The NBA season is looooong.
It's not quite the interminable grind of baseball, but it's not too far away from it either. Playing 82 games over 6 months, teams go through tactical changes, players move in and out of lineups, get traded, and intra-team dynamics change for better or for worse.
And so looking at the season-long results or stats might not really represent their "current" state accurately. The season-long stats are just so attenuated by averaging outcomes over a long time.
So I decided to build these graphs which look at individual game performances. Here are these charts by the league's top teams:
- Each marker represents one game
- The two axes are offensive & defensive ratings (scores adjusted by pace)
- The more recent the game, the larger the marker
- Games from the last 20 days are highlighted in different colours
- Gray dots are league-wide games
- The top left quadrant (green background) is good
- The bottom left quadrant is the corner of shame (bad O, bad D)
Yes, the recent games are smaller samples and some teams' results are just all over the place (looking at you, Bucks/Jazz). But some teams' recent results are clearly different enough from the rest of their season that I think it is unlikely to be a sample size thing. Especially:
- The Cavs should be worried out their offense.
- The Warriors need Draymond back - their defense is 😞.
- The Grizzlies are firing on all cylinders!
Oh, and then there's the Celtics.
Three of the Celtics' best four games, or four out of seven, have come in the last 20 days. Is it really just coincidences?
At the other end lies the Nets, who... erm... have had some recent difficulties.
They've been firmly planted in the bad D, bad O naughty corner for the most part, not that their D has been very good anyway.
Putting a summary of the ratings info all into one chart, we see:
I don't know whether this just small sample size theatre or not, I am no basketball expert. But some of the outliers like the Celtics and Nets being good/bad make sense given what's going on in the team.
More work would be needed to take guesses at separating components of luck and changes to actual form here. But it's interesting to think about.
For what it's worth, this is a chart captureing teams' consistencies (measured by standard deviation of off/def ratings). This isn't necessarily that predictive of the current variance in form as far as I can tell.
I have started to look at correlations on a game-by-game basis as well, and that's kind of interesting too, as some teams' offensive performances tend to vary together with defensive performances. This could be due to a number of factors, but it certainly could be a case of some teams getting down on itself or front-running.
Anyway, to sum it up, here's everybody's form condensed to just one chart, comparing teams' recent net ratings against their season-long net ratings. It doesn't quite tell the full story, but it's a pretty decent short summary nonetheless.
This is all pretty early, and a lot of it is me thinking out aloud. I think I will come back with more on this topic for sure.
Streamographs are usually not good - but this is beautiful, and one time where the overall impression is probably way more important than accuracy.
This issue was at least partly inspired by:
Highlight of the week: I got retweeted by Seth Partnow! Also, is Seth right? should I be doing stats astrology instead?
Thanks everybody for reading.
It's been really good to have this as an outlet and a way to engage with people during the pandemic. I hope to keep it up :)
Pls say hi on Twitter / let me know if you enjoyed this issue!