APBRmetrics

http://blog.cacvantage.com/2013/12/the- ... -shot.html

I'll be publishing more at some point looking at open mid-range vs. contested 3s .

Interesting stuff. Did you pick those three guys (Kosta, Taj, and Amir) because of their reputation as defenders (eg high rapm) or because they were among the leaders of your DPS/100 (or was it just because they all have unique first names)? I'm also curious which players are among the leaders in this metric, if you'd be willing to share.

If you have the data to calculate the opponents' FG% when a shot is contested, can't you also calculate their FG% when a similar shot is not contested? That seems like a better baseline to work from in terms of determining how many points are saved by a particular player contesting a shot.

jbrocato23 wrote:Interesting stuff. Did you pick those three guys (Kosta, Taj, and Amir) because of their reputation as defenders (eg high rapm) or because they were among the leaders of your DPS/100 (or was it just because they all have unique first names)? I'm also curious which players are among the leaders in this metric, if you'd be willing to share.

I did pick those guys because of their reputation as defenders and was aware of their high RAPM. Also, those three guys generally get less press then the Dwights of the world.

I actually haven't calculated the DPS/100 out for all the players in the league nor I don't feel comfortable sharing without getting permission from Vantage. I have released some numbers for a couple other players on my twitter. In regards to the leaders, as I said, I'd have to calculate it for the whole league first. To be honest, there are still some adjustments to the metric I'd like to make. For example, I've used the league wide frequency while I think it would be better to use specific position averages. Since we're looking big men, there is no point in inflating all of their numbers by including PGs in the league frequency.

At the moment, I am focusing on open mid-range vs. contested 3s but I will certainly come back to this and make the adjustments while publishing the leaders.

xkonk wrote:If you have the data to calculate the opponents' FG% when a shot is contested, can't you also calculate their FG% when a similar shot is not contested? That seems like a better baseline to work from in terms of determining how many points are saved by a particular player contesting a shot.

That is possible however their FG% when not contested is going to be mostly affected by the shooter and not the defender. So this is where luck could play a big role- i.e. if the defender consistently doesn't contest shots but the shooter he's guarding always misses. He'll end up with a higher points saved while not really playing good defense.

Additionally, I've looked at position averages for opponents' FG% on contested shots near the basket and while there is a difference between the FG% allowed for Centers and PGs, the difference is pretty small relative to the difference in contest frequency. My guess is that when a PG gets stuck under the rim, they don't put the effort in to contest the shot because they just assume it will go in. That is actually false since contested shots near the rim guarded by PGs only go in about half the time. One of the studies I want to look at is if PG defense is "irrelevant" as so many people say or if PGs are just lazy defensively. All of the evidence I have at the moment points to the latter but I'd like to break it down further.

You mentioned a contested FG% in the same manner as BABIP in baseball. I'm hesitant to do this because you're assuming Duncan contesting the shot is the same as, say, Ty Lawson. Or think about Battier covering the shooter's eyes when he contests. On an aggregate level across the league, it's a fine assumption, but we're missing special players who over-perform on defense because they do things in either a unique way or at an elite level when contesting shots. I think you'd also need a few years of data to check trends to see whether or not a player's low contested FG% is stable or not. In baseball, there are certain players who consistently over- or under-perform their BABIP (knuckleball pitchers are often the example.) I'm sure we could find the same in basketball.

knarsu3 wrote:That is possible however their FG% when not contested is going to be mostly affected by the shooter and not the defender. So this is where luck could play a big role- i.e. if the defender consistently doesn't contest shots but the shooter he's guarding always misses. He'll end up with a higher points saved while not really playing good defense.

That seems like it could be true, but it's part of something else that could be plausible - players deciding when to contest a shot depending on who's shooting. I could see it going either way- defenders could let guys shoot because they think the shooter is poor and they want him to shoot, or they could let better players shoot because better players are more likely to draw fouls or more likely to make it anyway. That's something you could tease out, but only if you have all the shooter-relevant data.

