Untitled Document
Brown Columbia Cornell Dartmouth Harvard Penn Princeton Yale



 Page 5 of 6 « First<3456
Username Post: Red and Blue Scrimmage        (Topic#15309)
mrjames 
Professor
Posts: 6062

Loc: Montclair, NJ
Reg: 11-21-04
10-30-13 01:32 PM - Post#154846    
    In response to palestra38

We're talking about two different things here. I don't think anyone is arguing that a covered shot is harder to hit than an open shot. That is absolutely true, and it's why, once out of layup range, the percentage might decline quickly and have an inelastic relationship to distance.

The problem is that this relationship faces everyone and doesn't necessarily explain year-over-year deviations.

For that, we now have two explanations:

1) YOURS: Coaches scout 2-pt jumper shooting percentage and tell their players to guard certain shooters (those with well-above average conversion rates last year) tougher and leave others (those with well-below average conversion rates last year) wider open. Under this theory, coaches MUST be passing along different guarding instructions for the better and worse shooters last year.

2) MINE: There exists a huge element of luck in jumpers and not a ton of specialization at that shot. Thus, deviations from the average hit rate (especially in the positive direction) are far more likely to be an expression of luck than skill. The safest bet for a player who exceeded the average last season is not their own rate from last season, but the average from last season.

My assumption is that most coaches don't even chart jumpers versus non-jumpers, so I don't really buy the fact that they have their players guarding certain players tougher on 2-pt jumpers based on their conversion rate last year.

 
palestra38 
Professor
Posts: 32680

Reg: 11-21-04
10-30-13 01:44 PM - Post#154847    
    In response to mrjames

I do believe that coaches look at shooting percentages in terms of instructing their players on whom to double cover and whom should not be left alone. That's why you see bad shooters regularly left wide open at the top of the key---teams figure you are doing them a favor if you take that shot. Additionally, I think that last year, Henry was indeed left open a lot and given little respect. People doubled Doc, DNH, Hicks and Cartwright before giving him a look. My belief is that despite Henry's percentage shooting last year, that lack of respect will continue because frankly, Penn has more dangerous offensive players. So I predict Henry will not fall back to the mean on mid-range jumpers.

 
TheLine 
Professor
Posts: 5597

Age: 60
Reg: 07-07-09
10-30-13 01:54 PM - Post#154851    
    In response to mrjames

I'm guessing it's both though it's impossible to measure. Some of it is regression to the mean; some of it is that coaches notice a player has been scoring more lately - not necessarily HOW he is scoring but THAT he is scoring - and so he gets more attention.


 
mrjames 
Professor
Posts: 6062

Loc: Montclair, NJ
Reg: 11-21-04
10-30-13 02:17 PM - Post#154852    
    In response to palestra38

Good luck to you in that endeavor.

I'm always fascinated how people can, in the face of overwhelming evidence, still believe that their special interest is the exception. There's no doubt that it's a very human way to think. Psychologically, though, it's fascinating.

To some extent, it's the difference between the fan mindset and the gambler mindset. You want hope. I want to be right.

 
palestra38 
Professor
Posts: 32680

Reg: 11-21-04
10-30-13 03:31 PM - Post#154853    
    In response to mrjames

Wait a second---here's where I always have a problem with you. This isn't Nate Silver aggegating current possibilities from data on present intent. What people say they will do in a few weeks is a good predictor of what they will in fact do. You are basing a prediction of future performance (on limited data) from PAST performance. There are a myriad of variables you cannot account for. You have no data on this year's performance (the equivalent of a poll right before an election) and may I suggest to you that it is possible that this is the only case you have where everyone from the prior year's team is back. Obviously, if you take a guy who shot well, but all his supporting players are gone the next year , the likelihood is that his percentage goes down. I have real problems when you speak of "overwhelming evidence" being something which may be correlation or coincidence--you just don't know. When I see Silver take 30 polls in the present, I accept his methodology. When you take a fairly limited sample of past performances and have no data to correlate it to the present year, how can you possibly call it "overwhelming evidence?" It is interesting data and I am sure there is some meaning in it---but you take it as overwhelming evidence of future performance and you just can't do that.

