On October 3rd of this year, Alexander Franks, Alexander D’Amour, Daniel Cervone and Luke Bornn (hereinafter referred to as AAD&L) published a paper entitled “Meta-Analytics: Tools for Understanding the Statistical Properties of Sports Metrics.”
Honestly, I have heard of Meta-Humans (thanks to the CW’s “The Flash” television series - you know humans with super abilities that were acquired by a particle accelerator explosion); but until now, had no idea that such a thing as Meta-Analytics existed. More specifically, Meta-Metrics.
What do Meta-Metrics do? They identify metrics that provide the best information to decision makers based on three criteria.
The three criteria are (1) stability, (2) discrimination, and (3) independence. Stability is defined as measuring the same thing over time; whereas, discrimination is defined as differentiating between players (i.e. player position); and independence is defined as providing new information.
Basically, in a world with new metrics being developed daily, AAD&L attempt to sift through the metrics to find the ones that work by cutting through noise and answering difficult questions.
Two sports are studied in the paper - basketball and hockey. I’m not going to discuss the hockey portion, but here are some of the basketball findings:
- Rebounds, blocks, and assists are strong indicators of player position (i.e. highly discriminative and stable)
- Raw three point percentage is the least stable and least discriminative of all the metrics looked at
- Rate based stats (stats based on per-game or per-minute such as Win Shares per 48 - WS/48, Offensive Rating - ORtg, Defensive Rating - DRtg, Box Plus/Minus - BPM) tend to be less discriminative but more stable than those that incorporate total playing time (such as Win Shares - WS, and Value Over Replacement Player - VORP)
- Rate based metrics are good for estimating player skill, whereas metrics incorporating totals are good for identifying overall contributions
AAD&L ended up with a preference for BPM for the rate based stats and VORP for total stats.
If you have the time, the paper is worth the read.
Since this edition of Friday Morning for the Weekend is Episode 007, I thought I’d go with a James Bond clip; enjoy!
In episode 006, both Chowda and 15,806 assists likened Star Wars characters to Utah Jazz players, coaches, and personalities; however, it is Fesenko for President with his suggestion that Amar is George Lucas and My_Lo is Amar’s muse who suggested the obvious (i.e. staff SLC Dunk with fan posters for the 2016-17 NBA season) that wins the Episode 006 Worst Analogy Championship Belt.
Thank you for all who participated; it was difficult to see 15,806 assists go down as he’s consistently competed at a high level week in and week out.
Where do Jazz players rank in BPM and VORP? See below:
VORP (measuring player contribution; negative two is average replacement player value)-
#15 in NBA, Rudy Gobert at 1.1 (tied with 4 players)
#21 in NBA, George Hill at .8 (tied with 5 players)
#33 in NBA, Gordon Hayward at .6 (tied with 15 players)
#48 in NBA, Rodney Hood at .5 (tied with 16 players)
BPM (measuring player skill; zero is league average)-
#13 in NBA, George Hill at 7 (tied with one other player)
#25 in NBA, Rudy Gobert at 5.1
#50 in NBA, Gordon Hayward at 3
#103 in NBA, Rodney Hood at 1.1 (tied with one other player)
There are anomalies in BPM since it is a per minute type of measurement; one of which, our own Joel Bolomboy comes in at #22 in the NBA with a score of 6.1.