AcrossTheCourt wrote:You mentioned a contested FG% in the same manner as BABIP in baseball. I'm hesitant to do this because you're assuming Duncan contesting the shot is the same as, say, Ty Lawson. Or think about Battier covering the shooter's eyes when he contests. On an aggregate level across the league, it's a fine assumption, but we're missing special players who over-perform on defense because they do things in either a unique way or at an elite level when contesting shots. I think you'd also need a few years of data to check trends to see whether or not a player's low contested FG% is stable or not. In baseball, there are certain players who consistently over- or under-perform their BABIP (knuckleball pitchers are often the example.) I'm sure we could find the same in basketball.

I actually agree with this and there is some evidence to support this theory. For example, Battier does have a lower contested FG%. So does Hibbert. But then there are the curveballs- like Lawson (albeit on very few attempts) having a lower contested FG% than Noah (who was surprisingly high). Ultimately, I am a bit early in my analysis to reach a definitive conclusion one way or another but the evidence at a league wide level supports contested FG% being somewhat random. However, I do plan on testing the stability of contested FG%. I have 2 years of data currently so the best I can probably do at the moment is test 10-15 game samples over time.

xkonk wrote:

knarsu3 wrote:That is possible however their FG% when not contested is going to be mostly affected by the shooter and not the defender. So this is where luck could play a big role- i.e. if the defender consistently doesn't contest shots but the shooter he's guarding always misses. He'll end up with a higher points saved while not really playing good defense.

That seems like it could be true, but it's part of something else that could be plausible - players deciding when to contest a shot depending on who's shooting. I could see it going either way- defenders could let guys shoot because they think the shooter is poor and they want him to shoot, or they could let better players shoot because better players are more likely to draw fouls or more likely to make it anyway. That's something you could tease out, but only if you have all the shooter-relevant data.

Yup, that's possible. I did have plans to eventually do a shooter adjusted shot defense metric. The data is there, not sure when I'll get around to it (time is not kind to you when taking graduate courses). That will be a longer term project though. However, maybe a post looking at the defenders who guard the best shooters/players most often would be interesting.

I think over a large sample though, you aren't likely to see that. Maybe a few games here and there but I can't imagine a player repeatedly choosing not to contest a shot cause they keep facing poor shooters. Eventually that player is going to guard a good shooter. Likewise, the opposite is true.

knarsu3 wrote:I actually agree with this and there is some evidence to support this theory. For example, Battier does have a lower contested FG%. So does Hibbert. But then there are the curveballs- like Lawson (albeit on very few attempts) having a lower contested FG% than Noah (who was surprisingly high). Ultimately, I am a bit early in my analysis to reach a definitive conclusion one way or another but the evidence at a league wide level supports contested FG% being somewhat random. However, I do plan on testing the stability of contested FG%. I have 2 years of data currently so the best I can probably do at the moment is test 10-15 game samples over time.

Lawson's contested fg% may be lower than Noah's because most of the time he guards point guards while Noah guards centers. Center's make their shots at a much higher rate near the basket than point guards. That is why you can't really compare this stat across positions.

italia13calcio wrote:

knarsu3 wrote:I actually agree with this and there is some evidence to support this theory. For example, Battier does have a lower contested FG%. So does Hibbert. But then there are the curveballs- like Lawson (albeit on very few attempts) having a lower contested FG% than Noah (who was surprisingly high). Ultimately, I am a bit early in my analysis to reach a definitive conclusion one way or another but the evidence at a league wide level supports contested FG% being somewhat random. However, I do plan on testing the stability of contested FG%. I have 2 years of data currently so the best I can probably do at the moment is test 10-15 game samples over time.

Lawson's contested fg% may be lower than Noah's because most of the time he guards point guards while Noah guards centers. Center's make their shots at a much higher rate near the basket than point guards. That is why you can't really compare this stat across positions.

I can adjust for all of this (I have shooter position and defender position data) and if you noticed in the article, I only looked at big men. My first conclusions about the BABIP-contested FG% comparison was actually based off looking at only big men. In regards to Lawson specifically, as I said, very few attempts which is why I would assume his is lower than Noah's. And yes, obviously PGs will be guarding PGs more when attacking the rim. However, I looked at PGs vs. PGs and Centers vs. Centers and their contested FG% is about the same near the basket. If you get cross matchups, like PGs guarding Centers near the basket, that's where you obviously see a big difference in contested FG% but that rarely happens.

Ultimately, I do need to test the stability of contested FG% over a time period. When I get the data, that will be one of the first things on my to do list.