RealGM

A question for the advanced stats experts. We all know that assists are kind of random as a basic number. Also, some home scoring tables are quite liberal with what they consider an assist. I have personally seen (while watching the game and score keepers live) a pass on the 3pt line to a guy who then takes three (!) dribbles into the paint, including several spin and cross over moves, being counted as an assist. Naturally, this has caused an inflation in assist numbers which are then unfairly being compared to earlier eras of NBA basketball where the rules seemed to be much more strict (when i started watching, there was a maximum of one dribble after receiving the assist in place, otherwise it was an unsassisted basket). I believe the idea behind the number. the real art of the assist is is to account for the fact that one guy puts another guy into a position to score easily that he wouldnt have had otherwise; tough to write down but i have a feeling of what a "real" assist is vs. what a "stat" assist is.

My question: Is there a reliable independant statistic for "real" assists, and what are the criteria?

LesGrossman wrote:A question for the advanced stats experts. We all know that assists are kind of random as a basic number. Also, some home scoring tables are quite liberal with what they consider an assist. I have personally seen (while watching the game and score keepers live) a pass on the 3pt line to a guy who then takes three (!) dribbles into the paint, including several spin and cross over moves, being counted as an assist. Naturally, this has caused an inflation in assist numbers which are then unfairly being compared to earlier eras of NBA basketball where the rules seemed to be much more strict (when i started watching, there was a maximum of one dribble after receiving the assist in place, otherwise it was an unsassisted basket). I believe the idea behind the number. the real art of the assist is is to account for the fact that one guy puts another guy into a position to score easily that he wouldnt have had otherwise; tough to write down but i have a feeling of what a "real" assist is vs. what a "stat" assist is.

My question: Is there a reliable independant statistic for "real" assists, and what are the criteria?

I think passing it to someone who then attacks the close out should count as an assist

I think you are saying your subjective view of an assist is more "correct" than someone else's. Okay. We all tend to think we know best. But its still subjective, not "real".

I've found this video, its from 2018 but he goes into detail and comes to the same conclusions (home cooking, very loose rule set leading to basically inflated and meaningless numbers):



And no, i dont think my view is neccessarily the right one, just that the rules have been loosened up to a travesty when looking at what counts as an assist these days (especially at home). We cant really tell who the best passing players are from looking at assist numbers because the way they are counted is designed to create spectacular numbers and the illusion that we see better players today than in the past.

Texas Chuck wrote:I think you are saying your subjective view of an assist is more "correct" than someone else's. Okay. We all tend to think we know best. But its still subjective, not "real".

Just to clarify, i was asking wether theres advanced statistics that might be more accurate than just the very rough and arbitrary assist number, i wasnt lobbying to change the way box score assists are counted - wont happen anyway. Kind of like stats for contested vs. uncontested rebounds, woudl you agree it makes a huge difference wether you fight and win a tough contested offensive board vs. just picking up the missed free throw when no opponent is within 10 ft? We saw you can get MVP by manipulating those simple stats...

LesGrossman wrote:

Texas Chuck wrote:I think you are saying your subjective view of an assist is more "correct" than someone else's. Okay. We all tend to think we know best. But its still subjective, not "real".

Just to clarify, i was asking wether theres advanced statistics that might be more accurate than just the very rough and arbitrary assist number. Kind of like stats for contested vs. uncontested rebounds, woudl you agree it makes a huge difference wether you fight and win a tough contested offensive board vs. just picking up the missed free throw when no opponent is within 10 ft? We saw you can get MVP by manipulating those simple stats...

I think individual rebound numbers are far less important than team rebounding numbers particularly on the defensive end. DeAndre Jordan and Andre Drummond aren't better rebounders than Nene or the Lopez twins despite the individual totals suggesting they are.

It's a team endeavor on the defensive glass and always has been.

I think potential assists is a better stat than assists anyway. Because its independent of if your teammates finishes the play. If what you are wanting to measure is the quality/volume of the creation ability of a player I would start there. Actual assists are subject to shooting variance.

Some places have 'at-the rim assists', which adds another useful layer.

Texas Chuck wrote:

LesGrossman wrote:

Texas Chuck wrote:I think you are saying your subjective view of an assist is more "correct" than someone else's. Okay. We all tend to think we know best. But its still subjective, not "real".

Just to clarify, i was asking wether theres advanced statistics that might be more accurate than just the very rough and arbitrary assist number. Kind of like stats for contested vs. uncontested rebounds, woudl you agree it makes a huge difference wether you fight and win a tough contested offensive board vs. just picking up the missed free throw when no opponent is within 10 ft? We saw you can get MVP by manipulating those simple stats...

I think individual rebound numbers are far less important than team rebounding numbers particularly on the defensive end. DeAndre Jordan and Andre Drummond aren't better rebounders than Nene or the Lopez twins despite the individual totals suggesting they are.

