Saturday, October 31, 2009

How to not use stats properly

A lot of people have been complaining about the shot ratings this year, especially the Close ratings. The best way to descibe close is it's the 4-14 foot range. Inside is the direct vicinity of the basket, mid-range is from the FT line out. Close is that weird inbewteen stage where you'd prefer to get closer to the basket if at all possible, and the region is staunchly guarded which makes it difficult to score there.

Since most players (especially guards) would prefer not to shoot from this range, it is common for players to have few shot attempts from this range. What that creates is a small sample size, and thus the opportunity for stats to be mishandled.

Eddie House has the following shot ratings

Inside 81
Close 50
Medium 86
3pt 87

House has 80s in all his shot ratings but the minimum 50 in his close rating. 2K Sports didn't take player history into account, which is pretty darn important when using a small sample (as it increases the pool of data to get a better idea. Most of you know this by now but I'm just laying out the details for any newcomers.)

Eddie House Close
2009: 9-30 (.300)
2008: 14-34 (.411)
2007: 12-41 (.292)
2006: 13-38 (.342)
2005: 23-43 (.534)
Total: 61-186 (.327)

What we have here is a player with a fluctuating percentage. He's as low as 29% one year and as high as 53% in another. House gets slapped with a 50, which is more due to his low sample size (as it seems 2K gave a 50 to any player with a small sample and a low percentage). Just going by this data we know House does not deserve to have the bare minimum rating of 50. I'm still collecting data for players, but it should be obvious House's numbers are better than say...

J. Jeffries
2009: 1-15 (.066)
2008: 2-12 (.166)

2007: 6-20 (.300)
2006: 7-29 (.241)
2005: 10-35 (.285)
Total: 26-111 (.234)

There's a player much more deserving of a 50.

Let's say House deserves a 55 for arguments sake. That still looks very odd next to an 81 Inside and 86 Medium. So to inspect further, let's take a look at those Inside numbers...

E. House
2009: 17-30 (.567)
2008: 12-35 (.343)
2007: 18-35 (.514)
2006: 30-67 (.448)
2005: 17-42 (.405)
Total: 94-209 (.449)

2K gave House an 81 rating because he went 17-30 over a full season. That's actually pretty pathetic when you think about it. How could they not know what a bad finisher he is? He shot 34% in a similar sample just a year prior, and has only been above 50% two times in five years. That 81 is just two points below...

J. Jeffries
2009: 92-175 (.526)
2008: 83-169 (.491)
2007: 70-126 (.556)
2006: 142-261 (.544)
2005: 171-318 (.537)
Total: 558-1049 (.531)

It should be fairly obvious which of these players is the significantly better finisher. Jeffries had a better year last year than House's last 5 seasons combined. What we see here is House's inside rating doesn't match up with his dreadful close rating.

61 inside
55 close
86 medium
87 3pt

Doesn't that look so much better? His strengths are clearly illustrated now, and we didn't even have to put any thought into it (2K hates doing that). All it took was merely analyzing readily available data.

But then, I guess a company that didn't update any FT ratings between this year and last has no use for somethng like that.


jonathon said...

That's awesome rashidi, I don't think I have read a better article about basketball than that right there, and I think anybody could understand that breakdown if you didn't know basketball that well.
Superb job man and much props when it comes to your knowledge of the game, It's amazing how you understand the rating system better than that joke of an insider who to me is just lazy, and the company who designed it.

Anonymous said...

Much respect to your hard work. They should be paying you cause wit out your rosters my 2k10 would be back in gamestop.

Rashidi said...

I don't think it was the greatest article. I certainly could have used more player examples.

I try to limit the length of my articles because the more time I spend on them, the less time I'm editing. There is still quite a bit of work to be done.

I just try to pick and choose my battles and try to keep people in the know and usually am writing about something I'm working on.

I spent some time collecting close/inside data now that hotspots is back and I suppose this article is my way of saying I'll have improved ratings in my next update because I now have that info at my disposal.

Anonymous said...

yeah I´m not an english people but I try it:

I like your blog very much, because it´s very competent written and I have a lot of fun during reading it, but why don´t you do some rosteredits for ps3 ;-).

I think 2k needs to jump over their shadow (don´t know if it´s right in english^^), and give you the job of the mo´fucking hoe 2kInsider, so they will have good rosters and you can earn some money with it.