We've Moved!

We've Moved!
Click here to head to riddickandreynolds.wordpress.com!

Tuesday, February 2, 2010

Talent Usage Rating: Quantifying The Amount Of Talent On The Floor


We often hear folks say, "this team is more talented than ours..."

But just how much more talent? More importantly--how much of this talent is actually making its way onto the hardwood?

A bench full of 5-star talent that never sees the floor ultimately doesn't have any impact the outcome of a game. That makes it mighty tough to use the "more talented" argument.

So, in an effort to help quantify just how much talent each team is bringing to bear each game, I've put together a spreadsheet featuring the "Talent Usage Rating," a formula that is a cumulative of a team's player's minutes divided by 200, multiplied by their average recruiting ranking.

In other words, a team's Talent Usage Rating = (Player A's Avg Rating * (Avg Mins Played/200)) + (Player B's Avg Rating * (Avg Mins Played/200)) + etc.;

Here the spreadsheet (factoring in games as of Monday evening):



A couple of things to note about the numbers:
  • The average minutes played for each player is calculated by taking their total minutes played divided by the total number of games played by the team. Doing so eliminates the possibility of a player's average minute stats being inflated by not playing in one or more games.
  • The average star rating for every player is determined by averaging the Scout and Rivals rankings. There are more ranking services out there, but these two are the most respected. Though there was some difference in the rating of certain players, for the most part both sites agreed on the star ratings of most players, and there was never more than a one-star difference for any player.
  • Players that did not show up on the Scout or Rivals sites or were listed as "NR" were given 1-star ratings.
As you can see, a team simply having the highest average star rating doesn't necessarily mean they have more talent on the floor than the next guy. Florid State's TUR is nearly 2% higher than Carolina's, despite the fact that UNC's roster averages a nearly 4% higher star rating per player. In other words, while FSU's overall talent level on the roster may be behind Carolina's, they put more of their higher-rated players out on the floor for more significant minutes. So when the Heels face off against the Noles on February 24th, don't assume that the Heels have a talent advantage on the court.

With the 11th-worst TUR and 10th-worst overall average roster star rating, Virginia is clearly getting the most "bang for their buck." They have talent where they need it (Sylven Landesberg) and are getting great contributions elsewhere from lower-rated players.

On the flip side, the Tar Heels are clearly under-performing relative to their TUR, but that's no new news to the entire college basketball community.

For those that want to see the entire breakdown of each roster that led to the rating calculations, here's the complete spreadsheet:

12 comments:

  1. this is really good work! you must have put a lot time into the research. at some sites during the period before the duke game i made some comments about State's overall lack of talent and that State was starting four three star recruits and one four star recruit. i concluded that with the addition of more five star recruits State would have better results. lately i have been commenting on Lowe's micromanaged slow tempo inside out game and wondering if the addition of five star recruits will make much different. right now i feel that the offense State is running will have difficulty with an undersized inside player no matter how talented he is-and Tracy is pretty talented. you made good comments with regard to Virginia's success and UNC's lack of success despite the talent level difference, so something more than talent is at play. in Virginia's case better coaching (at least a better defensive and offensive scheme) and a very good four star player is in play. i incorrectly thought Landesberg was a five star top twenty player, so i was surprised to see that he was a four star player. sbas2

    ReplyDelete
  2. i might also add that this information is something different than what is seen at other sites. in the past this site has given one the same game information as "Backing the Pack". good to see a new direction. sbas2

    ReplyDelete
  3. Normalized star rating (0-5) over normalized RPI rating (1-347) then gives us a measure of the coaching of available talent post-recruiting.

    1 Clemson (1.34)
    2 Maryland (1.32)
    3 Wake Forest (1.30)
    4 Georgia Tech (1.29)
    5 Virginia (1.26)
    6 Boston College (1.26)
    7 Virginia Tech (1.23)
    8 NC State (1.19)
    9 Florida State (1.17)
    10 Duke (1.13)
    11 Miami (1.04)
    12 UNC (0.96)

    One could then try to project where the team will be once the average star rating goes up. But, our star rating is heavily impacted by Vandenberg, who should have never been a 2-star recruit. That kid has talent and perhaps the highest potential of anyone currently on the team.

    ReplyDelete
  4. i agree on Vandenberg. i think it was difficult considering that Vandenberg was in Australia for the recruiting services to get an accurate feel for his ability.
    i noticed in the unc game that Vandenberg attempted a sky hook. of course when he missed Lowe immediately took him out. bring in Tommy to teach him as with his height such a shot can be effectively used much as Burleson used it.

    sbas2

    ReplyDelete
  5. ^ Good stuff. But one question: Is the Normalized star rating factoring in all players' ratings equally? Part of the reason for doing this was to get a better feel for the talent that's actually seeing the court, not just an average of all star ratings.

    I think I need to add one more tweak to it, though: By dividing the TUR by 200 (5 players by 40 mins.), the TUR would give a representation of average talent level on the court on a scale of 1-5.

    For example: FSU's Avg. Star rating of all players is 3.91, but taking their TUR and dividing it by 200 gives us a figure of 4.16. It's a relatively small difference (a quarter of a star), but it's a difference nonetheless.

    I'd be interested in seeing how your rankings would be impacted by using this new figure versus the normalized star rating. There are a couple of teams with close enough ratings (Virginia - Virginia Tech) to perhaps shake up the standings.

    ReplyDelete
  6. Nice post James. These are definitely some good statistics to have in the back pocket.

    ReplyDelete
  7. Updated spreadsheets to reflect new tweaks and make it easier to compare average star ratings with talent usage ratings.

    ReplyDelete
  8. Thanks CGB. The next evolution of this spreadsheet will be to factor in only the last 4-8 games played or to start at that point in the season (usually around mid-January) when most teams have finalized their rosters. Doing so eliminates minutes played by scrubs earlier in the season and takes into account those players (like a Tyler Zeller) that are injured and have not played recently.

    ReplyDelete
  9. Great information here. from first look, it looks like we have alot less stars than virtually everyone in the ACC and thats a problem. packbackr04

    ReplyDelete
  10. I love the work that you've done - it really shows that we have talent problems at the guard spot. Maybe that's why we can't seem to shoot a higher percentage. We get plenty of open looks, but can't seem to put them in the hole.

    I actually thought that we are in the top 1/3 of the talented teams in the league, but evidently not. Hopefully, Lowe will do better once we have guards that have higher than a 3-star rating.

    I thought for sure that Maryland had less talent than we have, but evidently not. Another way of modifying the spreadsheet would be to adjust the star ratings based on how many years that player has played in the league. Is a 3-star senior as good as a 5-star freshman?

    ReplyDelete
  11. ^ Funny you should mention that...I was just today pondering adding an experience quotient to further modify and clarify the numbers.

    I'm already thinking of things to include in Version 2.0...

    ReplyDelete
  12. Interesting system, needs tweaking but it sounds like you're already doing that. A couple quick fixes that need to happen - both Josh Davis and Jordan Vandenberg were Rivals 3* but are listed as 1* in your db. Same with Jin Soo Choi (who changed his name from Jin Soo Kim) for Maryland, though that's somewhat irrelevant b/c Choi has left MD and returned to Korea.

    ReplyDelete