APBRmetrics

Ok, I'm shelving the question of noise for now. Playing with the full dataset I think the shooting data shows clear validity. I'm much less certain about the assist and rebounding data, which appear to be quite noisy.

Here are the links to data in Tableau tables. I'm looking for help with the analysis, so I'll drop these links here, briefly talk about this season's data and let you have a look at the rest.

Shooting Mentality (behind/tied): https://public.tableausoftware.com/view ... y_count=no
The chart shows the shooting mentality of all NBA players who have taken >850 true shot attempts so far in 2014-2015. There's some interesting stuff here. For one, I love how I've got Jamal Crawford at the top at +9.5 and Lou Williams at the bottom at -7.4. It shows you can't really generalize about 6th-man gunners. I also find the contrast between Lebron's -3.8 and Kyrie's +4.3 noteworthy.

TS% change: https://public.tableausoftware.com/view ... y_count=no This chart shows a player's change in TS% in comebacks. This number is in percentage points and it is not adjusted for league average. There are some interesting connections here with the other charts. Blake's TS% goes down a tone in comebacks, but he also passes a hell of a lot more.

Passing Mentality (behind/tied): https://public.tableausoftware.com/view ... zHome=no#1
The chart shows the passing mentality of every player with >250 assists so far in 2014-2015. The sample sizes are smaller here, so I expect there are some issues with Blake's insane +16.3. That said, Mike Conley *doesn't* have any sample size problems and he's got an almost equally insane +13.3. Also noteworthy, Kyle Lowry's still sitting at the bottom of this chart. He's still one of the guys with one of the biggest splits between shooting mentality and passing mentality. John Wall's the opposite.

Rebounding Mentality (behind/tied): https://public.tableausoftware.com/view ... zHome=no#1
This one is new. It shows the rebound mentality of all players with >500 rebounds so far. I'm actually surprised to see any significant ahead/behind splits at all in rebounding, but there they are. You've got Duncan up there at +8.3 and Love trailing near the bottom at -9.1.

Finally, here's a tool you can use to find a player's career numbers in all three measures. https://public.tableausoftware.com/view ... zHome=no#1
I love this thing because it lets you look at career tendencies. Tim Duncan and Steve Nash, for instance, have both played every single full season of their careers with positive shooting mentalities. Nash even posted a ridiculous +16.3 in 06-07. Chris Paul and Tracy McGrady meanwhile posted negative shooting mentalities for most of their careers. When you play around with this tool, you'll notice right away the noisiness of the passing and rebounding numbers.

And, as a bonus. I've also used the same methods in close-score situations. So, here's a full set of the same links for player performance in scores within +/- 5 points.
Shooting Mentality (Close): https://public.tableausoftware.com/view ... y_count=no
Passing Mentality (Close): https://public.tableausoftware.com/view ... y_count=no
Rebound Mentality (Close): https://public.tableausoftware.com/view ... y_count=no
Player finder (Close): https://public.tableausoftware.com/view ... y_count=no

Anyway, what do you think?

*Edited to update the Tableau links

You're into some interesting stuff here.
When players don't like to shoot, or don't shoot well, when behind, are they also less likely to do well in close games?
You have trouble with <= data for players with good teams; but when these teams get to the playoffs, they're less "good". I wonder if the same players mentioned above also tend to falter in the playoffs.

I signed up for Tableau, but I get this message from those links above:

The interactive content that usually shows here could not be found. It may have been removed or renamed. Please check back soon.

Meanwhile, I did try to follow this thread, but I find it hard to read. I don't understand why anyone converts anything into percent format. Then you just have to multiply and divide by 100 all the time.

Does anyone really not know that .550 is the same as 55.0% ? Note that .550 doesn't have that extra character.
When a player whose Ast% is 10.0 goes up by 50%, is it now 15? Or is it 55? One can only guess.

That's an extreme example. But I am sure that when a player's eFG% goes from .500 to .550, some here would say it increased by 5%, and others would say it increased by 10%.
If you say it went up by .050, I believe you could not be misunderstood.

.. the percent difference in rebound rate while trailing vs. overall ...

TRb% (bb-ref terminology) is known as a %, and 10.0 is avg. When you refer to a "% difference", I think of it as a fraction of the given quantity -- in this case, a % of a %. But you may be referring to the fraction of 100%.

I'd really like to be able to comprehend what you've done, and in most lists you've shown, we 'get the idea' if not the quantities. Just prefer not to assume something that may be wrong.

Mike G wrote:You're into some interesting stuff here.
When players don't like to shoot, or don't shoot well, when behind, are they also less likely to do well in close games?
You have trouble with <= data for players with good teams; but when these teams get to the playoffs, they're less "good". I wonder if the same players mentioned above also tend to falter in the playoffs.

I signed up for Tableau, but I get this message from those links above:

The interactive content that usually shows here could not be found. It may have been removed or renamed. Please check back soon.

Meanwhile, I did try to follow this thread, but I find it hard to read. I don't understand why anyone converts anything into percent format. Then you just have to multiply and divide by 100 all the time.

Does anyone really not know that .550 is the same as 55.0% ? Note that .550 doesn't have that extra character.
When a player whose Ast% is 10.0 goes up by 50%, is it now 15? Or is it 55? One can only guess.

That's an extreme example. But I am sure that when a player's eFG% goes from .500 to .550, some here would say it increased by 5%, and others would say it increased by 10%.
If you say it went up by .050, I believe you could not be misunderstood.

.. the percent difference in rebound rate while trailing vs. overall ...

TRb% (bb-ref terminology) is known as a %, and 10.0 is avg. When you refer to a "% difference", I think of it as a fraction of the given quantity -- in this case, a % of a %. But you may be referring to the fraction of 100%.

I'd really like to be able to comprehend what you've done, and in most lists you've shown, we 'get the idea' if not the quantities. Just prefer not to assume something that may be wrong.

Thanks for the feedback. I've updated the Tableau links. Sorry about that. I was adjusting the titles and that broke the links. They should work now.

On percentages. I get your point. I expect I've been unclear. In the example you quote I said "percent difference," which is ambiguous. I'm using percent change not percentage point change unless I'm specifically stating otherwise.

Let me explain why, and see if you still think I'm doing it wrong. The reason I'm using percent change in looking at the rate stats (TSA rate, Assist Rate and Rebound rate) is because I'm adjusting for a significant positive league average. My goal with this is not to show what someone's assist rate actually becomes in comeback situations. I want to show what it becomes vs. what you'd expect it should be. This way, I can say that in 2006/7 Nash had a +16.3 shooting mentality, which tells you he shot 16.3% more often in comebacks than you would expect.

If I show this change in percentage points, and also correct for league average (which I plan to keep doing), I think people will expect you can add the base rates to the change to get the player's actually TSA rate in comebacks, which you can't. That's not the point of this.

That said. I am going to add commonly used efficiency stats to the tableau database. Here's an example: https://public.tableau.com/shared/X884R ... _count=yes In those cases I'm going to present those in percentage points rather than percent change. At the moment I am multiplying these values so that a value of 1.0 equals a the change from .50 to .51. The reason I do that is so that I can show 4 significant digits in a more readable form. I'll consider changing that if others agree with your objection.

These are awesome! Going to try and tie it in with some other data I have, great Tableau work too.