Shooting variance

[qianlong]

I was wondering about streaky players, and in particular streaky 3 point shooters. In statistics of other kinds averages are often followed by variance or standard deviation. I think it would be usefull in this case. Obviously if you can watch enough games you could realize if a player is a hot/cold player or a regular player, but if you are looking for players on other teams it could be usefull.

However i am not sure which would be the best statistic, variance of made? of attempts? some kind of ratio? Please help me to figure it out or suggest me if there is something similar on the web.

[Doctor MJ]

Debatable what's best, but to start out with it'd be great to just see variance for FG% & TS%.

My guess would be that for players receiving comparable attention from the defense, and shooting at similar volume, FG% variance largely comes how far away you tend to shoot the ball.

[qianlong]

Debatable what's best, but to start out with it'd be great to just see variance for FG% & TS%.

Totaly agree.

My guess would be that for players receiving comparable attention from the defense, and shooting at similar volume, FG% variance largely comes how far away you tend to shoot the ball.

I disagree on this, or better i agree in general, but that's what i specifically want to emphasize. This is true for most players, but there are some players that are just hot/cold. For this players i expect a greater variance and is not depending on harder shots but on off nights. There are some players that often go 0/6 or 5/6 or at least it seems to me. Having data to prove or better discover it would be nice.

Thank you for the reply.

[Doctor MJ]

Debatable what's best, but to start out with it'd be great to just see variance for FG% & TS%.

Totaly agree.

My guess would be that for players receiving comparable attention from the defense, and shooting at similar volume, FG% variance largely comes how far away you tend to shoot the ball.

I disagree on this, or better i agree in general, but that's what i specifically want to emphasize. This is true for most players, but there are some players that are just hot/cold. For this players i expect a greater variance and is not depending on harder shots but on off nights. There are some players that often go 0/6 or 5/6 or at least it seems to me. Having data to prove or better discover it would be nice.

Thank you for the reply.

Well, there's been a good amount of analysis on "the hot hand" in high levels of competitive basketball, and every study I've seen says it doesn't exist. It's just normal randomness. (Google "hot hand" basketball)

Now, I don't buy this totally because I know I've seen players begin to hesitate when they lose confidence - but among these top players, such things may very well be quite rare, which would agree with the studies.

So, I would agree with you that if we found really good evidence that certain players are inherently more variable than others that would be great, but I wouldn't expect that to happen.

[azuresou1]

I don't think it's statistically possible to prove being on fire, but anyone who's ever played knows that it exists. When you've missed a few in a row, you become tentative and hesitant, and you give the defense that extra split second to close in, making you even more hesitant since you now have a more contested shot. Then you start thinking about how much power/touch you need to put into your shot, and if you do take the shot, it's just not a great shot in general.

Whereas when you're on fire, you're ready to pop that shot as soon as the ball touches your hands, and you don't worry about the defender at all, and all your attention is on the bucket, and you just "do." I mean, I've had games where I've gone probably 85% from a college 3, where I'd fire it over taller guys and it'd just splash in.

[nonemus]

I did some research on this earlier:

So, I was thinking, everyone keeps talking about efficiency when talking about a player's offensive capability. Kobe is know to have many high scoring games, but then I thought, why is it that his TS% and PPG are lower than the other's in the elite scorer club? I mean, say there was a 30ppg player and a 28ppg player. If the majority of games of the 30ppg player fall between 10-50 points, but the 28ppg player is consistently around 23-33ppg, the second player has the clear edge.

So, thats what inspired me to do the following, to find out which of the NBA's best scorers is the most consistent.

The 5 players who will be compared are as follows:

1. Durant
2. Kobe
3. LeBron
4. Anthony
5. Wade

All of the values are from the '09-'10 seasons of the aforementioned players and were retrieved from basketball-reference.com.

The methodology is simple, players who are more consistent will have a smaller variance and thus will have a smaller range in which 84% of their scores (1 standard deviation plus and minus away from the average) fall into.

For example, if Player A had a average of 25 ppg and a SD of 5 ppg, 84% of their scores would fall between 20-30 points.

