Data Engineer @ Utah Jazz

Home for all your discussion of basketball statistical analysis.
Post Reply
Posts: 2
Joined: Fri Feb 02, 2018 12:34 am

Data Engineer @ Utah Jazz

Post by independent_variable » Fri May 03, 2019 9:33 pm

Hi all,

My name is Cory, I run the Analytics group for the Utah Jazz. We’re currently hiring, and I thought I would post this here in case anyone has a data engineering background and might find this interesting. Feel free to drop me a DM or email if you have any questions, thanks. 7

Posts: 6251
Joined: Thu Apr 14, 2011 11:10 pm

Re: Data Engineer @ Utah Jazz

Post by Crow » Mon May 06, 2019 11:29 pm

Have you considered having more data analysts?

I know you want / need data processes and products (and engineers to produce them); but whatever you have produced is only as good as the analysis from & promotion of it. More data analysts probably could help find and further analyze more things.

I don't know what process and products teams currently have on lineups; but from the outside looking at lineup patterns and results, I'd say most teams team need way more lineup analysis and way more dialogue / instruction on lineup management based on that analysis.

For the Jazz, the regular season lineup management moved to fine as the season progressed. But the playoff lineup results were quite bad. Every 10 most used quad, trio and pair was negative. Most bad to horrendous. The 4 most used 5 man lineups that played almost half the game were all really bad performers. Things didn't go as planned. Who is more likely to find out why, the folks who made those plans by themselves or folks who made those plans in conjunction with outsiders?

Less than 26% of the selected playoff lineups were cumulatively positive for the series. That is pretty low. Only 3 playoff teams were lower. Jazz used 74 lineups in playoffs. Probably close to mean and close to median. But that many were probably not "needed" and probably not optimal.

In regular season Jazz tested 2 lineups to the level of about 500 minutes and then everything else is 150 minutes under. Lineup management is more art than science given sample sizes, especially if the team doesn't aggressively test a small group of lineups. Was it necessary or wise for Jazz to use 643 lineups in regular season? 636 would qualify as "dink lineups" under a 1 minute per game for season standard. The 20 most used lineups got 54% of the regular season minutes. I'd have to check further to see where that falls within the range for the league; but it could have been higher and improved the level of lineup testing. In playoffs, the 20 most used lineups got 82% of the minutes. That is good concentration, but they might not have been the "right" 20 or in the right proportion or sequence. Apparently they weren't and not even close... or players just failed to perform adequate to the plan. Spending 46% of the regular season minutes outside the 20 most used lineups and then not using about 90% of those smaller lineups doesn't seem optimal management to me.

A few quick thoughts, fwiw.

Post Reply