Calhoun
9 Jim Calhoun Jerry Tarkanian
10 Lute Olson Lute Olson

Table 17 shows that the name lists of top 10 coaches generated from the above two
methods are identical, while there is a slight difference among the ranks. Therefore,
we can reach the conclusion that our model is stable.

7.2 Testing the influence of changing the comparative im-

portance among the four factors
Now we change the comparative importance among the four factors in the evaluation
of college men’s basketball coaches to see if there is a significant difference between
the outcomes.

Table 18.The comparative importance among factors after adjustment

Wins Pct. NCCAB Champions Coach Awards
Wins 0.5 0 0 0
Pct. 1 0.5 0 0
Team #26636 Page 25 of 38

NCCAB Champions 1 1 0.5 1

Coach Awards 1 1 0 0.5

Table 19 displays the rank of college men’s basketball coaches after adjustment. Table
20 compares the outcome in this section with the outcome in the previous section.

Table 19.Rank of college men’s basketball coaches after adjustment

NCAAB Coach Grey rela-
NAME WINS Pct.
Champions Awards tional degree
John Wooden 664 0.804 10 7 0.873444278
Adolph Rupp 876 0.822 4 5 0.649232275
Mike Krzyzewski 975 0.764 4 5 0.626894346
Dean Smith 879 0.776 2 4 0.531824804
Roy Williams 711 0.792 2 3 0.493260791
Jim Boeheim 942 0.75 1 2 0.493108718
Bob Knight 902 0.709 3 1 0.45717316
Jerry Tarkanian 729 0.784 1 0 0.438544324
Jim Calhoun 873 0.703 3 0 0.433420385
Rick Pitino 664 0.735 2 2 0.422098796
Denny Crum 675 0.696 2 3 0.4189328
Lute Olson 780 0.736 1 1 0.416837441
Eddie Sutton 804 0.71 0 2 0.411274136
Ray Meyer 724 0.672 1 3 0.406011386
Phog Allen 746 0.739 1 0 0.402993475
Henry Iba 764 0.693 2 0 0.391780608
Edgar Diddle 759 0.715 0 0 0.382957853
Bob Huggins 710 0.727 0 0 0.382252792
Jim Phelan 830 0.613 1 1 0.380833646
Mike Montgomery 656 0.684 0 2 0.378288688
Lefty Driesell 786 0.666 1 0 0.375969596
Don Haskins 719 0.671 1 0 0.366927201
Lou Henson 779 0.654 0 0 0.362422603
Ralph Miller 657 0.632 0 2 0.361857247
Norm Stewart 728 0.661 0 0 0.356422014
Gary Williams 668 0.637 1 0 0.350356494
Cliff Ellis 647 0.611 0 0 0.333333333

Table 20.Comparison of outcomes based on two different Fuzzy Consistent Matrixes

RANK Fuzzy Consistent Matrix1 Fuzzy Consistent Matrix2
1 John Wooden John Wooden
2 Adolph Rupp Adolph Rupp
3 Mike Krzyzewski Mike Krzyzewski
4 Dean Smith Dean Smith
Team #26636 Page 26 of 38

5 Jim Boeheim Roy Williams

6 Roy Williams Jim Boeheim
7 Bob Knight Bob Knight
8 Jerry Tarkanian Jerry Tarkanian
9 Jim Calhoun Jim Calhoun
10 Lute Olson Rick Pitino

We can see from Table 20 that the outcome is slightly different when compared with
the outcome in the previous section. Based on section 6.1 and 6.2, we can reach the
conclusion that our model is stable.

8 Sensitivity test
In previous section, we assume that the distinguishing coefficient   0.5 . This as-

sumption of this value is based on experience. Therefore, we need to change the value
of  to see if the value of  will exert significant influence on our model.

Table 21 shows the rank of coaches when   0.5 , while Table 22 and Table 23show

the rank of coaches when  is equal to 0.4 and 0.6 respectively.

