APBRmetrics :: View topic - Home court scorekeeping bias

APBRmetrics :: View topic - Home court scorekeeping bias - blocks

APBRmetrics
The statistical revolution will not be televised.

FAQ

Memberlist

Usergroups

Profile

Home court scorekeeping bias - blocks
Goto page Previous 1, 2, 3 Next

APBRmetrics Forum Index -> General discussion

View previous topic :: View next topic

Author

Message

Mike G

Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

Posted: Mon Aug 24, 2009 10:39 am Post subject:

HoopStudies wrote:

...vagueness when it comes to a block and a defensive rebound vs a steal, especially when a player is starting to go up for a shot.

There is indeed a negative correlation (-.107) between home blocks advantage and home steals advantage. A notable exception is the Blazers, who rank at the top of both lists.

By 'home advantage' I mean relative to their opponents, at home vs away. Blazers out-steal opponent by 21% at home, vs being out-stolen by 6% on the road, for a home advantage of 1.29 (121/94).

Portland also led the league in home/away point differential -- 1.11 . Over the whole league, I find these correlations between pt-diff and --
Ast .58
Reb .50
Blk .45
3fg% .45
Stl .27
TO -.34
PF -.52

And so, there does seem to be a correlation effect involving being 'better' at home, and especially in the very subjective scorekeeper details of Ast and Blk. Rebounds and fouls are just about equal and opposite. None of these are anywhere near zero.

Quote:

The Nuggets arena, for example, awards many fewer steals at home than are seen on the road...

Yet, Den got 1.13 times as many steals as their opponents at home, vs 1.07 on the road.
Saying "the arena" created the steals, the blocks, or the home runs, is a bit off. In 1927, Yankee Stadium had twice the AL-average number of homeruns. Does that mean you or I would be twice as likely to hit a homer in Yankee Stadium? Or does this spring from the fact that the Yankees hit 3 times as many homers as everyone else? Being a player chosen at random from a Yankee Stadium game doesn't make you a Yankee.

Someone wrote that Utah was an 'easy' place to get assists in the '90s. That's only true if playing in Utah made you more likely to be John Stockton. For no one was this the case.
_________________
`
36% of all statistics are wrong

Last edited by Mike G on Mon Aug 24, 2009 11:07 am; edited 1 time in total

Ilardi

Joined: 15 May 2008
Posts: 265
Location: Lawrence, KS

Posted: Mon Aug 24, 2009 11:05 am Post subject:

Mike G wrote:

There appear to be 3 possible sources of reliable home/away differences on various metrics:

1) scorekeeper bias
2) home/away performance difference due to referee bias
3) non-ref-induced home/away performance difference

Since the only metric not amenable to ref-induced biasing effects (FT%) shows no home/away difference, a parsimonious assumption is that the effect of #3 above is minimal or nonexistent.

Moreover, one could presumably begin - at a first approximation - to tease apart the effects of ref-induced performance bias from scorekeeping bias on a given metric for a given team by:

(1) predicting (in regression) the team's expected home/away performance-related difference on that metric based on the team's home/away point-differential (relying on Mike G's analysis of leaguewide correlations between pt-diff and the metric in question); and

(2) attributing any observed surplus (i.e., actual-expected) on the metric to potential scorekeeper bias (and estimation noise).

Mike G

Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

Posted: Mon Aug 24, 2009 11:30 am Post subject:

But Steve, if there's no better play by players/teams at home, other than that 'induced' by refs, then why is there such disparity in the improvements seen by teams at home? Blazers are 11% better (point-differential) at home; Wolves were 2% worse. FT% is one stat that's removed from the 'action'. It involves a player's peace of mind, concentration, work habits, etc. But not any teamwork or quick-thinking. Once at the line, the ref's whistle-tendencies are (in theory) no part of the task at hand. Both Por and Min are young teams, but they seem to have opposite reactions to playing home/road. _________________ ` 36% of all statistics are wrong

Ilardi

Joined: 15 May 2008
Posts: 265
Location: Lawrence, KS

Posted: Mon Aug 24, 2009 11:37 am Post subject:

Mike G wrote:

But Steve, if there's no better play by players/teams at home, other than that 'induced' by refs, then why is there such disparity in the improvements seen by teams at home? Blazers are 11% better (point-differential) at home; Wolves were 2% worse.

