APBRmetrics

[Mountain]

Sorry, I meant...

If you weighted the closest on a team last season say a double share and the farthest a half share and the others standard weight, I wonder if the weighted average of the returnees' predictions for this season would lead the solo marks.

[Mountain]

When the predictions varied by 9+ last season the top was right 3 times, more right than the bottom once, less right twice. Check what happens this season. So far top looks like better 5 times, worse 5 times.

Injuries account for some of the biggest shortfalls but you could argue that coaching played a big role in 7 biggest shortfalls, some I'd say where predictable (Thomas, Carlesimo), some not- at least as big as occurred (Skiles, Johnson, Riley) and some in the middle though maybe lean toward betting against (Dunleavy, Iavaroni).

I didn't give proper weight for the impact of the loss of Artest on defense for the Kings. I knew I wasn't but I went ahead anyways. I didn't think it would be this bad and it certainly isn't all him but he was a pretty big piece.

[Mountain]

As a test of the value of small samples or at least the value to me I will re-predict the season based on first 10% of season:

W East West W
57 Bos LAL 65
51 Det Uta 57
43 Phi Phx 53
46 Ind Hou 53
44 Tor Dal 44

50 Orl NO 58
56 Cle SA 46
43 Chi Por 43
28 Was Den 42
46 Atl GS 33

23 Cha Min 29
25 Mil LAC 23
23 NJ Sac 34
42 Mia OKC 18
33 NY Mem 22

Maybe now makes more since than a month ago.

If any of the systems wants to re=project after 10-20 games that might be worth seeing and this step might represent "progress" or not ("cheating" / changing from pure prediction ) depending how you want to view it.

[Mike G]

[Ed Küpfer]

[Mike G]

[Ed Küpfer]

[Mountain]

"There is not much data in 7 games"...
but if it pulls my average error down significantly there was enough for me to apply judgment to and revise prior judgments productively.

[Ed Küpfer]

I'm not sure what you're trying to say. I thought this was an experiment to see how well small sample forecasts did. I posted my forecasts based on one of my prediction models. Now we wait 5 months to see how close we were.
_________________
ed

[Mountain]

David's Sparks research boils down to me to advocating use of above average levels of Scorer's Opposites, Perimeter Scorers, Pure Perimeters and Pure Interiors.

Boston did all four with the first two in the star roles.

Last season Lakers and Houston did the first 2 but not the second two- at team aggregate minutes used level

http://spreadsheets.google.com/pub?key=pjtolzxemBV4dHRx1PMkaPw

A full dose of Gasol Bynum will get Lakers to 3 and probably runaway leader on Scorer's Opposite. Fisher-Farmar isn't going to get you above average on pure perimeter,

A healthy Ming isn't going to help Rockets with Pure Interior as he an Interior Scorer and Dikembe isn't either. But not everybody has to go pure interior.

Artest is a good type mix but may not change the Rockets standing compared to the 4 averages much on his own. But maybe Alston, perhaps a more mature or somewhat re-directed Brooks (if tried, if it works), Barry - if he plays enough- McGrady -when he plays that way- and Artest could achieve the Pure Perimeter need. Not very traditional or straightforward path to solving this need but time will tell if it is enough.

A lot still rides on Alston to distribute more often and smarter than last season but instead his assists rate is down 20%.

Of 52 guards 6-3 or less who've played 100+ minutes so far this season Brooks has the 9th lowest rate of assists per minute. Folks are busy praising his shooting and that is fine but I'd focus on what he is not doing as much or more. The scoring punch off the bench argument is valid when it works but ignores the times that doesn't work and all the times he doesn't lift others by strong passing. But I guess Houston has McGrady or Barry or both on the court with him so maybe the onus of passing is designed to fall on them. Total team assisted % is quite low with Brooks on. not that much different from when off, but compared to other good teams it is about as low I could find on quick partial check with only Lowry found to be lower.

Suns were at 3 last season probably before and after Shaq-Marion trade but getting scorer's opposite Barnes and using him a lot may have been very smart by Kerr and could put them at all 4 fully above average.

Golden State 2 right, 2 wrong. Davis gone; at least they have chance to add pure perimeter- Nelson or otherwise. Don't have a predominantly scorer's opposite.

Spurs 3 of 4 but lacked pure perimeter. That isn't Parker predominantly and Vaughn is not enough of a balancer and Mason and Hill probably aren't enough either.

Hornets 2 right, 2 wrong. Posey will help with both weaknesses and was another sharp choice by Bowers, perhaps even at the price which never stopped him before on the key calls.

2 right is pretty common. The now Thunder got 2 right.

Portland was at 1 last season. Though they probably move to 3 with Rudy and Oden.

Dallas was at 0 technically though close on 2 and Kidd would make 3 if those other 2 were tweaked up a little.

Denver was at 2, moves to 3 and isn't far off on scorer's opposite.

Utah at 3 and needs a Pure Interior.



Despite the poor record Miami got the types right in 3 of 4 cases.