Table 21. Rank of college men’s basketball coaches (  =0.5)

RANK NAME Grey Relational Degree (  =0.5)

1 Mike Krzyzewski 0.816390818

2 Jim Boeheim 0.673244141
3 Bob Knight 0.653648096
4 Dean Smith 0.54928851
5 Adolph Rupp 0.535868778
6 Jim Calhoun 0.508643536
7 Jim Phelan 0.488272356
8 Eddie Sutton 0.470881312
9 Lefty Driesell 0.463138417
10 Lute Olson 0.435701385

Table 22. Rank of college men’s basketball coaches (  =0.4)

RANK NAME Grey Relational Degree (  =0.4)

1 Mike Krzyzewski 0.816421

Team #26636 Page 27 of 38

2 Jim Boeheim 0.673238

3 Bob Knight 0.653626
4 Dean Smith 0.549279
5 Adolph Rupp 0.535847
6 Jim Calhoun 0.508644
7 Jim Phelan 0.488259
8 Eddie Sutton 0.470878
9 Lefty Driesell 0.463127
10 Lute Olson 0.435696

Table 23.Rank of college men’s basketball coaches (  =0.6)

RANK NAME Grey Relational Degree (  =0.6)公式

1 Mike Krzyzewski 0.831661964

2 Jim Boeheim 0.70427652
3 Bob Knight 0.687185051
4 Dean Smith 0.590853876
5 Adolph Rupp 0.574939217
6 Jim Calhoun 0.54878576
7 Jim Phelan 0.531490084
8 Eddie Sutton 0.511362919
9 Lefty Driesell 0.507122379
10 Lute Olson 0.479680225

By comparing the above three tables, we find that the top 10 college men’s basketball
coaches are identical, which means that our model is insensitive to the change of  .

9 Strengths and Weaknesses

9.1 Strengths
 It is convenient and broadly applicable to use our model to select the top col-
lege coaches. Utilizing the grey correlation analysis method, our model has high
precision and is convenient for researches to assess the performances of coaches.
The four factors we select can be universally adopted when analyzing different
sports in NCAA or coaches of different genders.
 Our model has good stability. Our outcomes are nearly identical when different
method of analyzing stability and sensitivity is adopted.
Team #26636 Page 28 of 38

9.2 Weaknesses
 The standard for selecting factors is rigorous to some degree. In terms of se-
lection of best all-time college men’s football coaches, we choose Division I
coaches with more than 200 wins. However, some famous coaches, although fail
to achieve 200 wins, should have been granted the qualification to be selected as
best all-time coaches.
 The collection of the numbers of Coach Awards is subjective. Some awards
should not have been included.
 The analysis of time line horizon is imperfect. For example, it is difficult for us
to set a correction factor for coaches who started their careers before 1950s.

Reference
[1] Rushall, B. S., and K. Wiznuk. "Athletes' assessment of the coach--the coach
evaluation questionnaire." Canadian journal of applied sport sciences. Journal cana-
dien des sciences appliquées au sport.10.3 (1985): 157.

[2] Cassidy, Tania, Paul Potrac, and Alex McKenzie. "Evaluating and Reflecting Upon
a Coach Education Initiative: The CoDe^ 1 of Rugby." Sport Psychologist20.2 (2006):
145.

[3] Nelson, Lee J., and Christopher J. Cushion. "Reflection in coach education: the
case of the national governing body coaching certificate." Sport Psychologist20.2
(2006): 174.

[4] Lotfi F H, Fallahnejad R. Imprecise Shannon’s entropy and multi attribute deci-
sion making[J]. Entropy, 2010, 12(1): 53-62.

[5] Saaty, T.L. The Analytic Hierarchy Process; McGraw-Hill: NewYork, NY, USA,
1980.

[6] Hwang, C.L.; Lin, M.J. Group Decision Making under Multiple Criteria: Methods
and Applications; Springer: Berlin, Germany, 1987.

