Friday, August 21, 2009

Worst Coach in the W?



With Sue F. of the They're Playing Basketball blog asking what makes a good coach, it sparked me (no pun) to write a post about bad coaches.

One of the great haters of the WNBA - who shall be known by his appropriate initials of "B. S." - wrote an ESPN post that characterized what could indicate a team that has a bad coach:

1. A team with a poor record in close games.
2. A team that couldn't take care of the ball, reflected in turnovers.
3. A team that gave up too many offensive rebounds.
4. A team that couldn't put together win streaks.
5. A team that gave up too many 3-point shots.
6. A team that had a bad road record.
7. A team that doesn't have a consistent rotation.
8. A team where the coach makes egregiously stupid moves.

I decided to do some number crunching.

"Close games" I could measure - Swanny's Stats had the most recent stats that indicated team records in games decided by six points or less. (Yeah, B. S. used five points or less. I don't have better stats, and B. S. can bite me.)

Turnovers, opponent 3-point shooting and road records are provided at the WNBA website. Offensive rebounds allowed per game are provided at each of the team sites. I created a new stat called "Streak Wins", which is a team's total wins counting only winning streaks of three games or more. For example if a team has winning streaks of 5, 3, 1, and 1 the team has eight streak wins.

I only judged 10 coaches - John Whisenant, Rick Mahorn and Anne Donovan all get free passes. For the remaining 10 coaches, they were judged on a 0 through 9 scale in each of the categories above. I don't have a metric for rotation consistency, and no metric on earth can measure stupidity.

This leaves me a number I call the "B. S. Coaching Number." It measures how bad a coach you are. Finishing from #1 to #10 are:

1. Marynell Meadors
2. Steven Key
3. Dan Hughes
4. Jennifer Gillom
5. Michael Cooper
6. Julie Plank
7. Brian Agler
8. Mike Thibault
9. Corey Gaines
10. Lin Dunn

"But Pet," you might ask, "aren't you just measuring how bad a team is instead of how bad a coach is?" Good point. I decided to correlate my final scores with the teams "losing percentage", or 1 minus winning percentage. The correlation was 0.40 - which is about a medium correlation. This means that a team's overall record accounts for *some* of a coach's placement on the list, but not *all* of it.

This very primitive metric - it's not scaled, which means that you get an extra point whether you finish 1 percent or 100 percent above your nearest competitor - seems to confirm at least in the #1 and #2 positions what people already believe about certain WNBA coaches: you can't find a WNBA message board without anyone screaming about Meadors and/or Key. Dan Hughes's poor placement, on the other hand, might have to do more with the team than with him. Likewise, Lin Dunn's good placement - the "least bad of the bad" - might have to do more with the Fever than with her.

I might have to think about this a bit more. But it's a nice mental exercise.

4 comments:

Q McCall said...

That's a really solid way to statistically operationalize "coaching quality".

About rotations, the Arbitrarian once had a post on Hardwood Paroxysm about quantifying rotation consistency...

Two thoughts:

1) I think maximizing the talent you have -- even if it doesn't show up in win% -- is an important quality of a coach. Implementing a strategy that inhibits a player's ability to perform is the mark of a bad coach.

2) So when considering that standard, this system is rather unfair to Gillom. She has a young team that lost their star player and has them in playoff contention. That's great coaching to me.

It's easy to win when you've been given tons of talent, much more difficult when you have to make due with less...

pt said...

Q, I'll need to look up that article at Hardwood Paroxysm.

With regard to Gillom - due to that 0.40 correlation, it means occasionally that the metric will confuse a bad coach with one who simply has a bad team. The Lynx did lose their star player. However, the Lynx haven't been playing very well at all recently, so you never know.

Q McCall said...

Here's the rotation piece... he looked at rotation sizes and then used the standard deviation as a means to measure variability...

http://hardwoodparoxysm.blogspot.com/2008/09/arbitrarian-empirical-estimates-of-team.html

I don't know if that's helpful or not for this..

As for Gillom, I think they were way overachieving early on and have fallen back to earth... but your point is solid.

Q McCall said...

OK... I should probably be doing other things with my day that have financial rewards...

But I just found two things that might be of interested:

1) Hollinger using wins vs. expected wins to evaluate coaches:
http://sportsillustrated.cnn.com/2004/writers/john_hollinger/09/20/best.coach/index.html

2) Berri had an article about coaching as well...
http://dberri.wordpress.com/2008/11/18/coaches-and-deck-chairs/

Crunch the numbers so I can use your data!