share

Manipulation of Statistics

TheArtistPaysth...
TheArtistPaysthePrice's picture
Registered User
Joined: 02/23/2012
Posts: 1089
Points: 2925
Offline
Manipulation of Statistics

Advanced metrics and statistics are taking over NBA decision making and I think its a move in the right direction. I think its moving way to fast but still the right direction.

What if a team plays the first and second unit separately? Would this destroy everything except for the eye test? Guys who may be role players with niche skills might be forced to create and be outside of what they could be effective at. If you primarily play with only better or worse players then you can manipulate or even nullify advanced stats.

Last year whenever Westbrook was out the game, Harden was always in so this corrupts a little bit of how well he could possibly be with a ball dominant pg or how well a ball dominant pg could play with him. So how well he can play him Lin could have been easily over valued. This is a okay example but there are countless others. Like this year the cav's through the first 7 games played the entire second unit as a unit. Kyrie Irving hasn't played a minute with more than one bench player on the court, unless a injury occurred. Donald Sloan (back up pg), hasnt played with Varejao (all around advanced stats stuffer) at all that I can remember.

How does advanced stats rectify this?


larryfisken
Registered User
Joined: 06/23/2011
Posts: 95
Points: 166
Offline
U would not destroy

U would not destroy everything because u adjust for strenght of the oponent. At least thats what i think is true. Also it is hard to judge an individual because it depends on the players ur playing with thats why u often see stats for lineups becuase thats whats have the most impact on the end result the most important stat in the game.

samosas
Registered User
Joined: 09/20/2012
Posts: 255
Points: 475
Offline
i'll take a stab at this one.

i'll take a stab at this one. this situation in particular is going to be extremely dicey and most likely could not be calculated directly. first of all, we would have to have an arbitrary measure as to what constitutues ball dominant. taking > 20% of the shots while on the floor? what about having the ball in their possession for > 15 seconds per shotclock on average? Personally, I think you'd have to get down to some real nitty gritty, and by that my mean situations that mirror playing with a ball dominant guard. So, if James Harden is playing with someone other than Russell Westbrook, but that person is emulating him in terms of possession of ball per shot clock, contested shots, shots per possession, whatever, that's the best way to measure a statistic like that.

In your last example with the Cavs, I don't think that you'd ever glean anything useful from a stat like 'starting point guard increases efficiency when playing with more than one sub'. There's two problems you run into when analyzing large datasets: one, the sample size is so large that everything happens too often and everything seems significant (this happens all the time). or two, the event doesn't happen often enough and it would take another year or two worth of data to get enough events to be considered significant. With the second option you have to ask yourself, if it takes 5 years worth of data to prove this point...is it really proving anything at all?

so, while a lot of these things are very interesting, you can't ever rely totally on math. it can be a very powerful tool in the hands of a good scout/coach, though.

thparadox
thparadox's picture
Registered User
Joined: 06/13/2008
Posts: 438
Points: 513
Offline
What you're talking about is

What you're talking about is a major problem with one particular advanced stat: plus/minus

Plus/minus is less reliable than the other advanced stats. It can be useful if you want to verify the eye test and other advanced metrics, but by itself it's somewhat biased.

PER is also not that great, because it heavily relies on raw production.. which means it favors players who have a very high usage.

Something like Offensive Rating is different. It tracks efficiency in scoring, passing, rebounding etc. per possession. Win Shares and WARP are based on Dean Oliver's "four factors". Anyway, it's fairly complicated, and I can't do it justice here. If you really want to learn about it, google "Dean Oliver Four Factors". Even better, read Dean Olivers book "Basketball on Paper".

Then check out Basketball Prospectus (WARP) and basketball-reference.com (Win Shares)

In any case, there will never be a perfect advanced stat in basketball. As you alluded to, basketball relies very heavily on team chemistry.

samosas
Registered User
Joined: 09/20/2012
Posts: 255
Points: 475
Offline
plus minus is a terribly

plus minus is a terribly misleading stat by itself, your right. i'm sure you can pair it with something to get something meaningful

thparadox
thparadox's picture
Registered User
Joined: 06/13/2008
Posts: 438
Points: 513
Offline
FYI - the problem you're

FYI - the problem you're discussing still exists with the other stats. Especially on the defensive end. For example, it's much easier to defend your man when you have Dwight Howard behind you.