[7] Fan, Z.P. Complicated multiple attribute decision making: Theory and applica-
tions. Ph.D Dissertation; Northeastern University, Shenyang, China, 1996.

[8] Chu, A.T.W.; Kalaba, R.E.; Spingarn, K. A comparison of two methods for de-
termining the weights of belonging to fuzzy sets. J. Optimiz. Theor. App. 1979, 27,
531–538.

[9] Shannon, C.E. A mathematical theory of communication. Bell Syst. Tech. J. 1948,
27, 379–423.
Team #26636 Page 29 of 38

[10]Sun, L. J.; Xing, X. J.; Zhou, D. Q. Improvement of Entropy Method. Statistics and Deci-
sion. J. 2010,21.

[11]Lu, T. C.; Kang, K. The Application of Entropy method and AHP in determining the
weights . Computer Programming Skills & Maintenance. J. 2009, 22.

[12]Chen, X. The Application of fuzzy analytic hierarchy process in the preferred aspects of
the program. Computer Engineering and Design. J. 2004, 10.

Appendix

Men’s basketball college coaches

a NCCAB Champions
b Coach Awards
c NABC Coach of the Year
d Basketball Times National Coach of the Year
e National coach of the year
f AP national coach
g Henry iba awards
h Coach Year

Name Year Wins Pct. a b c d e f g h

Mike
Krzyzewski 38 975 0.764 4 5 1 2 0 2 0 1976
Jim Boeheim 37 942 0.75 1 2 1 0 0 1 0 1977
Bob Knight 42 902 0.709 3 1 0 0 0 1 0 1966
Dean Smith 36 879 0.776 2 4 0 0 4 0 0 1962
Adolph Rupp 41 876 0.822 4 5 0 0 5 0 0 1931
Jim Calhoun 39 873 0.703 3 0 0 0 0 0 0 1973
Jim Phelan 49 830 0.613 1 1 0 0 1 0 0 1989
Eddie Sutton 37 804 0.71 0 2 0 0 0 2 0 1970
Lefty Driesell 41 786 0.666 1 0 0 0 0 0 0 1961
Lute Olson 34 780 0.736 1 1 0 0 1 0 0 1974
Lou Henson 41 779 0.654 0 0 0 0 0 0 0 1963
Henry Iba 41 764 0.693 2 0 0 0 0 0 0 1930
Edgar Diddle 42 759 0.715 0 0 0 0 0 0 0 1959
Phog Allen 48 746 0.739 1 0 0 0 0 0 0 1906
Jerry Tarkanian 31 729 0.784 1 0 0 0 0 0 0 1970
Norm Stewart 38 728 0.661 0 0 0 0 0 0 0 1968
Ray Meyer 42 724 0.672 1 3 1 0 0 2 0 1943
Team #26636 Page 30 of 38

Don Haskins 38 719 0.671 1 0 0 0 0 0 0 1962

Bob Huggins 30 710 0.727 0 0 0 0 0 0 0 1985
Roy Williams 24 711 0.792 2 3 1 0 0 1 1 1989
Denny Crum 30 675 0.696 2 3 0 0 3 0 0 1972
Gary Williams 33 668 0.637 1 0 0 1 0 0 0 1979
John Wooden 29 664 0.804 10 7 0 0 7 0 0 1947
Rick Pitino 27 664 0.735 2 2 1 0 1 0 0 1979
Ralph Miller 38 657 0.632 0 2 0 0 2 0 0 1952
Mike Mont-
gomery 31 656 0.684 0 2 1 0 1 0 0 1979
Cliff Ellis 37 647 0.611 0 0 0 0 0 0 0 1976

Men’s football college coach

a NCCA Champions
b Coach Awards
c Bobby Dodd Coach of the Year Award
d Walter Camp Coach of the Year Award
e Amos Alonzo Stagg Award
f AFCA Coach of the Year
g Paul 'Bear' Bryant Award
h George Munger Award
i Home Depot Coach of the Year Award
j Sporting News College Football Coach of the Year
k Associated Press Coach of the Year
l Joseph V. Paterno Award
m National Coach of the Year

