Interesting article on Nate Silver's website

[AccuRater]

http://fivethirtyeight.com/features/the-hidden-value-of-the-nba-steal/

The author's (Benjamin Morris) data supports the notion that the number of steals are the most important statistic contributing to offensive performance. The data are pretty dramatic. Here's a quick summary:

Rank Order of Box Score Stats relative to Points Per Game

1. Steals
2. Blocks
3. Turnovers
4. Assists
5, Rebounds

That is, among the listed stats, Steals contribute the most to total points per game while Rebounds contribute the least.

A suggestion: Don't dismiss it until you read it.

 

[skurey]

it's a terrible article

 

[AccuRater]

it's a terrible article

Now there's a well reasoned response. I guess that settles it.

 

[Millhouse]

http://fivethirtyeight.com/features/the-hidden-value-of-the-nba-steal/

The author's (Benjamin Morris) data supports the notion that the number of steals are the most important statistic contributing to offensive performance. The data are pretty dramatic. Here's a quick summary:

Rank Order of Box Score Stats relative to Points Per Game

1. Steals
2. Blocks
3. Turnovers
4. Assists
5, Rebounds

That is, among the listed stats, Steals contribute the most to total points per game while Rebounds contribute the least.

A suggestion: Don't dismiss it until you read it.

not buying it

steals might be a big bucket that captures every other omitted variable

 

[OttoinGrotto]

http://fivethirtyeight.com/features/the-hidden-value-of-the-nba-steal/

The author's (Benjamin Morris) data supports the notion that the number of steals are the most important statistic contributing to offensive performance. The data are pretty dramatic. Here's a quick summary:

Rank Order of Box Score Stats relative to Points Per Game

1. Steals
2. Blocks
3. Turnovers
4. Assists
5, Rebounds

That is, among the listed stats, Steals contribute the most to total points per game while Rebounds contribute the least.

A suggestion: Don't dismiss it until you read it.

Haven't read it yet, but I think it has face validity. A steal prevents a shot by the opponent and potentially gives you a shot. Makes sense if that's really valuable.

 

[CuseFaninVT]

I'd imagine that steals lead to the most high percentage shots (fast break dunks) so that's why they are so highly rated. Which is sad considering our two guards were the best at steals in the ACC yet our fast breaks were so few and far between.

 

[Millhouse]

Haven't read it yet, but I think it has face validity. A steal prevents a shot by the opponent and potentially gives you a shot. Makes sense if that's really valuable.

That doesn't add up to nine though. Taking all the immeasurable benefits of good team defense and chalking it up to steals

 

[sdhusky]

Haven't read it yet, but I think it has face validity. A steal prevents a shot by the opponent and potentially gives you a shot. Makes sense if that's really valuable.

The basic argument is that a steal or a block is the hardest to replace.

For example, if you sit a guy who scores 15 points a game, your team isn't going to score 15 points less per game. The replacement player and the rest of the team is going to make up most of that 15 points.

However, if you sit a guy who makes 3 steals per game or a guy who blocks 3 shots per game, you are likely going to lose that production. "everyone" can score the basketball, not everyone can steal or block the basketball.

 

[PoppyHart]

And somewhere Aaron Craft laughs and smiles.

 

[OttoinGrotto]

That doesn't add up to nine though. Taking all the immeasurable benefits of good team defense and chalking it up to steals

Skimmed it, looks like he's saying that steals at the NBA level are more function of individual ability than good team defense, which is why they weigh so heavy in his regression.

Meh. I kinda like it. Feels like you should too. I'm drawing parallels in my mind to punting vs go for it on 4th stuff.

 

[Knicks411]

I can also buy that a steal, not only gives you the ball, but is more likely to lead to a good offensive shot than a block or a rebound.

Could this also be because there are so few steals, relative to the other stats? Not sure if that skews this data or something.

 

[OttoinGrotto]

I can also buy that a steal, not only gives you the ball, but is more likely to lead to a good offensive shot than a block or a rebound.

Could this also be because there are so few steals, relative to the other stats? Not sure if that skews this data or something.

I think that's a part of it.

 

[Millhouse]

Skimmed it, looks like he's saying that steals at the NBA level are more function of individual ability than good team defense, which is why they weigh so heavy in his regression.

Meh. I kinda like it. Feels like you should too. I'm drawing parallels in my mind to punting vs go for it on 4th stuff.

it's not the same. the punting stuff uses actual results. there's no expected value calculation here unless a steal leads to Dan Cortese hitting a 20 pointer from the special circle.

if every steal stopped an and-1 three pointer and led to an and-1 three pointer, this would still be flawed.

being really good leads to steals. getting a steal doesn't make you really good.

gambling for steals is going to have a big payoff. when you don't get them, you lose. and then the regression guys say aha steals are really valuable! there's no control for failed steals though.

 

[OttoinGrotto]

it's not the same. the punting stuff uses actual results. there's no expected value calculation here unless a steal leads to Dan Cortese hitting a 20 pointer from the special circle.

if every steal stopped an and-1 three pointer and led to an and-1 three pointer, this would still be flawed.

being really good leads to steals. getting a steal doesn't make you really good.

gambling for steals is going to have a big payoff. when you don't get them, you lose. and then the regression guys say aha steals are really valuable! there's no control for not getting missed though.

 

[Knicks411]

Correct me if I'm wrong, but in a scenario where say Tyler is guarding someone, he knocks the ball away and it rolls to CJ, and CJ picks it up, Cj is the one credited with the steal, no? just asking.