So, here is what I had found:
---------------------------------------------------------------------------------------------------------------------------------
Kevin Durant

Code: Select all

Regular Season (82 games): 30.1 ppg with a SD of 6.8; TS% of .607 with a SD of .111
Playoffs (6 games): 25.0 ppg with a SD of 5.3; TS% of .499 with a SD of .086

---------------------------------------------------------------------------------------------------------------------------------
Kobe Byrant

Code: Select all

Regular Season (73 games)*: 27.0 ppg with a SD of 8.9; TS% of .545 with a SD of .126
Playoffs (23 games): 29.2 ppg with a SD of 7.9; TS% of .567 with a SD of .100

*Played with an injured finger.

---------------------------------------------------------------------------------------------------------------------------------
Lebron James

Code: Select all

Regular Season (76 games): 29.7 ppg with a SD of 8.15; TS% of .604 with a SD of .09
Playoffs (11 games): 29.1 ppg with a SD of 8.94; TS% of .607 with a SD of .174

---------------------------------------------------------------------------------------------------------------------------------
Carmelo Anthony

Code: Select all

Regular Season (69 games): 28.2 ppg with a SD of 7.6 '; TS% of .548 with a SD of .102
Playoffs (6 games): 30.7 ppg with a SD of 8.6; TS% of .564 with a SD of .140

---------------------------------------------------------------------------------------------------------------------------------
Dywane Wade

Code: Select all

Regular Season (77 games): 26.6 ppg with a SD of 7.7; TS% of .562 with a SD of .113
Playoffs (6 games): 33.2 ppg with a SD of 7.7; TS% of .650 with a SD of .081

---------------------------------------------------------------------------------------------------------------------------------

Some observations:
a) Of course, Kobe's playoff data stands for more than the rest due to the size of his sample pool.
b) In the Regular Season, LBJ is by far the most consistently efficient scorer.
c) Although LBJ posted up great numbers in the '09-'10 playoff run, he was highly inconsistent. This inconsistency caused his team many games.
d) Durant, although a close second for have the best consistency in the regular season behind LBJ, is had very good consistency in the playoffs. The thing is, he was consistently bad.
e) Wade's performance against Boston was phenomenal. Yeah, they probably weren't trying as hard as when they were in the ECF or in the Finals, but Boston regardless is a top notch defensive team.

Of course, consistency is only one part of the equation of who is the better offensive player, but it is an important one. If you are consistent, it means that you are reliable, and this wins games, and many times tells more than a whole average vs average argument.

So, given the stats, who would you want on your team from a purely scoring perspective ( meaning not including passing/assists) for the Playoffs/Regular Season?

My list looks like this:

Regular Season:

1. LBJ
2. Durant
3. Wade
4. Melo
5. Kobe

Playoffs

1. Kobe/Wade (Probably Kobe)
2. Kobe/Wade
3. Melo
4. LBJ
5. Durant

The reason it's Kobe/Wade is due to

a) Kobe went up against much more aggressive defenses

and

b) Wade's small pool of data (5 games)

[Doctor MJ]

I don't think it's statistically possible to prove being on fire, but anyone who's ever played knows that it exists. When you've missed a few in a row, you become tentative and hesitant, and you give the defense that extra split second to close in, making you even more hesitant since you now have a more contested shot. Then you start thinking about how much power/touch you need to put into your shot, and if you do take the shot, it's just not a great shot in general.

Whereas when you're on fire, you're ready to pop that shot as soon as the ball touches your hands, and you don't worry about the defender at all, and all your attention is on the bucket, and you just "do." I mean, I've had games where I've gone probably 85% from a college 3, where I'd fire it over taller guys and it'd just splash in.

Oh the thing the stat guys look for is simple and I think dead on: Does making your last shot make you more likely to make your next than when you missed your last shot? And over again, they come back saying it doesn't at the NBA level.

I agree that because of the hesitation factor, there is no doubt that the hot hand exists in some form, but it's entirely plausible that at the highest level in the game, players aren't often so frazzled to make a difference large enough to show any correlation.

Still, you see Kobe go for 81 points, and it's hard to believe there wasn't something different about him that day.

[Doctor MJ]

nonemus, that's very cool stuff.

One thing strikes me as odd: I'd expect the SD for the playoffs would be much bigger than in the regular season and it isn't. Thoughts on that?