Name Years Wins Pct. a b c d e f g h i j k l m

Bowden,
Bobby 44 377 0.743 22 3 1 1 1 0 0 0 0 0 0 0 0
Bryant, Bear 38 323 0.78 15 3 0 0 0 3 0 0 0 0 0 0 0
Warner,
Glenn Scobey 49 319 0.733 4 1 0 0 1 0 0 0 0 0 0 0 0
Stagg, Amos
Alonzo 57 314 0.605 2 1 0 0 0 1 0 0 0 0 0 0 0
Paterno, Joe 46 298 0.685 24 17 2 3 1 5 1 3 1 1 0 0 0
Beamer,
Frank 33 266 0.666 9 7 1 1 0 1 1 1 0 0 1 1 0
Edwards,
LaVell 29 257 0.716 7 3 1 0 1 1 0 0 0 0 0 0 0
Osborne, Tom 25 255 0.836 12 2 1 0 0 0 0 0 0 0 0 0 1
Team #26636 Page 31 of 38

Holtz, Lou 33 249 0.651 12 5 0 1 0 0 2 0 0 2 0 0 0

Brown, Mack 30 244 0.666 13 2 1 1 0 0 0 0 0 0 0 0 0
Moore, Jerry 31 243 0.642 3 0 0 0 0 0 0 0 0 0 0 0 0
Hayes,
Woody 33 238 0.759 5 4 0 1 1 1 0 0 0 1 0 0 0
Schembechler,
Bo 27 234 0.804 5 4 1 1 1 1 0 0 0 0 0 0 0
Ault, Chris 28 233 0.681 2 0 0 0 0 0 0 0 0 0 0 0 0
Fry, Hayden 37 232 0.564 7 2 0 0 1 0 0 0 0 1 0 0 0
Tressel, Jim 25 229 0.742 6 12 12 0 0 0 0 0 0 0 0 0 0
Laycock,
Jimmye 34 222 0.573 1 0 0 0 0 0 0 0 0 0 0 0 0
Spurrier, Ste-
ve 24 219 0.733 9 0 0 0 0 0 0 0 0 0 0 0 0
Nehlen, Don 30 202 0.609 4 3 1 1 0 1 0 0 0 0 0 0 0
Dooley, Vince 25 201 0.715 8 5 1 1 1 1 0 0 0 1 0 0 0

Men’s ice-hockey college coach

a NCAA Champions
b Coach Awards
c Lester Patrick Trophy(1966 起评)
d Spencer Penrose national Coach of the Year(only for division one)
e Jack Adams Award
f NCAA Coach of the Year

Name Years Wins Pct. a b c d e f

Jerry York 40 935 0.61 5 2 1 1 0 0
Jack Parker 40 897 0.64 3 3 0 3 0 0
Red Berenson 30 770 0.66 2 2 0 1 1 0
Don Lucia 26 623 0.64 2 1 0 1 0 0
John "Snooks" Kelley 36 501 0.67 1 3 1 2 0 0
Ron Mason 36 924 0.7 0 1 0 1 0 0
Rick Comley 37 783 0.56 0 2 0 2 0 0
R.H. "Bob" Peters 36 744 0.7 2 0 0 0 0 0
Len Ceglarski 35 689 0.66 0 3 0 3 0 0
Jeff Sauer 31 655 0.55 2 0 0 0 0 0
John MacInnes 26 555 0.65 3 6 1 2 0 3
Don Brose 34 540 0.59 0 0 0 0 0 0
Dick Umile 23 519 0.64 0 0 0 0 0 0
John P. Riley, Jr. 36 542 0.61 0 0 0 0 0 0
Bruce Marshall 25 351 0.48 0 0 0 0 0 0
Team #26636 Page 32 of 38

