Wednesday, November 18, 2009

82games = Misleading stats

I haven't used for statistical reference in a few years now, and this is one of the reasons why.

According to them, Rodney Stuckey is the Pistons' starting SG, and Ben Gordon is the starting PG.

They also list Jared Jeffries as a SF/SG when the dude hasn't played a minute outside of PF this season. They honestly think Wilson Chandler is a PF in Jeffries' lineups, or that Gallinari is a SG when all three are on the floor.

This matters because they are logging defensive stats by position, meaning the wrong data is being assigned to the wrong players. Now that "advanced statistics" are en vogue, I have seen many a blogger write entire articles based upon incorrect statistics that they view as infallible.

Such as this one...

The fact is there is no explanation by 82games on the methodology and reasoning behind the stats they collect, and because basketball fans have few other sources we look the other way and are forced to put blind faith into something we know little about.

Right now on ESPN the Mavs are playing the Spurs. Kidd is on the floor with Beaubois. Kidd is being guarded by Richard Jefferson, while Beaubois is guarded by George Hill. Does that mean Kidd is the SG? Or is Richard Jefferson the PG?

I have emailed site owner Roland Beech about the incorrect position data (last season and this season) and have yet to get a response. If you would like to leave your own message asking for some type of explanation, you can do so here:


jk said...

Give Roland Beech a chance. A couple of seasons ago Sam Vincent was giving Jeff McInnis WAY too many minutes and playing him alongside Felton. In these situations, Beech had Felton marked as the PG when he was actually playing the 2. A friend e-mailed him and Beech made the changes without argument.

I think 82games is a pretty small operation. I imagine Beech would be very happy to have additional 5-man unit information from credible sources. How long ago did you e-mail him?

Rashidi said...

Regarding last season's Knicks, months ago. Regarding this season's Knicks, three days ago.

82games might have a low-budget homepage but they're a decent-sized operation. It's not like one man is collecting all that data.

I applied as a data collector about four years ago when they did some expansion. , My job would have been to track one quarter of a basketball game. So you can figure out from there how many people they have working under them.

Rashidi said...

One thing I didn't mention in my post is they don't explain the difference between "Close Shot" and "Jump Shot". Their shot data doesn't matchup with's hotspots, which is what 2K uses to determine their shot ratings (which is why I found it ironic that 2K contracted 82games to work for them).

jk said...

Wow, that is disappointing. I was a fan of 82games and happy for them to get involved with 2k.

Anonymous said...

You know what? You're absolutely right both in your observation and your specific example cited. I use their numbers on my Bucks Blog, and I've noticed they NEVER list Moute as a shooting guard when he frequently guards that spot.

They should categorize them exclusively by defense and height/weight. Thus in your example Kidd and Jefferson are both small forwards. Categorizing them by "division of labor" is way too inexact, and the only relevance of their "position" data is the "counterpart" comparison, which is a defensive observation.