[azuresou1]

I think that's too simple of an analysis though. If you make a few shots in a row, you're more likely to take worse quality shots. This is why we see Kobe hit 3 in a row, and then toss up a contested fadeaway 3. Hell, there's even a term for these things (heat checks).

Conversely, if players haven't been making their shots, most players will pass up shots they would normally take and work a bit harder to get better quality looks. Instead of taking an off-the-dribble three, they'll instead swing it around and wait until they can curl off a screen and get an open mid-range jumper.

I think the better the caliber of player, the less hesitation plays a role. However, simultaneously, the better the caliber of player, the more heavily shot quality degradation comes into play.

http://www.youtube.com/watch?v=goAXMXebqtc

[Nivek]

I agree that the hot hand exists because I've felt it. Whether the patterns of misses and makes deviate from what the numbers would predict -- probably not. That's not proof that there's no hot hand, however. There used to be (might still be, actually) a lot of debate among baseball stat guys about the existence of clutch hitting. Many studies proved there was no such thing. Finally, Bill James said wait a minute -- we're searching for "clutchness" (not his word) by subdividing to a small sample size and then deriving numbers from there. He posited that perhaps ALL pitcher-batter confrontations were measures of clutchness -- that guys don't get to the major leagues without the ability to come through in high-pressure situations. Or something like that -- I'm paraphrasing/summarizing from something I read probably 4 years ago.

My point is that "hot hand" in basketball may be like that.

As for nonemus's work -- it's interesting. I took a similar approach several years ago looking at consistency across an array of statistical categories. I ran the numbers on many, many players. What I did was create a consistency score that was SD/AVG. So, if we're looking at Brendan Haywood's rebounding consistency for 05-06 (to choose a player at random from my spreadsheet) -- Haywood averaged 11 rebounds per 40 minutes. The standard deviation was 3.99. So it's 3.99/11.0 = 0.36, or 36%. That season, Haywood's per minute rebounding varied game-to-game by 36%.

0% means no variation -- perfectly consistent. The higher the number, the greater the variation. That season, Haywood was the Wizards 2nd most consistent rebounder behind Jamison's 35%.

[qianlong]

Great work nonemus very telling

One thing strikes me as odd: I'd expect the SD for the playoffs would be much bigger than in the regular season and it isn't. Thoughts on that?

Maybe the pace reduction in playoff could reduce variability, could it be? according to nbastuffer the pace was 93.1 in regular season and 90.5 in post season

[jman3134]

I don't think it's statistically possible to prove being on fire, but anyone who's ever played knows that it exists. When you've missed a few in a row, you become tentative and hesitant, and you give the defense that extra split second to close in, making you even more hesitant since you now have a more contested shot. Then you start thinking about how much power/touch you need to put into your shot, and if you do take the shot, it's just not a great shot in general.

Whereas when you're on fire, you're ready to pop that shot as soon as the ball touches your hands, and you don't worry about the defender at all, and all your attention is on the bucket, and you just "do." I mean, I've had games where I've gone probably 85% from a college 3, where I'd fire it over taller guys and it'd just splash in.

Oh the thing the stat guys look for is simple and I think dead on: Does making your last shot make you more likely to make your next than when you missed your last shot? And over again, they come back saying it doesn't at the NBA level.

I agree that because of the hesitation factor, there is no doubt that the hot hand exists in some form, but it's entirely plausible that at the highest level in the game, players aren't often so frazzled to make a difference large enough to show any correlation.

Still, you see Kobe go for 81 points, and it's hard to believe there wasn't something different about him that day.

I don't see why the players would not be frazzled enough to show a difference in correlation. I would be more interested to view the statistical breakdown for NBA stars as opposed to individuals either coming off the bench or taking on a more limited team role. (either as a supporting starter or a backup) I think that measuring the hot hand effect has more to do with the sample of players you use than anything.

Think about the human nature of a star vs. a bench player. One bad shot will get one player tossed out of a game, whereas another player might have free reign to shoot whenever he wants with limited repercussions. This is why, imo, there are such things as system players and strategic team fits- when a 15ppg Mike James moves from the Raptors, he might not even be guaranteed a roster spot. This must impact one's in-game playing tactics, and also explains why it is so easy to slip out of the NBA.