Tuesday, April 29, 2008

Coach Rankings Part 1

Although I've discussed this at length before, let's stick to the numbers. The perception of the top coaches is mostly based on NCAA tournament appearances and wins. Here are the coaches who have the most appearances plus wins in the last 10 years (NCAA tournaments 1999-2008).

NCAA Appearances – Wins, Coach - School
10 – 28, Roy Williams - North Carolina
10 – 26, Mike Krzyzewski - Duke
10 – 24, Tom Izzo - Michigan State
10 – 22, Bill Self - Kansas
10 – 16, Rick Barnes - Texas
9 – 22, Billy Donovan - Florida
9 – 17, Tubby Smith - Minnesota
9 – 15, Lute Olson - Arizona
9 – 9, Mark Few - Gonzaga
8 – 22, Jim Calhoun - Connecticut
8 – 9, Bob Huggins - West Virginia
7 – 17, Gary Williams - Maryland
7 – 11, Jim Boeheim - Syracuse
7 – 10, Bo Ryan - Wisconsin
6 – 17, Ben Howland - UCLA
6 – 12, Thad Matta - Ohio State
6 – 10, Bruce Webber - Illinois
6 – 8, Mike Montgomery - California
6 – 7, Jay Wright - Villanova
6 – 6, Al Skinner - Boston College
6 – 5, Mike Brey - Notre Dame
5 – 12, John Calipari - Memphis
5 – 9, Rick Pitino - Louisville
5 – 7, John Thompson III - Georgetown
5 – 7, Bruce Pearl - Tennessee
5 – 6, Ernie Kent - Oregon
5 – 6, Jamie Dixon - Pittsburgh
5 – 6, Paul Hewitt - Georgia Tech
5 – 5, Mark Gottfried - Alabama
5 – 5, Herb Sendek - Arizona State
5 – 5, Tom Crean - Indiana
4 – 7, Mike Davis - UAB
4 – 6, John Brady - Arkansas State

My rankings evaluate how coaches obtain these appearances and wins. Is it through recruiting? Is it through player development and by earning a high seed in the regular season. Or is it by exceeding expectations in the tournament?

Let’s start with coaches that exceed expectations in the tournament. Based on past NCAA tournament data, here are the wins expected for each seed.

ExpWins - Seed
3.42 - 1
2.41 - 2
1.82 - 3
1.49 - 4
1.15 - 5
1.24 - 6
0.88 - 7
0.67 - 8
0.58 - 9
0.64 - 10
0.49 - 11
0.50 - 12
0.25 - 13
0.18 - 14
0.04 - 15
0.00 - 16

This spring I discovered that ESPN also uses this concept. ESPN calls it PASE or Performance Against Seed Expectations. Here are the coaches with the highest PASE in the last 10 years, averaged over the number of NCAA appearances, with a minimum of three NCAA appearances.

PASE Coach -School
0.72 Tom Izzo - Michigan State
0.71 Billy Donovan - Florida
0.67 Ben Howland - UCLA
0.66 Jim Larranaga - George Mason
0.56 Mike Davis - UAB
0.55 Roy Williams - North Carolina
0.52 Jim Calhoun - Connecticut
0.47 Sean Miller - Xavier
0.47 Gary Williams - Maryland
0.46 Rick Pitino - Louisville

Jim Larranaga and Mike Davis show that if you rarely make the tournament, but then make the Final Four, you look like a good tournament coach.

PASE Coach -School
-1.14 Mike Montgomery - California
-1.04 Oliver Purnell - Clemson
-0.53 Rick Stansbury - Mississippi State
-0.47 Bob Huggins - West Virginia
-0.46 Lorenzo Romar - Washington
-0.44 Greg McDermott - Iowa State
-0.42 Dana Altman - Creighton
-0.42 Brad Brownell - Wright State
-0.36 Fran Dunphy - Temple
-0.35 Bobby Lutz - Charlotte

California is getting a good regular season coach, but Mike Montgomery had some colossal tournament flops before he left Stanford for the NBA. Oliver Purnell took Dayton to the tournament as a 4 seed and 11 seed and took Clemson as a 5 seed, but has won zero tournament games. Rick Stansbury has taken Mississippi St. to the tournament five times in the last decade, but he has never advanced to the Sweet Sixteen. Bob Huggins improved his numbers with a deep tournament run this year. In the last decade, Fran Dunphy took Penn to the tournament 6 times and Temple once, but he does not have a tournament win to show for it.

Some coaches earn their NCAA tournament wins and appearances through recruiting.
I ran a regression to estimate the impact of talent, looking within coaches over time. Not surprisingly, McDonald's All-Americans and upperclassmen that were top 100 recruits had the biggest impact on success. Based on these regressions I assigned a value to each type of recruit. I also lowered the value of McDonald's All-Americans based on the fact that most leave early for the NBA. With these numerical evaluations, I assigned a value to each recruiting class for the last 9 years. I only evaluated coaches at their current school, and I threw out the first year in the current job. Based on these numbers, the top recruiters are:

RECR Coach - School
1.80 Mike Krzyzewski - Duke
1.57 Bill Self - Kansas
1.41 Roy Williams - North Carolina
1.31 Billy Donovan - Florida
1.13 Thad Matta - Ohio State
1.10 Ben Howland - UCLA
1.03 Tom Izzo - Michigan State
1.02 Rick Barnes - Texas
1.02 Paul Hewitt - Georgia Tech
0.93 Jim Boeheim - Syracuse

Stewart Mandel of Sports Illustrated had a great article pointing out that Coach K gets a lot of McDonald's All-Americans, but the causality may be wrong. His recruits may be called McDonald's All-Americans simply because they decided to go to Duke, not because they are really top 20 players. If this is true, you may want to lower Coach K's recruiting rank.