Murray Armstrong 21 460 0.67 0 2 1 1 0 0

Bill Wilkinson 26 437 0.48 0 0 0 0 0 0
Shawn Walsh 17 423 0.68 0 1 0 1 0 0
John Gasparini 17 407 0.6 0 0 0 0 0 0
Doug Woog 14 390 0.66 0 0 0 0 0 0
Amo Bessone 31 387 0.46 1 1 0 1 0 0
Ned Harkness 23 380 0.74 3 1 0 1 0 0
Jeff Jackson 14 367 0.68 2 1 0 1 0 0
Vic Heyliger 23 352 0.69 6 1 0 1 0 0
Bob Daniels 21 350 0.48 0 2 0 2 0 0
Mike McShane 32 633 0.65 0 0 0 0 0 0
George Gwozdecky 27 592 0.6 2 2 0 2 0 0
Joe Marsh 26 482 0.53 0 1 0 2 0 0
Doug Ross 27 457 0.6 0 0 0 0 0 0
Rick Gotkin 25 453 0.58 0 0 0 0 0 0
Mike Gilligan 25 407 0.54 0 0 0 0 0 0
Craig Dahl 24 407 0.5 0 0 0 0 0 0
Mike Sertich 21 375 0.49 0 1 0 1 0 0
Rand Pecknold 19 367 0.6 0 0 0 0 0 0
Bruce Marshall 25 351 0.48 0 0 0 0 0 0

Woman’s basketball college coach

a NCAA Champions
b Coach Awards
c NCAA Coach of the Year
d Naismith Coach of the 20th Century
e NAIA

Name Year Wins Pct. a b c d e

Pat Summitt 38 1098 0.841 8 8 7 1 0
C.Vivian Stringer 42 916 0.732 0 0 0 0 0
Tara VanDerveer 34 910 0.817 2 4 4 0 0
Sylvia Hatchell 38 908 0.739 1 3 2 0 1
Barbara Stevens 36 900 0.785 0 0 0 0 0
Jody Conradt 38 900 0.744 2 6 6 0 0
Debbie Ryan 34 739 0.695 1 1 0 1 0
Kay Yow 38 737 0.682 0 0 0 0 0
Sue Gunter 34 708 0.697 0 0 0 0 0
Muffet McGraw 31 714 0.735 1 1 0 1 0
Amy Ruley 29 671 0.772 5 0 0 0 0
Theresa Grentz 33 671 0.683 0 0 0 0 0
Team #26636 Page 33 of 38

Nancy Fahey 27 639 0.844 5 0 0 0 0

Program

1)Apply Fuzzy Consistent Matrix to determine weight

A=[0.5 0 0 1
1 0.5 0 1
1 1 0.5 1
0 0 0 0.5]

c=sum(A,2);
b=sum(A);
m=c(1,:)+b(:,1);
for i=1:4
for j=1:4
r(i,j)=(c(i,:)-c(j,:))/(2*m)+0.5;
end
end

d=sum(r);
for j=1:4
for i=1:4
s(i,j)=r(i,j)/d(:,j);
end
end

k=power(prod(s,2),1/4);

for i=1:4
w(i,:)=k(i,:)/sum(k);
end
w=w’

2)Apply Entropy Method to determine weight

x=xlsread('E:\2014 美赛\B 题\数据\basketball.xls');

x=x';
k=mean(x,1)
t=sum(x)
s3=0;
s4=0;
Team #26636 Page 34 of 38

for i=1:27
a3=(x(i,3)-k(3))^2;
s3=s3+a3;
end
d3=1/(sqrt(s3))
c3=t(:,3)/(sqrt(s3))

for i=1:27
a4=(x(i,4)-k(4))^2;
s4=s4+a4;
end
d4=1/(sqrt(s4))
c4=t(:,4)/(sqrt(s4))
y=[]
[m,n]=size(x);
for i=1:n
y(:,i)=x(:,i)/sum([x(:,i)]);
end

for i=1:27
y(i,3)=y(i,3)*d3+c3;
y(i,4)=y(i,4)*d4+c4;
end

for i=1:n
y(:,i)=x(:,i)/sum([x(:,i)])+1;
end

for l=1:n
s(1,l)=0;
for j=1:m
p(1,l)=y(j,l)*log(y(j,l))
s(1,l)=s(1,l)+p(1,l)
end
end
k=(log(m))^(-1)
e=-k*s
h=ones(1,n)-e
w=h/(n-sum(h))
w=w/sum(w,2)

