Defense metrics
Re: Defense metrics
Thanks for the notice.
I'd replied to author's twitter.
If interested, check his post there or at my feed @bballstrategy.
I'd replied to author's twitter.
If interested, check his post there or at my feed @bballstrategy.
Re: Defense metrics
That's a pretty good article. Nothing seen to disagree with, other than the Einstein 'quote' coined by someone else.
He doesn't seem to suggest 'substitution patterns' as a confounder in the plus-minus measures: Do you keep Bruce Bowen on Kobe, to keep his numbers down? or go with Brent Barry to make Kobe work on defense?
One strategy (Bowen v Kobe) might make Bowen look worse on defense than Barry, since he's only in the game when Kobe is. That arrangement also makes Kobe look better on D, since Bowen isn't helping the Spurs' offense.
Going with Barry, the Spurs should look worse on D; unless the Lakers counter with a defensive specialist to give Kobe a break.
Player tracking or lineup analysis may sort these things out. But when you break it into tiny intervals, your data gets confounded by stuff like a player checking in between free throws, or with a few seconds on the clock, etc.
https://quoteinvestigator.com/2010/05/2 ... -einstein/
He doesn't seem to suggest 'substitution patterns' as a confounder in the plus-minus measures: Do you keep Bruce Bowen on Kobe, to keep his numbers down? or go with Brent Barry to make Kobe work on defense?
One strategy (Bowen v Kobe) might make Bowen look worse on defense than Barry, since he's only in the game when Kobe is. That arrangement also makes Kobe look better on D, since Bowen isn't helping the Spurs' offense.
Going with Barry, the Spurs should look worse on D; unless the Lakers counter with a defensive specialist to give Kobe a break.
Player tracking or lineup analysis may sort these things out. But when you break it into tiny intervals, your data gets confounded by stuff like a player checking in between free throws, or with a few seconds on the clock, etc.
https://quoteinvestigator.com/2010/05/2 ... -einstein/