Also, according to my evaluation, recruits aren't everything. Tom Izzo has enough of a tournament coaching advantage (see above) over someone like Bill Self, that he can make up for having less talented recruits. (The key problem for Izzo in recent years is that he has struggled in the regular season.)

Here are the worst recruiters at a BCS school for at least 4 years (i.e. minimum 3 recruiting classes.)

RECR Coach - School
0.00 Bill Carmody - Northwestern
0.04 Ed DeChellis - Penn State
0.14 Al Skinner - Boston College
0.20 Oliver Purnell - Clemson
0.23 Norm Roberts - St. John's
0.31 Jeff Lebo - Auburn
0.32 Seth Greenberg - Virginia Tech
0.33 Frank Haith - Miami
0.34 Dennis Felton - Georgia
0.42 Bruce Webber - Illinois

The worst of the worst recruiters tend to lose their jobs at about the 3 year mark, so I should add that Greenberg, Haith, Felton, and Webber are hardly horrible. In fact, the only way you keep your job if you can't recruit at the elite level is by knowing how to coach.

You can also earn wins and appearances during the regular season by developing players and earning a high seed. For the last 6 years, I take expected wins based on seed and subtract off expected wins based on talent and get regular season performance. (I only have 9 years of recruiting data, and I can only evaluate each team's seniors for the last 6 years, so I can only make this calculation for the last 6 years.)

REG Coach - School
2.47 Roy Williams - North Carolina
2.33 Mike Krzyzewski - Duke
2.19 Bill Self - Kansas
2.14 Rick Barnes - Texas
2.04 Jamie Dixon - Pittsburgh
2.02 John Calipari - Memphis
2.02 Bo Ryan - Wisconsin
1.99 Mark Few - Gonzaga
1.96 Jim Calhoun - Connecticut
1.90 Ben Howland - UCLA

Not surprisingly, the best coaches at winning in the regular season were also some of the best recruiters. (Players like to play for a winning coach.) The interesting feature here is that Jamie Dixon and Bo Ryan show up on this list even though they weren't on the recruiting list. Even with less talent, these coaches have been consistent winners in the regular season.

REG Coach - School
-0.69 Leonard Hamilton - Florida State
-0.38 Jeff Lebo - Auburn
-0.23 Norm Roberts - St. John's
-0.21 Andy Kennedy - Mississippi
-0.18 Travis Ford - Oklahoma State
-0.09 Jerry Wainwright - DePaul
-0.02 Bill Carmody - Northwestern
0.03 Scott Drew - Baylor
0.03 Ed DeChellis - Penn State
0.06 Bobby Gonzalez - Seton Hall

On the flip side, Leonard Hamilton has turned some great recruiting classes into zero NCAA tournament appearances at Florida St. I also give less weight to success at non-BCS schools based on the fact that most coaches making the transition fail. Even so, I find it a little surprising that Travis Ford earned the Oklahoma St. job after FAILING to qualify for the NCAA tournament with a talented team.

REC , REG , PASE , TOT, School - Coach
1.41 , 2.47 , 0.55 , 4.43, North Carolina - Roy Williams
1.57 , 2.19 , 0.25 , 4.00, Kansas - Bill Self
1.80 , 2.33 , -0.34 , 3.79, Duke - Mike Krzyzewski
1.10 , 1.90 , 0.67 , 3.68, UCLA - Ben Howland
1.31 , 1.57 , 0.71 , 3.58, Florida - Billy Donovan
0.91 , 1.96 , 0.52 , 3.39, Connecticut - Jim Calhoun
0.90 , 2.02 , 0.20 , 3.12, Memphis - John Calipari
1.02 , 2.14 , -0.07 , 3.09, Texas - Rick Barnes
1.13 , 1.45 , 0.32 , 2.90, Ohio State - Thad Matta
1.03 , 1.09 , 0.72 , 2.84, Michigan State - Tom Izzo

This list should look a lot like the list of wins and appearances at the start of this post. I've simply broken apart where the success comes from. Bill Self is a star recruiter; John Calipari fattens up in the regular season and earns a high seed and better NCAA tournament path; and Tom Izzo makes his money by exceeding expectations in the tournament.

The above number says Bill Self should average 4.0 wins plus appearances per year. But Bill Self's numbers in the initial list were 10 appearances and 22 wins, or 3.2 per year. The key difference is that the recruiting figure is ONLY for Kansas. This does not include Self’s recruiting at Tulsa and Illinois which was at a slightly lower level. The regular season and tournament figures are derived from all of Bill Self’s jobs. Even though he only earned 3.2 wins plus appearances per year, if he stays at Kansas, this model suggests that Bill Self should average 4.0 wins plus appearances in the future.

For someone like Coach K at the same school, the discrepancy between the actual 3.6 wins and 3.8 wins is explained by the different time horizons for the various factors. I evaluate RECR for 9 years, REG for the last 6 years, and PASE for the last 10 years.
Differences will also occur when there is an unusual amount of early entry at a school. I currently do not penalize coaches for early entry.

I'll be back later with a conference by conference breakdown.