Comments on: Layups: Hollinger’s Forecasts Are Here! (2010-11 Edition)

By: Ivan Kudasik

Ivan Kudasik — Mon, 14 Nov 2011 14:06:51 +0000

Hi Craig! I'm a personal trainer in SD, CA and would love to share my passion for TRX use with my clients! I have used TRXs at the gyms i've worked at in the past and believe it to be an invaluable tool for mobility/strength/core stability work and SO much more! If i had my own TRX I would be able to share this tool with many and hopefully inspire success and strength with it!

By: Neil Paine

Neil Paine — Mon, 04 Oct 2010 13:57:44 +0000

All that means is that team point differential can predict team performance within +/- 2 wins. There's a team adjustment that makes sure WP matches up with wins predicted by pt. differential, but that doesn't necessarily say anything about the validity of the metric.

By: Leroy Smith

Leroy Smith — Sun, 03 Oct 2010 21:51:03 +0000

With every advanced stat there are problems. Wins Produced is not different. What I am propsing is that, when you put box score stats into wins produced for any random team in the past you will get within 2 wins or so of what the team actually produced. Dan Rosenbaum and you guys make great points about the weaknesses of WP, but at the end of the day it predicts. That was never denied in any of the links you kindly provided.

Don't think I'm one those that are a slave to any one system. But WP has helped me pull off some neat party tricks.

By: Neil Paine

Neil Paine — Sun, 03 Oct 2010 19:46:11 +0000

I feel obliged to point out, Leroy, that Wins Produced also has a number of well-known, extremely serious flaws:

http://sonicscentral.com/apbrmetrics/viewtopic.php?t=877

http://www.sonicscentral.com/apbrmetrics/viewtopic.php?p=18975

http://sonicscentral.com/apbrmetrics/viewtopic.php?p=17287#17287

http://sonicscentral.com/apbrmetrics/viewtopic.php?t=1589

http://sonicscentral.com/apbrmetrics/viewtopic.php?p=13946#13946

Et cetera, et cetera.

By: Leroy Smith

Leroy Smith — Sun, 03 Oct 2010 18:19:23 +0000

PER is worthless and any predictions that use it is therefore worthless as well. All PER does is reward scoring, whether or not is efficient. Look at the top 10 performers in PER and look at the top in scoring average and you will see very similar (not identical) lists. The only advanced stat that actually (and accurately) predics wins, baring injury to key players, is Wins Produced (WP) by Dave Berri. Search wages of wins in google and you will be convinced.

By: BSK

BSK — Wed, 29 Sep 2010 01:36:07 +0000

Thanks, Neil. The "somewhat arbitrary" aspect of it is what gets me. I get that it's hard to arrive at anything 100% definitive, but we certainly ought to try. And, as you point out, my biggest complaint is when someone purports a stat to be something other than it is. I get that every stat has it's strengths and weaknesses, but as long as we are clear about that and use stats where they are strong and not where they are weak, I'm willing to entertain just about anything. But Hollinger often acts as if PER is the end-all, be-all, and whatever it says to be true is true (even with his own admitted caveat about the defensive aspect). Now, I suppose there may be more rigorous analysis behind his conclusions that he doesn't put into every column for the sake of his audience (he does write for ESPN, afterall). But it's still hard to make much of PER when it seems like a mysterious number plucked out of no where. I do think his contributions with regards to pace factor have been very important. Must give credit where credit is due.

By: Neil Paine

Neil Paine — Tue, 28 Sep 2010 12:59:14 +0000

Right, the weights are basically a more evolved version of what Tendex had in the early 90s. Give a somewhat arbitrary positive weight to "good stuff", a somewhat arbitrary negative weight to "bad stuff", and add everything up. The only different wrinkle with PER is that it adjusts for pace and is scaled such that the NBA average is always 15. In essence, it's a dimensionless quantity, with an easy-to-understand scale, that correlates well with offensive rating at the team level (which doesn't necessarily mean it correlates well with individual offensive impact) and poorly with defensive rating at the team level. I don't think it claims to be much more, although Hollinger does sometimes treat it as though it were a more complete metric.

By: ManchvegasBob

ManchvegasBob — Tue, 28 Sep 2010 11:43:01 +0000

Game Score is the simplified metric Hollinger has thrown out there that emulates PER to some extent. From http://www.nbastuffer.com

Game Score Formula =(Points)+0.4*(Field Goals Made)+0.7*(Offensive Rebounds)+0.3*(Defensive rebounds)+(Steals)+0.7*(Assists)+0.7*(Blocked Shots)- 0.7*(Field Goal Attempts)-0.4*(Free Throws Missed) - 0.4*(Personal Fouls)-(Turnovers)

By: BSK

BSK — Tue, 28 Sep 2010 02:16:31 +0000

Chuck-

I understand all that. But the numbers themselves are not explained. At least not in a way that I get.

Neil-

I agree. I still read JH, but I struggle with PER. I think he sometimes offers some really great insight into the game. But sometimes he seems to be number-heavy for the sake of it, and ends up muddling around.

By: Neil Paine

Neil Paine — Tue, 28 Sep 2010 01:05:00 +0000

Re: #2 - I wouldn't get too hung up on PER when reading Hollinger's stuff. All of the criticisms Walter mentioned are valid, and JH does frequently refer to it, but it's possible to appreciate his previews and player comments while still taking PER with a grain of salt. Speaking for myself, I almost never use PER at all, but I value the non-PER content of what he writes.