Thread by @ElGee35

In response to some questions about passer rating, here’s a thread about the history of valuing assists, how passer rating was born and what it captures. (Links at the end) #thinkingbasketball 1/14

I’ve always been disappointed in assists by themselves - they need context when used to measure impact or skill. Assist:turnover ratio was often cited in the past, but it’s not always helpful. 2/14

A previous (unpublished) version of the passer rating regression actually pegged ast:tov as a slight *negative* — too much conservatism might indicate an unwillingness to try risky passes that are high ROI. It’s the dink and dunk of quarterbacking for basketball. 3/14

The aim of my first major passing study in 2010 was to divorce assists from “shot creation.” For example, Brevin Knight crushed MJ in assists, but Jordan created far more shots for teammates by causing the D to react. This led to the birth of BOX CREATION. 4/14

The key insight from box creation was that too much scoring cannibalizes chances for teammates (bc the defense reacts to the threat of a scorer with doubles and stunts) BUT, too little scoring and the defense won’t react. There’s a balance at the heart of offensive stardom. 5/14

The ability to quantify creation gave birth to a more comprehensive version of usage called OFFENSIVE LOAD. Load includes passing & creation, not just shots and turnovers, so it estimates a player’s total “direct involvement” in the offense. 6/14

In 2017 I started tracking detailed passing tendencies for players. It’s hard to reduce passing to one dimension or an overall score, but I think of it as the composite of all the possible types of passes. eg outlets, lobs, skip passes, touch passes, etc. 7/14

PASSER RATING is an attempt to measure this overall passing ability. Few if any excel in every component of passing, and time and circumstance will influence passing ability. The key insights of passer rating are: 8/14

First, a high ratio of assists to load is a major indicator of passing skill. The more a player accrues assists per involved-possessions, the more likely it is that he is finding the easiest shots for his teammates. 9/14

The second is obvious - layup assists are generally an indicator of good passing. They are the highest EV spot on the court and finding them regularly *as a percentage of one’s overall assists* is generally a positive. It indicates less dink n dunking to outside shooters. 10/14

There also seems to be a relationship between height and passing (cue @ZachLowe_NBA ). Specifically, when the other signals are strong and the player is tall, he is almost always an excellent passer. 11/14

I do think there’s a positional bias in the current version — guards pick up more idle assists — that I haven’t been able to adjust for but would like to in a future version. For now, I recommend some mental curving for tall, efficient guards who grade out well. 12/14

Finally, my aim was to create something historically retrospective. With the granularity of the new data we have today, I think "modern" versions of this could be even more precise/illuminating. Thanks for reading. Look forward to more conversations on this. Links below. 13/14

LINKS:
fansided.com/2017/08/11/nyl…
backpicks.com/2017/10/16/off…
backpicks.com/2018/07/15/nba…
fansided.com/2018/07/17/nyl…
14/14