 
The Quad 
Sophomore
Posts: 137

Reg: 12-16-04
Re: Stats
10-30-13 03:59 PM - Post#154854    
    In response to mrjames

Stats are helpful, often predict well and frequently insightful but there is a two word response familiar to Penn fans:

Rob Belcore


 
QuakerPTPer 
Sophomore
Posts: 184

Reg: 10-31-06
10-30-13 04:01 PM - Post#154855    
    In response to palestra38

If past activity is such a horrible thing to use for predicting the future, you probably hate amazon for daring to show you a product suggestion.

 
palestra38 
Professor
Posts: 32680

Reg: 11-21-04
10-30-13 04:07 PM - Post#154857    
    In response to QuakerPTPer

The equivalent here is past activity of other lawyers in their mid-50s, not counting me.

If he were basing this on Brooks' past performance, I would not complain. He is predicting a fall-off in Brooks' shooting based on the fact that others have fallen off, and in doing so, he refuses to account for any possibility for those results except the default--luck.

 
palestra38 
Professor
Posts: 32680

Reg: 11-21-04
Re: Stats
10-30-13 04:11 PM - Post#154858    
    In response to The Quad

But there was a reason to predict Belcore's bad performance---his past horrible performance. As I point out below, this prediction on Brooks is not based on Brooks' performance, but the performance of others. There are numerous possibilities for this, including but not limited to: (1) small sample size, (2) changes in the lineup (3) changes in shot selection, (4) greater emphasis on defending this player or (5) injuries or other events affecting performance.

 
QuakerPTPer 
Sophomore
Posts: 184

Reg: 10-31-06
10-30-13 04:28 PM - Post#154859    
    In response to palestra38

I don't think mike is "refusing" anything. He's merely submitting his view on the likelihood of something.

If 450 out of 500 people resembling P38 bought 'sign here' post-it notes on amazon, it's not a bad bet for them to push that product to you. Sure, maybe you have an office manager who takes care of it for you, so of course you don't want it. Maybe all those customers are ambulance chasers, amazon can't account for that! What a terrible model!

It's not a bad bet for mike to make his point, given all the past data. Sure there's things not accounted for, but the repeated notion that it should be tossed into the garbage is getting tired.

 
Jay O 
Masters Student
Posts: 547

Loc: Philadelphia
Reg: 11-16-09
10-30-13 04:35 PM - Post#154860    
    In response to palestra38

Jeez, next weekend can't get here soon enough.

Both sides of this are right, but everyone's phrasing isn't. Here's what I think you're trying to say:

P38: You don't like that Henry Brooks's jump-shooting past success is being written off for this season because other players have generally regressed by this metric. If Brooks plays the same role in this year's offense, why should his percentage drop off?

mrjames: You're saying that it's extremely unlikely that Brooks will shoot 2-point jumpers so well this season because other players who have shot similarly have also regressed. It's not that Brooks won't shoot quite as well (or better), just that if he does, we'll be watching something unusual and uncommon (and special?).

Myself, I'll be impressed if Brooks shoots as well because it'll be mostly without precedent. If he's asked to take these shots regularly and at such a high rate (with success), he'll add a great wrinkle to the Penn offense. I'm not expecting that he will, though, and I hope the success of the team doesn't hinge on his 2-point shooting.

Still, I don't want to put words in anyone's mouth -- where am I wrong?

 
Silver Maple 
Postdoc
Posts: 3765

Loc: Westfield, New Jersey
Reg: 11-23-04
10-30-13 05:09 PM - Post#154863    
    In response to Jay O

In essence, the question here is this: what's the better predictor of Henry Brooks' shooting performance this coming season-- Henry's own past performance, or the past performance of others?

I've been conducting and analyzing quantitative analysis for more than 25 years, and I'd say that there's no definitive answer to this question. There certainly isn't any 'overwhelming' evidence to support either position.

 
palestra38 
Professor
Posts: 32680

Reg: 11-21-04
Red and Blue Scrimmage
10-30-13 05:09 PM - Post#154864    
    In response to QuakerPTPer

Who says they are like me---all we are talking about is basketball players with nothing else in common. Not all lawyers are alike either.

And by the way, Mike's sample was 30 players, not 500.

 
AsiaSunset 
Postdoc
Posts: 4350

Reg: 11-21-04
10-30-13 05:37 PM - Post#154866    
    In response to palestra38

It would be interesting to see a shot chart for Brooks. I'm thinking the shots in close might have actually lowered his shooting percentage. He's deadly with that 15 foot jumper.