FT% is one stat that's removed from the 'action'. It involves a player's peace of mind, concentration, work habits, etc. But not any teamwork or quick-thinking. Once at the line, the ref's whistle-tendencies are (in theory) no part of the task at hand.

Both Por and Min are young teams, but they seem to have opposite reactions to playing home/road.

There's a robust research literature documenting that refs are influenced by the intensity of home crowd cheering/jeering/etc. (The effect exists across many sports - and a sizeable portion of this literature actually comes from soccer.) I would hypothesize, therefore, that Portland's crowd - by virtue of their high level of expressed ardor - exerts a large biasing effect on refs, and that Minnesota's more phlegmatic crowd does not.

In fact, an interesting study might involve using a dB-meter to assess the mean/peak decibel levels of various home crowds, and seeing how strongly it correlates with observed home/away differentials.

Mike G

Joined: 14 Jan 2005
Posts: 3601
Location: Hendersonville, NC

Posted: Mon Aug 24, 2009 12:40 pm Post subject:

Ilardi wrote:

... an interesting study might involve using a dB-meter to assess the mean/peak decibel levels of various home crowds, and seeing how strongly it correlates with observed home/away differentials.

Ah, jeez. As good as that sounds in theory, we'd still have to somehow (digitally?) subtract the arena-induced sound effects from the general noise level, to arrive at actual 'crowd' decibels.
Wolf howls? Speedway roars? Ack.
_________________
`
36% of all statistics are wrong

Ilardi

Joined: 15 May 2008
Posts: 265
Location: Lawrence, KS

Posted: Mon Aug 24, 2009 2:23 pm Post subject:

Mike G wrote:

Ilardi wrote:

... an interesting study might involve using a dB-meter to assess the mean/peak decibel levels of various home crowds, and seeing how strongly it correlates with observed home/away differentials.

Yeah, I thought about the measurement confounds, as well. It's a non-trivial challenge, but still do-able in theory.

In lieu of such an ambitious undertaking, it's probably possible to get reasonably valid subjective ratings of each venue from a set of raters who actually attend a large number of games at each one.

DJE09

Joined: 05 May 2009
Posts: 148

Posted: Mon Aug 24, 2009 9:37 pm Post subject:

Ilardi wrote:

This is not true. Over all games, all venues FT% is the same for home as opposed to road team. But, when this is reduced to the team / venue level we get differences between observed Home / Away FT%. Therefore, at the team level we can say there is evidence for 3) occuring.

Ilardi

Joined: 15 May 2008
Posts: 265
Location: Lawrence, KS

Posted: Mon Aug 24, 2009 9:46 pm Post subject:

DJE09 wrote:

Ilardi wrote:

Then presumably there exists a roughly equal number of teams for whom the opposite of #3 is the case (i.e., for every team that performs better at home, there is another that performs commensurately better on the road) . . . which seems counter-intuitive. Are these team-level (venue-level) FT% differences meaningful, or just noise (or an artifact of teamwise home/away strategic differences in choosing whom to foul)?

Brian M

Joined: 25 Nov 2006
Posts: 40

Posted: Tue Aug 25, 2009 12:28 pm Post subject:

Ilardi wrote:

This seems a bit facile, given the drastic difference in context between playing a basketball floor game effectively vs shooting free throws effectively.

We know that many things great and small can affect performance. Location? Students can improve test taking performance by studying in the same physical location as where the test takes place. Social factors? Female students have been shown to perform worse on mathematics tests when males are also taking the test. White males have been shown to perform worse on athletic tasks when black males are also performing these tasks.

Given that robust effects on performance exist even for such subtle influences, by far the most reasonable null hypothesis is that the home/away difference affects actual player performance, as well as ref and scorekeeper performance. The factors of location and social influence in home/away games are sledgehammer versions of the more subtle effects of location and social influence noted above and known independently to exist. Of course, there's also the good old placebo effect: if you believe in the home court advantage, then you are likely to perform as if there is one.

It may be that professional athletes are so highly trained that they can suppress all of these influences on performance, but it would really take a lot of substantial evidence to demonstrate that, I think. An intermediate possibility is that there are individual differences in the extent to which home/away play differs, which could muddy the conceptual and statistical picture significantly. But on the average, from what we know about human performance in other domains, the priors would seem to be squarely in favor of the existence of a true home court advantage on player performance.