3)Total model
Weight and calculate grey relational degree
Team #26636 Page 35 of 38

A=[0.5 0 0 1
1 0.5 0 1
1 1 0.5 1
0 0 0 0.5]

c=sum(A,2);
b=sum(A);
m=c(1,:)+b(:,1);
for i=1:4
for j=1:4
r(i,j)=(c(i,:)-c(j,:))/(2*m)+0.5;
end
end

d=sum(r);
for j=1:4
for i=1:4
s(i,j)=r(i,j)/d(:,j);
end
end

k=power(prod(s,2),1/4);
for i=1:4
w(i,:)=k(i,:)/sum(k);
end
w=w';

a=xlsread('E:\2014 美赛\B 题\数据\basketball.xls');

for i=1:4
a(i,:)=(a(i,:)-min(a(i,:)))/(max(a(i,:))-min(a(i,:)));
end
[m,n]=size(a);
cankao=max(a')'
t=repmat(cankao,[1,n])-a;
mmin=min(min(t));
mmax=max(max(t));
rho=0.5;
xishu=(mmin+rho*mmax)./(t+rho*mmax);
correlation=w*xishu

xlswrite('C:\Users\Administrator\Desktop\correlation.xls',correlation)
Team #26636 Page 36 of 38

An article for Sports Illustrated

The best all-time college coaches

College coaches are critical to the performance of their teams. As the heat of National
College Athletic Association Championships spreads throughout the country, the
coaches that have led their teams to win a considerable number of trophies become
the legends among sports fans. Who is your favorite college coach?

May be he is John Wooden, the legend on basketball court, who won the most amount
of NCAA champions?

May be he is “Bear” Bryant, the hero in the field of football ,who held the record for
most wins as head coach when he retired as a coach in 1982?

May be she is Pat Summitt, the most successful woman basketball coach, who hold
the most all-time wins in NCAA basketball history above 1000 victories?

We know that you are tired of the endless debate on your choice of the best all-time
coaches. Thus we build a quantitative evaluation system to rank the college coaches
of all time. The results are just in! Check whether your favorite college coach is
among the top!

Top 5 College Men’s Basketball Coaches

1. John Wooden
2. Adolph Rupp
3. Mike Krzyzewski
4. Dean Smith
5. Jim Boeheim
Team #26636 Page 37 of 38

Top 5 College Men’s Football Coaches

1. Bear Bryant
2. Jim Tressel
3. Tom Osborne
4. Bobby Bowden
5. Woody Hayes

Top 5 College Men’s Ice Hockey Coaches

1. Jerry York
2. Vic Heyliger
3. Jack Parker
4. Ron Mason
5. Ned Harkness

In our evaluation system, we select four indicators to assess the performance of

coaches: The times that they lead their teams to win NCAA championships, the win-
ning percentage, the times of victory throughout their coaching career, and the times
they win the title of Coach of the Year. We use appropriate method to determine the
weight of each factor, and then use mathematical model to rank all the coaches.

One of the advantages of this model is that it eliminates the arbitrariness of human
judgment. We know that most of media use online-voting system to collect data and
rank the coaches, which is not convincible. Instead of that, we obtain the data of all
college coaches from NCAA database, and use mathematical model to determine
which indicator is more important.

Another advantage is that our model is universally applicable. We can apply this
model across both genders and various kinds of sports.

The function of our model might be limited, but your enthusiasm of sports is unlim-
ited!