 

[scottynguyen]

Correct me if I'm wrong, but in a scenario where say Tyler is guarding someone, he knocks the ball away and it rolls to CJ, and CJ picks it up, Cj is the one credited with the steal, no? just asking.

Whoever tips/strips the ball, regardless of who ends up with the ball, is credited with the steal.

So in that scenario Tyler should be given the steal.

 

[AccuRater]

The basic argument is that a steal or a block is the hardest to replace.

For example, if you sit a guy who scores 15 points a game, your team isn't going to score 15 points less per game. The replacement player and the rest of the team is going to make up most of that 15 points.

However, if you sit a guy who makes 3 steals per game or a guy who blocks 3 shots per game, you are likely going to lose that production. "everyone" can score the basketball, not everyone can steal or block the basketball.

Exactly correct. That's at the core of the author's interpretation of the data. There are other factors as well. For example, Morris suggested that steals often lead to fast breaks and high percentage shots. But what's most interesting are the data. I'm not yet ready to buy a nine point impact, but steals may be worth more than we thought. For sure, the study needs to be replicated.

Incidentally, have you read Nate Silver's book, "The Signal and the Noise"? I highly recommend it.

 

[Knicks411]

Whoever tips/strips the ball, regardless of who ends up with the ball, is credited with the steal.

So in that scenario Tyler should be given the steal.

Ok, then forget it. I could have sworn I've seen scenarios where the guy picking up the ball was given the steal. Carry on...

 

[PeteCalvin]

Which stat has more impact on the outcome of a Pitt game:

1. The number of clutches
2. The number of grabs
3. The number of moving screens
4. The number of defenders on the floor (6 with Dixon)
5. The number of guys named Karl Hess officiating the game


:noidea:

 

[SWC75]

I'm not big on "replacement value" for evaluating players. It's too speculative. A guy scores 15 points. If he didn't play, somebody else would have scored those 15 points. Really? How do we know nobody else would have made the steal?

Also, our backcourt this year was good at stealing the ball but not at covering the three pointer or preventing penetration to the top of the key. As a result we were a weaker defensive team. Last year, with MCW and Triche we averaged 9 steals a game and gave up 42.6% shooting on two pointers and 28.4% on three pointers. This year we averaged 8 steals and gave up 4.5% on two pointers and 34.5% on three pointers. Which are the more important stats? MCW and Triche were clearly better on defense but the difference in steals is minimal.

 

[Cusefan0307]

As an advanced stats geak it's really simple. If you look at statistics steals and blocks are the least likely stats to happen, followed by turnovers, assists and rebounds. Each stat has a point value tied to it. When he did his analysis he determined what directly lead to points. You can't score without possessions. Steals and blocks add possessions to your team and opportunities for easy baskets. Turnovers take possessions away from you and lead to easy baskets. Assists are mostly just a by product of scoring. Some players are better passers, but it's easier to replace assists and rebounds because shots will always be made and missed , that's what basketball is. This is how the Oakland A's were able to compete in the early oughts. People will find fault with any statistical measure, and for what it is worth Tyler Ennis lead our team in win shares this year, ironically followed by Trevor Cooney. Supposedly they added the most value to our team by statistical measures.

By advanced statistical measures we had three players last year outperform our entire team this year, and a fourth player from last year outperformed everyone but Ennis this year, so advanced statistics show that we were a better team last year compared to this year, especially the core four of mcw, southerland, fair, and triche. Advanced statistics show also that cj was much better last year compared to this year. I know people will argue that it doesn't take into account a players role on the team or variance in performance, but their is no doubt it is a great measure in analyzing the quality of a team.

http://www.sports-reference.com/cbb/schools/syracuse/2014.html

 

[Millhouse]

Haven't read it yet, but I think it has face validity. A steal prevents a shot by the opponent and potentially gives you a shot. Makes sense if that's really valuable.

the other team gets the ball back.

it's not like you get a fixed number of possessions per game.

more steals mean more possessions for everyone.

 

[OttoinGrotto]

the other team gets the ball back.

it's not like you get a fixed number of possessions per game.

more steals mean more possessions for everyone.

Sure, but when it's stolen from you you do lose an opportunity to take a shot on one of your possessions.

 

[Millhouse]

Sure, but when it's stolen from you you do lose an opportunity to take a shot on one of your possessions.

slightly off topic here but i am not that impressed with 538. it's like watered down buzzfeed for wannabe nerds. but who wants to be a nerd?

i think the crazy nine number comes from not being able to control for steals attempted. you wouldn't want to judge a player simply on shots made, what if they had to miss a lot of shots to make that number of shots? doesn't seem like the regression controls at all for the downside of going for a steal and not getting it. a missed steal just shows up as points for the other team and no additional steals for you. going for a steal and giving up easy points shows up in his regression as "boy you should get steals!" instead of "missed steals hurt"

 

[TheCusian]

slightly off topic here but i am not that impressed with 538. it's like watered down buzzfeed for wannabe nerds. but who wants to be a nerd? i think the crazy nine number comes from not being able to control for steals attempted. you wouldn't want to judge a player simply on shots made, what if they had to miss a lot of shots to make that number of shots? doesn't seem like the regression controls at all for the downside of going for a steal and not getting it. a missed steal just shows up as points for the other team and no additional steals for you. going for a steal and giving up easy points shows up in his regression as "boy you should get steals!" instead of "missed steals hurt"

Are you implying that you aren't a stat nerd ?