Ed Küpfer

Joined: 30 Dec 2004
Posts: 785
Location: Toronto

Posted: Tue Aug 25, 2009 1:31 pm Post subject:

Brian M wrote:

Why wouldn't this "extra homecourt effort" then appear in free throw shooting?

Again, there is only a single, unambiguous performance metric in basketball, as context-free as we can get: free throw percentage. That metric has never displayed a home/away bias at the team or player level. From the range of plausible answers to the question "what causes homecourt advantage", the single fact that home players shoot just as well as road players at the line would lend a lot of weight to the answers that do not involve players playing differently depending on what colour their uniforms are that day, or at least it seems so to me.

Sure, it is plausible (perhaps likely) that players play differently in different enviroments. How does this square with the fact that they don't play differently in the one area where defenders and officials cannot apply any influence? To me it suggest, and suggests strongly, that we should look at those external influences for causes.
_________________
ed

Ilardi

Joined: 15 May 2008
Posts: 265
Location: Lawrence, KS

Posted: Tue Aug 25, 2009 1:37 pm Post subject:

Brian M wrote:

Ilardi wrote:

Brian, as a neuroscientist and academic clinical psychologist, I'm fully aware of the contextual influences you've mentioned vis-a-vis performance. As I'm sure you know, these effects are not always robust, the effect sizes are sometimes rather small, and they don't always replicate across investigators. Additionally, the literature you've alluded to pertains principally to cognitive/academic performance, not complex motor behavior; nevertheless, I'll gladly stipulate that these studies may have relevance in athletic domains, as well - perhaps even in the domain of NBA basketball performance - but we just don't have enough direct empirical evidence to settle that issue.

My point was simply that in the one NBA performance domain in which we can actually rule out the well-established, robust influence of referee bias on athletic performance - FT% - there is absolutely no effect of home/away context. Now, I don't know about you, but I find free throw shooting during an actual game to be quite challenging - with fatigue effects, compelling distractions, potentially high performance pressure, etc. (Even though nothing is moving, this fact only serves to give FT shooting certain phenomenological affinities to putting a stationary golf ball - a task legendary for inducing "yips" among even highly skilled athletes.) Thus, a compelling a priori case could easily have been made for the huge expected contextual effects we should have expected to see on FT% (especially considering the fact that hostile crowds explicitly try to adversely influence performance on this task - more so than virtually any other activities that occur on the court).

But no such context effect exists on this difficult task (nor any expectancy driven "placebo effects"), and it was on that basis that I suggested it may be parsimonious to infer minimal home/away contextual influences across other NBA performance domains - at least insofar as they would be alleged to exist independent of the robust, well-established effects of officiating bias. However, I certainly respect your objection to my inference and the reasoning behind it; we just happen to see this one differently.

gabefarkas

Joined: 31 Dec 2004
Posts: 1313
Location: Durham, NC

Posted: Tue Aug 25, 2009 2:01 pm Post subject:

HoopStudies wrote:

FYI - There is an established method for evaluating Arena Factors, based on baseball's park factors:

AF_blk = (Blkrate + OppBlkRate)H/(Blkrate + OppBlkRate)R

So you sum the rates at home and divide the total by the sum of rates on the road.

.....

So Atlanta gives the fewest, Charlotte the most. It's not terribly different than what you did, but it has an easy interpretation: It's the ratio of how many total blocks are given in one arena vs for the same teams in other arenas.

So you could say that Atlanta's shot blockers are hindered at the start by getting only 72% as many home blocks. As I say, I have no idea how to turn this and other numbers into an assessment of bias.

Yes, but these rates are all entangled, since one team's home is another team's away.

In other words, Charlotte played in Atlanta twice, while Phoenix only played there once. So, Atlanta's stingy home scorers are also (artificially?) deflating Charlotte's road totals, making their AF_blk (artificially?) higher. Since Charlotte played their twice, and Phoenix only played their once, could we presume the impact on Charlotte (and their AF_blk) was twice as great?

In fact, I see ATL, MIA, and ORL near the top of the list, while WAS and CHA are near the bottom, and I wonder if that's truly random, or if they had an intercorrelated impact on each other.

Similarly, I see MIL 3rd from the top, while the other four teams in the division (CLE, CHI, IND, DET) all near the bottom.

And following it out in the other direction, I see all five teams in the Northwest (DEN, POR, UTA, MIN, OKC) fairly evenly distributed throughout the list.

