DaggaStats, Part II: College Basketball

With the Shaka Smart rumors flying tither and yon, I figured it was time to export the football coaching analysis to the world of hoops coaches. I don't have a lot of time to dig into it right off the bat, but I'll gloss over a couple of predictable points of interest and follow up on your questions in comments.

Method Recap

For the uninitiated, or forgetful:

DaggaStats is my method for analyzing the suitability of past and current college coaches for a current opening.

I assume that team quality (Q) = recruiting talent (T) + coaching (C), and then use rescaled statistical proxies for Q and T (40 = worst team, 100 = best team, 70 = average) to derive a measure of C.

Then I rate the past performances of coaches using C and T. I use different two methods for this.

The simple method is to multiply their C and T scores together (CxT). This scoring method is marginally better than Q alone (C + T), because adding C and T together disguises those cases where the coach is severely underachieving in one area and making up for it in another. Multiplying them together ensures that the top scores go to those who do well in both C and T.

The more complex method is called the Replacement Suitability Index, or RSI. This method weights the result for the level of talent at the coaching position you're trying to fill. If you're looking to fill a position at the least talented team in the country, C is your only concern; if you're trying to fill a position at the top school in the country, both C and T are equally indispensable and thus weighted 50/50. Schools in between are given a score on a sliding scale, based on their current talent level.

For college basketball, the Team Quality (Q) proxies I'm using are the RPI and the Kenpom rankings.

My Recruiting Talent (T) proxy is the 247 Composite Team Rankings.

Using these three sources, we have twelve solid years of data. Since we can't effectively start the analysis until we have four years of data, this analysis covers the last 9 seasons, starting in 2006-2007.

One final note about the 247 rankings: sadly, it does not fully cover 12 years of recruits for 351 teams. It mostly covers those recruits who were given a 5-, 4-, or 3-star ranking before they were signed. Of 3105 coaching seasons in the dataset, a whopping 1458 of those seasons were for teams that did not land a single player recognized in the 247 Composite in the previous four years. That's roughly 45% of the total.

I dealt with this issue the only unbiased way I could: I looked at how those teams actually fared (Q), picked the median value as the best guess of how any of those teams picked out of a hat would do, and used that median Q value as the baseline T value. That T score is 61.75.

So wherever you see a T score of 61.75, it means that coach did not have any players at all that made the 247 Composite. If the coach had a better score than that, they landed at least some ranked players.

The Results

Looking at current single-school coaching tenures of at least three years, here are the Top 30 in RSI using Texas as the RSI benchmark:

Gotta admit, it gives me warm fuzzies to look at the Top Ten produced by this method. All of those guys have a legit claim to be there.

But beyond that, here we have some insight into the two key names bandied around for the Texas position this week: Shaka Smart and Gregg Marshall.

Here's the last nine years at VCU:

By contrast, here's Wichita State:

See the key difference?

In eight years at Wichita State, Gregg Marshall has not successfully recruited a single high school, prep, or Juco player deemed worthy of consideration by the ranking services. Meanwhile, Shaka Smart and Anthony Grant have improved VCU's recruiting talent level by a point per year. That's exceptionally rare. No other small-school coach in the sample accomplishes anything close to that (though Larry Brown might, eventually). Not even Brad Stevens.

EDIT: And I'm afraid Wichita State was one of the schools impacted by the scraping bug (see comments). Marshall has indeed recruited some decent players to Wichita - his T score should actually be in the mid-to-high 70's, pretty good for a MVC team. However the numbers don't show any upward movement at all in his ratings since he arrived. At least as far as the 247 scores are concerned, Turgeon was a better recruiter for the Shockers.

I'll post an updated version of the spreadsheet when I can figure out how to scrape the data accurately. If it's doable at all, it's going to be a beast. I might need to switch proxies altogether, which would suuuck.

So HEEEEY 247, whassup - if any of your employees are reading this, tell your bosses that your Team Composite lists are steaming piles of horsedoody. You have that data on hand, right there in your web server even, scattered on individual team pages. There's no excuse for leaving dozens of teams out of the rankings.

I am doing this as a hobby, free of charge; you are trying to convince people to pay real money for your tools and analysis. For that, you need actual quality control. Get it together.

Open question: when will purveyors of tabular data realize that they need to provide the option to download their data in machine-readable format 100% of the time, always squared, zero exceptions? Do we have to wait for ALL the Baby Boomers to die off first? Just wondering.

For old times' sake, let's look at Rick Barnes:

If an elite recruiter at a highly visible school also wants to be known as a good coach, he needs to keep his C scores at or over 70. That score means his teams are living up to their talent.

Much like Mack Brown, Rick has consistently fallen short of that standard since early 2010. Super-high talent levels sustained a pedestrian level of team success for awhile but that gravy train either is approaching its final stop or has already reached it. Texas needs not only a quality coach, but an exciting recruiting presence to shore up some enthusiasm for a program in a recruiting area filled with capable poachers like Billy Kennedy, Larry Brown, and Scott Drew.

Long story short: save an unexpected defection from an established coach like Bo Ryan or Mark Few, the numbers say that Smart is the best guy available out of this pool. He's a plus recruiter whose teams consistently overperform. Plus he's young, so if he's any good at Texas he'll stand a solid chance of ushering the team into its new arena.

The main downside is the puns. The awful, awful puns.

The spreadsheet can be downloaded HERE.

I'll have inconsistent web access over the next couple of days, but if there's anything you'd like to see, let me know in comments and I'll get to it eventually.