So did Atlanta lacking only the Pure Perimeter which is why they went for Bibby and though I was not a fan of the trade last season he is playing better right now and the Atlanta surge may be due to the missing ingredient being there and now performing better.

Philly had 2 and picked up a 3rd check mark getting Brand. Miller is not a pure perimeter so that box remains empty by this method of judging it.

Cavs had Pure Interior and Perimeter Scorer out the wazoo but lacked Pure Perimeter and Scorer's opposite. Mo Wiliams doesn't help the pure perimeter need much but a healthy more heavily used Varejao would address the Scorer's Opposite need.

Nets had 3 of 4. Is Yi a Scorer's Opposite? Modestly. Will see how it evolves.

Orlando and Toronto at 3 and holding (unless Oneal plays more like a Scorer's opposite or Bosh does in the presence of ONeal or maybe the combo is enough).

Detroit was at 2 and with the trade stays at 2 but shifts which.

Chicago at 2 but has the hard to find scorer's opposite in spades, in fact higher than anybody- by minutes.

Indiana at 2 then, should move to 3 with the bigs acquired and not far off 4.

Washington good on 3; needs scorer's opposite. Blatche fits the type. Wait n see on McGee.

Having the right types and using them heavily is much but not all of the challenge. Choosing how to use them with each other and the rest of the rotation is another part and David Sparks' type pair research helps with that thinking some too. Of course good productivity is very important, above just getting a type distribution "right"; but you need to ensure that your productivity with other types is not hindering other players the way the overall averages suggests it is generally for pure and interior scorers or over-duplicating certain elements of an elite well-rounded hard to defend or defeat team and leaving important gaps.

[eyriq]

Great stuff Mountain, I really like that website the Arbitrarian. I get how to find the raw data to determine a players style, but I failed to catch how I'd create an index from which to weigh the calculations. Could I just use the league average stats?

Also, when you listed how many of each type each team had, was that from your own work or is it listed somewhere? Thanks.

Just from Davids quick (as he put it) study, it was interesting to see common wisdom upheld that playing inside out is the most effective. I'm looking forward to seeing how the Magic players measure up with this.

[eyriq]

I'm still not sure I have the appropriate index, but tentatively speaking it seems that the Magic in Hedo have a SP and in Dwight have an II. Nelson seems to make the cut for a PP, but that doesn't intuitively make sense, lol.

[Mountain]

Thanks eyriq.

On your first question not sure exactly what you are trying to do but I think I would mainly refer you back to David's work (corrected /better link)
http://arbitrarian.wordpress.com/2008/07/08/nba-playing-style-spectrum/
and
http://sonicscentral.com/apbrmetrics/viewtopic.php?t=1833&postdays=0&postorder=asc&start=30
If you wanted to make a table of the data from the carts in the second link (or found a previous one) that would be handy for you, me and others to refer to.


And then prior type split for the Magic players:

http://spreadsheets.google.com/pub?key=pjtolzxemBV7RdQhUfnWbTA

Howard Scorer Opposite - Interior Scorer
Turkoglu interestingly is a Perimeter Scorer first, Interior scorer second, Pure Scorer third.
Nelson decidedly more Scoring Perimeter than Pure Perimeter.


On the second question, I just sorted/filtered the first google spreadsheet I listed as a link for the 2007-8 team type data. And it is minutes played by players with performance at the time similar to a type rather than quantity of players though obviously that will influence it.

[eyriq]

Thank you very much Mountain.

By Index I meant a norm to which I could compare the individual player results, like for last year Nelson was 47%Scorer, 34%Perimeter, 19%Interior, so compared the the norm that would make him.... when I took the league average it was 52%S, 16%P, 30%I, so since he was under the Scorer norm but over the Perimeter norm by more than half I though PP. But that is team stats for the entire league, all positions, and so on so I wonder if that is how an comparison point is meant to be found. Maybe it would be more helpful to do it with stats from only other pg's, or I don't know. But yeah, that is my question, how do you build a scale of comparison?

I love the spread sheet! What does MEV represent?

I have to say that finding this site has been great, I'm learning so much just reading through here. Sometimes I get lost in the acronyms or some of the more complex ideas, but that comes with the territory. A great resource, in book version even, would be a dummies guide to understanding, interpreting, and utilizing all the great metrics available. I'm sure there are people out there like me who would love to be mini-statisticians, or really just more advanced observers and commentators, when it comes to their own favorite team.

[Mountain]

Gotcha, now.

"Maybe it would be more helpful to do it with stats from only other pg's"

Yeah I agree, but I don't think that has been presented. The third link I gave, if, for example, sorted for just 2008 and then coded for position and averaged could be the basis for such a type split by position index.

If I have forgotten a spreadsheet where David already did or more easily allows for this I hope he'd take a moment to point it out to us.

If you assemble and share such an index that would be quite interesting to see.


"What does MEV represent?"

David's Model Estimated Value.
I'll leave the description to his original work.

As for reference tools you could check out Basketball on Paper or some the shorter internet resources listed in the linkage thread. There is also Wages of Wins (and the forthcoming II). Maybe another book introducing and / or advancing modern basketball analysis gets written in the future.