Brian M

Joined: 25 Nov 2006
Posts: 40

Posted: Tue Aug 25, 2009 7:44 pm Post subject:

It just seems to me that there is a weak basis for generalizing FT performance to overall basketball performance. Even though FT shooting is an important part of the game, it differs in many important respects from live game situations. player interaction FT: none live: interact with 4 teammates and 5 defenders decision making FT: none live: implement coach's strategy; improvise quick decisions on the fly in reaction to evolving game situation time to gather concentration FT: effectively unlimited live: limited or none, depending on game situation motor action FT: stereotyped, overlearned, the same each time live: dynamic, changing depending on the context-- distance from basket, position of teammates and defenders, etc It does not seem hard to imagine how various factors could influence complex, dynamic, fast paced performance without showing up in performance on a simple, static task with no time pressure. I would not feel comfortable generalizing from the latter to make claims about the former.

DJE09

Joined: 05 May 2009
Posts: 148

Posted: Wed Aug 26, 2009 11:51 pm Post subject:

Ed Küpfer wrote:

Again, there is only a single, unambiguous performance metric in basketball, as context-free as we can get: free throw percentage. That metric has never displayed a home/away bias at the team or player level. From the range of plausible answers to the question "what causes homecourt advantage", the single fact that home players shoot just as well as road players at the line would lend a lot of weight to the answers that do not involve players playing differently depending on what colour their uniforms are that day, or at least it seems so to me.

The fact that there is no League wide Home/Away difference in the FT%, does not mean that for some teams or some players that they do not shoot FT better at home or on the road.
Treating FTA as a binomial event, with p as FT% for home and away, we get five teams with a % difference greater than 3* SD:

Code:

Team Home Away
CLE 0.779 0.733
MIN 0.743 0.795
OKC 0.762 0.812
SAS 0.736 0.788
TOR 0.851 0.796

About half the players in the league do not attempt enough Free Throws to draw any conclusions about their FT% (irrespective to how many DP it is quoted), and there appears to be some small sample effects (ie shooting 1-1 for 2 FTA for a game seems to unduly influence player's FT%) which seem to impact player FT% (around 10% of the league shoots 50-55% for FT - which is larger proportion players than all the other players who shoot under 65% - most of these players are low frequency however) which means there is a greater variability in player FTA and FT%. All of which means that when we apply the usual comparison tests we get the sorts of numbers we might reasonably expect eg. there are three players (our of 223 with meaningful Home / Away FT totals) where their Home and Road FT% differ by more than 3*SD.

There is no REAL difference in the values of the League wide Opponent 2FG% (Road 47.8% cf 49.2% home) and 3FG% (Road 36.3% cf 37.1% Home) that isn't explained by natural variability. If you are prepared to play the "Natural Variability" to explain variance in team and player performance, then we need to apply this to League level averages - which means we are back to saying there is only really any evidence for home court / officiating bias in Blocks from comparison of the Home/ Away Opponent Numbers.

DJE09

Joined: 05 May 2009
Posts: 148

Posted: Thu Aug 27, 2009 12:09 am Post subject:

BobboFitos wrote:

Only sorta scanned this post so I may be misreading, but to me it's pretty obvious why Orlando as an opponent yields a much lower blk% - they take more 3s (which have a very low blk rate!) than any other team.

Again, to echo Irishhand, it's all about context. And it's easy to explain why Orlando as an opponent in other arenas yield lower blk rates.

I agree we could expect ORL to have a much lower block % that average, but why is it 6.6% when they attempt 33% of shots as 3s at home, and 9.25% when they attempt 34% of shots as 3s on the road...

Besides, the argument "I can come up with a good basketball reason why we see this number ..." hinges on your definition of what constitutes a "Good" basketball reason. Is the argument "Chris Andersen was freakishly better at home than on the road this season" a good basketball reason why we see the discrepancy in Denver's Numbers.

Further, I was trying to make the point that Home court bias can be equally evident in the lack of awarding of blocks to the away player, as well as adding blocks to the home player - questioning the assumption that the away rate is "True" whilst the Home rate is biased.

Display posts from previous:

	APBRmetrics Forum Index -> General discussion	All times are GMT - 5 Hours Goto page Previous 1, 2, 3 Next
Page 2 of 3

You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum