Why WS, ORTG, DRTG on basketball reference are useless advanced stats

[Gold Dragon]

There was a poster asking about WS and I had a recent discussion with someone about individual ORTG and DRTG on basketball reference. The statistical information provided on basketball reference is invaluable and many of the advanced stats they list are very useful for analysis. But their calculations for ORTG and DRTG and any stats derived from them (including WS) are widely known as being essentially useless stats. That is because they depend on flawed assumptions and very complex estimations to convert box score statistics into a number that tries to approximate individual contributions to team ORTG and team DRTG.

In contrast, the individual ORTG and DRTG on stats.nba.com use tracking data to calculate the actual points for or against when a player is on the floor, which is a much more useful piece of data.

Here is how BREF explains their calculations and they also acknowledge some of the flaws in those calculations.

https://www.basketball-reference.com/about/ratings.html

Here are some more detailed breakdowns of the flaws in their DRTG formula if you are interested in the math. ORTG has similar flaws.

https://www.slcdunk.com/2013/1/31/3939352/the-offbeat-a-quick-note-on-individual-defensive-ratings

Here is a great discussion we had here a few years ago.

viewtopic.php?f=32&t=1545097

[Dr Positivity]

I always viewed basketball reference's ORTG as basically just an efficiency stat. It's like taking TS% and adding assists and turnovers information to it. For example Lowry has 116 and Powell has 108 despite Lowry having .567 TS% and Powell .575. ORTG is more friendly to Lowry because it includes his passing. Overall it does an acceptable job of judging how efficient someone uses a possession in my opinion

[vini_vidi_vici]

Just a few corrections. While Gold Dragon got the definitions mostly right, for clarity..

BBREF is individual RTGs (ex: iDRTG)
NBA.com is team RTGs while player is on the court. (ex: the teams NetRTG while KL is on the court is +9.3).

I dont want to delve too deep into things (we can if you would like), but the inherent issue with iDRTG is that, its reliant on a) the teams DRTG and b) it assumes all players are the same and what differentiates them is BLKs/STLs (2 defensive stats). It assumes everyone gets the same amount of individual stops and faces the same number of possessions. Guys who stop players but dont get those 2 stats are going to fall behind in iDRTG despite being better defenders. Fortunately we have alot of aggregate match up based stats, whether Synergy/MatchUps on NBA.com/etc..

For iORTG, you have to consider USG% (which does not use ASTs) and role definition, to make comparables. That in and of itself is tough enough.

Most boxscore stats are tough in general because of the assumption we have to make for individuals vs their team stats (this applies to ppls love of Win Shares too).