HotSnot
Registered User
Joined: 05/20/2010
Posts: 739
Points: -1421
Offline
@thparadox Its interesting

@thparadox

Its interesting you mention Howard in the +- discussion. So far this season he is way into the minus most games, but improving as of late. When I check the Lakers box scores Howard is usually 10 plus/minus points off the rest of the starters. I'm not sure if its because he's getting run with the second unit while the other starters rest. I just found it note worthy.

thparadox
thparadox's picture
Registered User
Joined: 06/13/2008
Posts: 438
Points: 513
Offline
Yeah, I thought about that

Yeah, I thought about that too. The Lakers are playing terrible defense this year....

one of the biggest problems with plus/minus is sample size. Especially in only about 10 games it's relatively meaningless.

Lebron's Hairline
Lebron's Hairline's picture
Registered User
Joined: 06/07/2012
Posts: 2158
Points: 5969
Offline
I stop looking at all these

I stop looking at all these stats after Chris Paul was ranked in the bottom 5 in defense in 2009; Paul is an excellent defender but they didn't factor in that he had to guard his man and Marco belinellis; who is an atrocious defender

kingofbums
Registered User
Joined: 10/30/2012
Posts: 345
Points: 681
Offline
but understand hes not

but understand hes not exactly a "great" on ball defender...paul will get the steals but he isnt that great moving laterally...at least not nowadays

thparadox
thparadox's picture
Registered User
Joined: 06/13/2008
Posts: 438
Points: 513
Offline
oh

oh really?

http://basketballprospectus.com/article.php?articleid=638

"Chris Paul, New Orleans - As I've written, Paul's net defensive plus-minus has made a 180 this season. Based on watching Paul--and his gaudy steals totals--I suspect last year was the fluke."

Hitster
Hitster's picture
Registered User
Joined: 08/17/2010
Posts: 3529
Points: 2590
Offline
The plus and minus stats are

The plus and minus stats are interesting and on occasions a player could have a very effective game but maybe only get 2 points. Yet the plus and minus stats of certain players could be increased depending on what team they play on for example the LeBron Cavs or how many players has Kobe helped have great plus stats.

An interesting team to look at regarding this might be the Spurs who hve a deep bench, use a few starting level guys from there and will tweak who starts alongside Tim Duncan.

apb540
apb540's picture
Registered User
Joined: 06/08/2009
Posts: 2324
Points: 1811
Offline
?

I'm not sure what you are asking. I mean, yea the players you play with matter for advanced stats, but that matters for other things such as winning, losing, and good ol' fashioned stats.

On the subject of advanced stats, I think this can be beyond usefull for NBA coaches because it can give them the purest assessment of what a player does. As far as being used for scouting, I think that there will never be anything that beats the "eye test". I mean maybe Presti & Co. used advanced stats to find Westbrook but I doubt it.

thparadox
thparadox's picture
Registered User
Joined: 06/13/2008
Posts: 438
Points: 513
Offline
It's definitely useful for

It's definitely useful for scouting. Advanced stats guys loved Dejuan Blair, for instance.... he should never have dropped that far. Another example is Kenneth Faried.

But I completely agree that the eye test is more important for prospects. They haven't matured yet, the sample sizes are lower, and you have to translate the stats to a different league

thparadox
thparadox's picture
Registered User
Joined: 06/13/2008
Posts: 438
Points: 513
Offline
It's definitely useful for

It's definitely useful for scouting. Advanced stats guys loved Dejuan Blair, for instance.... he should never have dropped that far. Another example is Kenneth Faried.

But I completely agree that the eye test is more important for prospects. They haven't matured yet, the sample sizes are lower, and you have to translate the stats to a different league

aamir543
aamir543's picture
Registered User
Joined: 04/11/2009
Posts: 5062
Points: 5541
Offline
Yes, here is a thread I

Yes, here is a thread I created about Metric stats a while back.

http://nbadraft.net/forum/are-metric-stats-useful

And also, according to metric stats, Kevin Durant is killing it on D this year, among the best in the league, much better than he was playing last year.

Hitster
Hitster's picture
Registered User
Joined: 08/17/2010
Posts: 3529
Points: 2590
Offline
When scouting in regard to

When scouting in regard to the draft teams will look at advanced stats as the piece from the Cav's draft room said they chose Waiters as Chris Grant liked the type of shots he took. They had broken his output down further than field goals, 3 pointers etc and really looked at the positions he shot from and decided he fitted their mix.

Likewise teams would look at how well players play different systems and how well they defend them, lets say a play excels on a pick and role offence but the team tracking him plays a half court offence and they see that his half court production isn't so good, so this could desuade them from drafting him.

Although it may not be stats related I always recall a story from a major football championship where the German team had scouts watching their opponents with more than one scout watching each individual player looking for weaknesses and how to reduce their effectiveness. I would imagine that the scouting of draft prospects would be done to this level to help produce the stats and give the sort of feedback that Chris Grant and other GM's had on the likes of Waiters.

RSS: Syndicate content