It's a team endeavor on the defensive glass and always has been.

I think potential assists is a better stat than assists anyway. Because its independent of if your teammates finishes the play. If what you are wanting to measure is the quality/volume of the creation ability of a player I would start there. Actual assists are subject to shooting variance.

That is another aspect that needs to be addressed separately. I think in europe they experimented with free throws for a while (if assist lead to a foul and free throws and one was made, the assist was awarded). My point is more a far more narrow definition of the assist as compared to a random pass somewhere at half court that does not help the scoring guy at all. Its really hard to define the assist, if you check out the video and the very fuzzy definition of the league itself you'll know what i mean. But i would like to know who actually puts his team mates into easy scoring opportunities the most time (uncontested or at least a mismatch).

I think Elgee's passer rating index is a pretty good barometer for playmaking/passing in place of assists.

I think a metric which incorporates some sort of % as a form of ts% into the shot created with a pass would be a good metric to have. As in if a guy passes to someone for an open 3 it would calculate the likelihood of that pass being converted into a 3 based on league averages(so maybe it's worth 1.6 pts) while a pass for an open dunk would be worth 2.0 points or close to it.

LesGrossman wrote:A question for the advanced stats experts. We all know that assists are kind of random as a basic number. Also, some home scoring tables are quite liberal with what they consider an assist. I have personally seen (while watching the game and score keepers live) a pass on the 3pt line to a guy who then takes three (!) dribbles into the paint, including several spin and cross over moves, being counted as an assist. Naturally, this has caused an inflation in assist numbers which are then unfairly being compared to earlier eras of NBA basketball where the rules seemed to be much more strict (when i started watching, there was a maximum of one dribble after receiving the assist in place, otherwise it was an unsassisted basket). I believe the idea behind the number. the real art of the assist is is to account for the fact that one guy puts another guy into a position to score easily that he wouldnt have had otherwise; tough to write down but i have a feeling of what a "real" assist is vs. what a "stat" assist is.

My question: Is there a reliable independant statistic for "real" assists, and what are the criteria?

I'm not aware of anything that represents a clearly superior approach to assist-tracking than the NBA for all NBA games, but I will say that when Ben Taylor was first getting started as an independent analyst, he did a lot of hand tracking and one of the stats he tracked was something he called Opportunities Created (OC).

The idea there is that a) his standard for actual assists was higher than the NBA, but b) he didn't focus on whether the shot actually went in. He was looking at what the playmaker was doing and judging whether the player was truly creating opportunities for others.

From there he ended up using regression to create a stat that let him estimate historically how many opportunities players were creating (Box Creation) BC.

I'm not going to post a bunch of details here but I will say when I look now at leaders in BC by season, here's what I see:

1. Westbrook '16-17
2. Harden '17-18
3. Nash '06-07
4. Nash '09-10
5. Nash '07-08

Others high on the list: Mark Price, John Stockton, Magic Johnson, Steph Curry, Chris Paul, LeBron James

I'd note that I wouldn't exactly call Ben super-high on Westbrook or Harden in general. This isn't a statement of who was a better passer or playmaker but a statement about how much opportunity creation is falling on one man, and so it shouldn't be a surprise that helios and classic floor generals rank high.

All of this stuff is beyond the Patreon wall at this time, so you'd need to pay for that.

In the end, I trust Ben's eyes for this analysis considerably more than a random scorekeeper, but he can't do it for every game, and as his work has taken off, he's more focused on watching games to create videos to explain on a team-level what's happening than he is to make stats.

LesGrossman wrote:A question for the advanced stats experts. We all know that assists are kind of random as a basic number. Also, some home scoring tables are quite liberal with what they consider an assist. I have personally seen (while watching the game and score keepers live) a pass on the 3pt line to a guy who then takes three (!) dribbles into the paint, including several spin and cross over moves, being counted as an assist. Naturally, this has caused an inflation in assist numbers which are then unfairly being compared to earlier eras of NBA basketball where the rules seemed to be much more strict (when i started watching, there was a maximum of one dribble after receiving the assist in place, otherwise it was an unsassisted basket). I believe the idea behind the number. the real art of the assist is is to account for the fact that one guy puts another guy into a position to score easily that he wouldnt have had otherwise; tough to write down but i have a feeling of what a "real" assist is vs. what a "stat" assist is.

My question: Is there a reliable independant statistic for "real" assists, and what are the criteria?

Yes. It's called Oppurtunies created.

You can also expand it qualitatively to
-> great oc's
-> middling oc's
-> weak oc's

By similar bball reference, ast%:tov% is better than assists anyway

the "real" criteira is impact on teammate effiency

MyUniBroDavis wrote:

LesGrossman wrote:A question for the advanced stats experts. We all know that assists are kind of random as a basic number. Also, some home scoring tables are quite liberal with what they consider an assist. I have personally seen (while watching the game and score keepers live) a pass on the 3pt line to a guy who then takes three (!) dribbles into the paint, including several spin and cross over moves, being counted as an assist. Naturally, this has caused an inflation in assist numbers which are then unfairly being compared to earlier eras of NBA basketball where the rules seemed to be much more strict (when i started watching, there was a maximum of one dribble after receiving the assist in place, otherwise it was an unsassisted basket). I believe the idea behind the number. the real art of the assist is is to account for the fact that one guy puts another guy into a position to score easily that he wouldnt have had otherwise; tough to write down but i have a feeling of what a "real" assist is vs. what a "stat" assist is.

My question: Is there a reliable independant statistic for "real" assists, and what are the criteria?

I think passing it to someone who then attacks the close out should count as an assist

only if you draw extra attention.

passing to someone else who ends up doing the work is just replacement value isn't it

OhayoKD wrote:

MyUniBroDavis wrote:

LesGrossman wrote:A question for the advanced stats experts. We all know that assists are kind of random as a basic number. Also, some home scoring tables are quite liberal with what they consider an assist. I have personally seen (while watching the game and score keepers live) a pass on the 3pt line to a guy who then takes three (!) dribbles into the paint, including several spin and cross over moves, being counted as an assist. Naturally, this has caused an inflation in assist numbers which are then unfairly being compared to earlier eras of NBA basketball where the rules seemed to be much more strict (when i started watching, there was a maximum of one dribble after receiving the assist in place, otherwise it was an unsassisted basket). I believe the idea behind the number. the real art of the assist is is to account for the fact that one guy puts another guy into a position to score easily that he wouldnt have had otherwise; tough to write down but i have a feeling of what a "real" assist is vs. what a "stat" assist is.

My question: Is there a reliable independant statistic for "real" assists, and what are the criteria?

I think passing it to someone who then attacks the close out should count as an assist

only if you draw extra attention.

passing to someone else who ends up doing the work is just replacement value isn't it

We will probably not heal the simple assist defintion which would include that scenario, but in a more meaningful sense that shouldnt count as an assist, because its not really assisting. The value added by the assist is the open scoring opportunity for someone else; there has to be a difference between dribbling to the defense, and then beating the defense (no assist) vs. getting a pass and then beating the defense.

LesGrossman wrote:A question for the advanced stats experts. We all know that assists are kind of random as a basic number. Also, some home scoring tables are quite liberal with what they consider an assist. I have personally seen (while watching the game and score keepers live) a pass on the 3pt line to a guy who then takes three (!) dribbles into the paint, including several spin and cross over moves, being counted as an assist. Naturally, this has caused an inflation in assist numbers which are then unfairly being compared to earlier eras of NBA basketball where the rules seemed to be much more strict (when i started watching, there was a maximum of one dribble after receiving the assist in place, otherwise it was an unsassisted basket). I believe the idea behind the number. the real art of the assist is is to account for the fact that one guy puts another guy into a position to score easily that he wouldnt have had otherwise; tough to write down but i have a feeling of what a "real" assist is vs. what a "stat" assist is.

My question: Is there a reliable independant statistic for "real" assists, and what are the criteria?

Several spin moves and crossovers on three dribbles? That must be one of the greatest drives of all-time so I agree that shouldn't have been awarded as an assist. In general however I think a liberal standard for scoring assists is appropriate.

rand wrote:

LesGrossman wrote:A question for the advanced stats experts. We all know that assists are kind of random as a basic number. Also, some home scoring tables are quite liberal with what they consider an assist. I have personally seen (while watching the game and score keepers live) a pass on the 3pt line to a guy who then takes three (!) dribbles into the paint, including several spin and cross over moves, being counted as an assist. Naturally, this has caused an inflation in assist numbers which are then unfairly being compared to earlier eras of NBA basketball where the rules seemed to be much more strict (when i started watching, there was a maximum of one dribble after receiving the assist in place, otherwise it was an unsassisted basket). I believe the idea behind the number. the real art of the assist is is to account for the fact that one guy puts another guy into a position to score easily that he wouldnt have had otherwise; tough to write down but i have a feeling of what a "real" assist is vs. what a "stat" assist is.

My question: Is there a reliable independant statistic for "real" assists, and what are the criteria?

Several spin moves and crossovers on three dribbles? That must be one of the greatest drives of all-time so I agree that shouldn't have been awarded as an assist. In general however I think a liberal standard for scoring assists is appropriate.

You dont really need much imagination, have you watched that video? There are several examples.

If as you say, a "liberal standard" is appropriate, what can you really derive out of that number when trying to figure out who really helps his team the most by setting up, identifying and executing really "helpful assists"? They disappear among the vast number of random passes imho. According to those numbers, Russ and Harden seem to be the best at finding the ideal pass...eye test suggests otherwise though.

I think the only way to track your subjective "real" assist is to watch every game and score them yourself