 
QuakerPTPer 
Sophomore
Posts: 184

Reg: 10-31-06
Re: d and Blue Scrimmage
10-30-13 05:39 PM - Post#154867    
    In response to palestra38

Sample size isn't your complaint though. He could have 100k guys in there and you'd still think it was bogus. No point in discussing further, I just figure I'm good for an annual attempt.

 
mrjames 
Professor
Posts: 6062

Loc: Montclair, NJ
Reg: 11-21-04
10-30-13 05:41 PM - Post#154868    
    In response to palestra38

I'm showing you 16 past players who fit this same profile and 15 of them saw a decline the next season.

Your point is that Henry isn't exactly like the other 16.

My point is that the 16 aren't exactly like each other either. They all faced different circumstances. And yet just ONE of the 16 saw a different outcome. If the true odds were 50/50, having 15 of 16 break the same way would happen essentially 0 percent of the time. As differing circumstances come, I think in order to find a unique one that would really make a difference, you need to be thinking on the order of making the hoop wider in year two.

And Nate Silver absolutely uses past data in his model. He uses it to decide how much to weight his polls and how to adjust for historical bias in those polls. There's also an element of his model based on the fundamentals of the state in question, which is all modeled out by seeing how past changes have impacted the way the state votes. The reason he includes all of that past data in his present forecast of future results is because he finds the past data to be instructive in identifying trends and fluctuations that without context could cause him to view historically irrelevant changes in present numbers as a reason to change his forecast of the future. Ken Pomeroy has found that his preseason metrics (based primarily on carryforwards from the past) enhance the predictiveness of his model, even when he has present year results flowing in. That's why he used to leave the preseason numbers in the model until January and now keeps them in throughout the season. This idea that the past doesn't have predictive value is fallacious.

The good news is that we could have as many as 17 players (assuming they take enough jumpers again) who are +/- 3 pct pts from the average to monitor this season. We'll see how things turn out.

 
mrjames 
Professor
Posts: 6062

Loc: Montclair, NJ
Reg: 11-21-04
10-30-13 05:55 PM - Post#154869    
    In response to mrjames

Also, lost in my more general approach to this is the specific point that while Henry might have shot 42.6% last season (108 attempts), he shot 29.1% the year prior (79 attempts).

So, the idea that his own past performance is saying anything other than what the rest of the data set is saying is dubious.



 
weinhauers_ghost 
Postdoc
Posts: 2125

Age: 64
Loc: New York City
Reg: 12-14-09
Re: nba shooting data
10-30-13 07:14 PM - Post#154873    
    In response to mrjames

You guys might want to take a look at the shot charts Kirk Goldsberry prepares for grantland.com .

 
palestra38 
Professor
Posts: 32680

Reg: 11-21-04
Red and Blue Scrimmage
10-30-13 07:48 PM - Post#154875    
    In response to mrjames

Nate Silver uses voter's predictions of how they are going to vote---very accurate representations of how they eventually vote. His use of "past" data is simply older polls.

I don't see how you see Brooks' freshman performance as somehow predicting reversion to that performance...you don't think improvement is possible?

I guess PTP is right in a way---while I believe your sample size is way short of sufficient to take your conclusions, I would still doubt that past performance of other players is predictive of how Henry will do this year with a better team. While I don't know what your data shows, I would assume that most of the players with very good mid-range shooting suffer at least some loss of talent around them the following year and also lose the "lack of respect" element that is so important to getting open from 12-15 feet. Let me ask you though, what would you say given your expectation that Henry has "zero" chance of equaling or bettering his performance this year, if he indeed does so?

 
mrjames 
Professor
Posts: 6062

Loc: Montclair, NJ
Reg: 11-21-04
Re: d and Blue Scrimmage
10-30-13 08:10 PM - Post#154878    
    In response to palestra38

You win. Henry is going to rock this year.

 
 Page 5 of 6 « First<3456
Icon Legend Permissions Topic Options
Report Post

Quote Post

Quick Reply

Print Topic

Email Topic

11471 Views





Copyright © 2004-2012 Basketball U. Terms of Use for our Site and Privacy Policy are applicable to you. All rights reserved.
Basketball U. and its subsidiaries are not affiliated in any way with any NCAA athletic conference or member institution.
FusionBB™ Version 2.1 | ©2003-2007 InteractivePHP, Inc.
Execution time: 0.228 seconds.   Total Queries: 18   Zlib Compression is on.
All times are (GMT -0500) Eastern. Current time is 07:40 AM
Top