Accuracy of assists

[rrravenred]

Just having a minor muse about this... how accurate do people feel that assist numbers are in describing someone's passing abilities?

On the Bucks board, we've had a succession of point guard prospects and some of the posters half-jokingly came up with the "almost-assist" where a player is given a perfect pass in good scoring position, but somehow bobbles the ball or misses the easy shot.

In that case, although the pass itself was "Made FG-quality", no assist could (or should) be recorded. Conversely, if a horrible floating pass is somehow caught and an awful off-balance shot is made in spite of the quality of the pass, an assist gets recorded.

I suppose the question I'm asking is how much validity (in a statistical sense) people feel an assist has and whether or not there is a better way of measuring passing and ball movement...

[mysticbb]

In basketball there is nothing like a "hockey assist", which would give a better impression about the passing ability. Anyway, 82games.com counts all the passing turnovers, relate that to the amount of assists and you get a better measurement in the first place. If you assume that the sample size can be big enough that the percentage of situations, in which a player plays a perfect pass, but the teammate wasn't able to make the open shpt, are roughly the same for every player, assists per passing turnover can be a good indicator for the passing abilities of the players. But this stat doesn't account for the system. If a offense is using pick&roll more often, it will create more assists, than a offense which uses only screens to get open shots for the ball handler or a drive to the basket. That can have a negative effect for the assists per passing turnover for a player.

So, there is no perfect measurement for passing abilities, the best value is the assists per passing turnover.

[rrravenred]

A/TO is my favourite stat w/r/t passing (especially when differentiated by class of turnover), but as I said, it's not ideal. Hell, I'd probably score an NBA assist if I played on a court with Peak Shaq, Peak Kobe, Lebron and Peak Ray Allen. That assist would have absolutely no reference to my passing ability (which would be, from a selfish perspective, a good thing ). The credit for the assist would entirely belong to me, without reference to the quality of my teammates

If it were possible to defining assists as a pass to a player which leads to a shot and then take it from there, and then split it into made shots (possibly also splitting it into 2p Fg, 3P FG and shooting fouls) you could then look at:

- percentage of made shots from assists
- adjusted FG differential for the player taking the shot (from their average)
- points per unadjusted assist

Of course, with this it's not really possible to isolate individual player-to-player relationships (at least not without some extraordinarily messy record-keeping), but it's still a bit more fine-grained...

[mysticbb]

A/TO is my favourite stat w/r/t passing (especially when differentiated by class of turnover), but as I said, it's not ideal. Hell, I'd probably score an NBA assist if I played on a court with Peak Shaq, Peak Kobe, Lebron and Peak Ray Allen. That assist would have absolutely no reference to my passing ability (which would be, from a selfish perspective, a good thing ;)).

Either you are a very good college player or you are underestimate the defensive abilities of the most players. They probably will force a couple of turnovers, therefore I guess your assists to turnover ratio will be pretty bad. ;)
Anyway, I got what you meant and I agree with you that A/To isn't perfect. Well, you will hardly find a perfect stat anyway.

If it were possible to defining assists as a pass to a player which leads to a shot and then take it from there, and then split it into made shots (possibly also splitting it into 2p Fg, 3P FG and shooting fouls) you could then look at:

- percentage of made shots from assists
- adjusted FG differential for the player taking the shot (from their average)
- points per unadjusted assist

Of course, with this it's not really possible to isolate individual player-to-player relationships (at least not without some extraordinarily messy record-keeping), but it's still a bit more fine-grained...

Not a bad idea. That would give at least a more accurate value for the passing abilities.

[Doctor MJ]

Just having a minor muse about this... how accurate do people feel that assist numbers are in describing someone's passing abilities?

On the Bucks board, we've had a succession of point guard prospects and some of the posters half-jokingly came up with the "almost-assist" where a player is given a perfect pass in good scoring position, but somehow bobbles the ball or misses the easy shot.

In that case, although the pass itself was "Made FG-quality", no assist could (or should) be recorded. Conversely, if a horrible floating pass is somehow caught and an awful off-balance shot is made in spite of the quality of the pass, an assist gets recorded.

I suppose the question I'm asking is how much validity (in a statistical sense) people feel an assist has and whether or not there is a better way of measuring passing and ball movement...

There are a lot of problems with using assists as the one number to rate a player's passing ability. To me the biggest is this: A great floor general has major impact on the vast majority of his team's possessions, but even the most prolific assist getters only get assists on 10-15% of those possessions. It's silly to put too much stock in a stat that essentially pretends such a player is twiddling his thumbs the vast majority of the time. Doesn't mean I don't look at the assist stat, I'd just never make a judgment only on that.

One other note, when I saw the title of the thread I thought you'd be talking about the subjective nature of scoring assists. Unlike the other major stats, the definition of an assist is not very clear. The result is that every team's scorekeeping has different standards, and the variance is significant.

The really frustrating part of it all, is that when statisticians do analysis of point guards, they focus on assists, and because of all the fuzziness in assists, and correlation between assists and success is not very clear, they tend to underrate point guards. It's one of the reasons I like +/- stats. Top point guards tend to really stand out there, which is more in line with how I see the game.

[mysticbb]

but even the most prolific assist getters only get assists on 10-15% of those possessions. It's silly to put too much stock in a stat that essentially pretends such a player is twiddling his thumbs the vast majority of the time.

Steve Nash gets an assists for 18% of the team's possessions in this season. If you only count the scoring possessions, it is 35% (which includes his own scoring possessions), and he gets an assists for 54% of his teammates field goals.

The really frustrating part of it all, is that when statisticians do analysis of point guards, they focus on assists, and because of all the fuzziness in assists, and correlation between assists and success is not very clear, they tend to underrate point guards.

If the comparison is made just between point guards, using assists isn't that bad at all.

It's one of the reasons I like +/- stats. Top point guards tend to really stand out there, which is more in line with how I see the game.

Really? In the last 10 years the players which really stand out in the adj. +/- were Kevin Garnett, Tim Duncan, Dirk Nowitzki, LeBron James and Dwyane Wade. The first point guard is Chris Paul listed at 6th (http://waynewinston.com/wordpress/?p=158). So, the top players will be on Top of the +/- list, that doesn't really depend on the position (well, the 3 of the 4 best were 3 PF).

[rrravenred]

That sort of says to me that the value of a good point guard is overrated in terms of contributions to offensive efficiency. An interesting measure could be done by looking at things like screens set per possession or successful post-passes (both "out" and "in"), possibly also factoring in defeating the double-team without trying to directly correlate that with a made bucket. Looking at those measures would yiield some interesting data about who does good "work" and what correlation that has with improved offensive efficiency.

Having said that, Lebron and Wade are ersatz PGs by role.

[Doctor MJ]

Steve Nash gets an assists for 18% of the team's possessions in this season. If you only count the scoring possessions, it is 35% (which includes his own scoring possessions), and he gets an assists for 54% of his teammates field goals.

Phoenix averages 96.8 possessions per game, Nash averages 11.8 assists. That's 12.2% of the team's possessions. But even if the answer were 18%, that's a small difference. Most possessions, he's not getting credit for an assist, but he's very actively involved in orchestrating the offense pretty much every possession.

If the comparison is made just between point guards, using assists isn't that bad at all.

Well, sure, when it comes to the factor that underrates point guards, comparisons between point guards isn't an issue. That goes without saying. Course, my other points still stand, and they do affect comparisons between point guards.

Really? In the last 10 years the players which really stand out in the adj. +/- were Kevin Garnett, Tim Duncan, Dirk Nowitzki, LeBron James and Dwyane Wade. The first point guard is Chris Paul listed at 6th (http://waynewinston.com/wordpress/?p=158). So, the top players will be on Top of the +/- list, that doesn't really depend on the position (well, the 3 of the 4 best were 3 PF).

I think what I said was unclear before. What I meant was that point guards are reasonable represented +/- in a way they typically are not by PER-style stats.

To respond to Winston's list: He really constructed it horrendously. I've made comments before. It's shocking that he'd make a list of the best players of the decade without factoring in how much players played.

Here's what I'll say:
-Kidd's been a consistent +/- leader. I'd be shocked if any decade list properly weighted didn't have Kidd in the top 10.
-Nash has been a dominant +/- performer in Phoenix, which is the only time he's been considered a dominant. If Winston made a list of the past 5 years, Nash would easily be on the list.
-Stockton and Paul wouldn't be on a properly weight list, but obviously for the time they played, they did well by this metric.

So it's pretty clear, not that point guards dominate +/-, but that they are represent well by it.

[Doctor MJ]

That sort of says to me that the value of a good point guard is overrated in terms of contributions to offensive efficiency. An interesting measure could be done by looking at things like screens set per possession or successful post-passes (both "out" and "in"), possibly also factoring in defeating the double-team without trying to directly correlate that with a made bucket. Looking at those measures would yiield some interesting data about who does good "work" and what correlation that has with improved offensive efficiency.

Having said that, Lebron and Wade are ersatz PGs by role.

Are you basing this off the Winston stuff? Don't. First, see my previous response. Then remember that +/- is both offense and defense. Take a look at the other thread on this board about adjusted +/-. It's got Nash rated as the #1 offensive player over the last 6 years.

[Roger Murdock]

Assists are impossible to measure accurately. A pass that leads to a wide open dunk is sooooo much better than a pass that leads to an open jumper.

[rrravenred]

Are you basing this off the Winston stuff? Don't. First, see my previous response. Then remember that +/- is both offense and defense. Take a look at the other thread on this board about adjusted +/-. It's got Nash rated as the #1 offensive player over the last 6 years.

Based it on the above post, nothing else.

I still hold that Adjusted +/- is very noisy, but it's certainly a basis for comparison.

[Doctor MJ]

Are you basing this off the Winston stuff? Don't. First, see my previous response. Then remember that +/- is both offense and defense. Take a look at the other thread on this board about adjusted +/-. It's got Nash rated as the #1 offensive player over the last 6 years.

Based it on the above post, nothing else.

I still hold that Adjusted +/- is very noisy, but it's certainly a basis for comparison.

Yup, the noisiness is a problem. Not a reason to not use the metric, but definitely a reason not to use the metric exclusively.

[TheSecretWeapon]

Doctor MJ touches on the biggest issue when it comes to assists -- they're subjective and biased. When I was tracking defense, I worked by taking notes directly on a printout of the official play-by-play. Out of curiosity, I started scrutinizing assists. It was a mini-study (only about 10 games and approximately 450 total assists), but I found assists over-awarded by 15-20%.

There was definite bias. PGs and "stars" received assists for some passes whereas lesser-knowns and big men making a nearly identical pass did not recieve an assist.

For example, two plays: Play One, Jason Kidd made a simple ball rotation pass. Caught the ball at the top of the circle and moves it to the next guy on the wing. That wing player fakes a jumper, dribble drives down into the lane (two dribbles) and finishes with a dunk. Assist, Kidd.

Play Two: Same rotation pass, same wing player, same ball fake (different defender), same dribble down the lane, another finish with a dunk. This time the rotation pass was made by Mikki Moore. And this time, no assist.

Many other plays like this.

And it's similar for steals, too.

[Mamba Venom]

Scorekeepers admit to running up assist numbers