NBA 2K14: Finding a "Shot Medium" Rating Formula - iTofu's Blog

iTofu's Blog

NBA 2K14: Finding a "Shot Medium" Rating Formula

Posted on July 13, 2014 at 05:22 AM.

(0)

This is a post in a series of blog posts describing the process of player modifications, ratings changes or signature skill changes that went into my PS4 NBA 2K14 Offseason Update Roster. In my opinion, trying to analyze NBA 2K14's shot medium rating was both the most interesting and difficult to understand.

When trying to duplicate 2K's rating formulas for medium range shot and 3 point shooting to update the current roster's ratings, I thought the best method would be to disconnect my PS4 from the internet, then look at the original roster and see how that compared to the 2012-13 season.

I took all players who attempted more than 100 3 point shots and 100 medium range shot shots in 2012-13. I used all shots 15 feet or longer that were for 2 points as being medium range shot, because when you look at the hot spots in free style practice the medium range shot hot spots start at the free throw line.

As seen in the three point post, the three point percentages and ratings correlate well, just pluck out the outliers and you have a pretty solid formula:

However, when looking at medium range shot shots, there wasn't a clear correlation:

That confused me considering how tight 3-point was. I chose 15 feet, because when in freestyle practice mode the medium range starts at the free throw line. I thought perhaps 2K used a different range. I tried 10+ feet 2 pointers, 12+ feet 2 pointers and 17+ feet 2 pointers with roughly the same result.

Then I noticed something, Ryan Anderson who was a ridiculous outlier in each instance (seen in the graph above as a 60 rating shooting a percentage of .443).

Then, I wondered, what if 2K never updated the ratings? So, I ran 15+ feet from 2011-12. The correlation was a little better by not significantly:

Upon closer inspection, it appeared as is some were updated and some weren't, perhaps 2K hadn't updated all the ratings. So I data fitted a formula to all the data points, then calculated which year seem to correlate better for each. Then took that data, fitted a formula and calculated which year seem to correlate better for each player again. I repeated this process until the each player couldn't be correlated any better:

That correlates better, but it's still not at the level that 3 point shooting correlated. I took the best fitting percentage of the past 3 years, which graphs out like this:

This was were I settled for finding a formula.

I know what you're probably thinking, "Logically, the data will always correlate better when you choose the best year's percentage for correlation. lol"

Just from looking at older 2K does not update medium range shot ratings for players consistently the way 2K updates the 3 point rating. To go back to the Ryan Anderson example, these are his medium range shot shots over the past 4 years:

2009-10: 10/30 33.3% NBA 2K11: 75
2010-11: 6/22 27.3% NBA 2K12: 75
2011-12: 16/50 32.0% NBA 2K13: 60
2012-13: 78/176 44.3% NBA 2K14: 60
2013-14: 32/67 47.8% End of season: 60

I'm guessing 75 may have been an educated guess. When 2K13 was released, previous seasons, particularly 2010-11 indicated Ryan Anderson didn't take many medium range shot shots and shot poorly when he did. In 2012-13 signed with New Orleans and Ryan Anderson's sample size grew as well as his shot percentage at 44.3%, yet he maintained his rating of 60 while players of a similar percentage and sample size were in the low 80s neighborhood. Despite continued success in the medium range shot his rating is still 60.

Of course, he is the greatest outlier, so it's possible he's an anomaly. Here are more examples with similar situations:

Martell Webster (74/69 is an outlier given low 40% range)
2010-11: 29/68 42.6% NBA 2K12: 74
2011-12: 21/52 40.4% NBA 2K13: 69
2012-13: 42/104 40.4% NBA 2K14: 69
2013-14: 36/86 41.9% End of season: 69

Chris Bosh (89/90 not a significant outlier for 50%)
2010-11: 190/413 46.0% NBA 2K12: 89
2011-12: 113/277 40.8% NBA 2K13: 89
2012-13: 186/360 51.7% NBA 2K14: 90
2013-14: 150/299 50.2% End of season: 90

The ratings not changing on yearly basis seems to be pattern among most of the people I spent time looking through. That being said, there are players who display more confusing ratings, such as Arron Afflalo:

Arron Afflalo's medium range shot (66-76 ratings are an outlier for low 40% range):
2011-12: 57/138 41.3% NBA 2K13: 76
2012-13: 105/247 42.5% NBA 2K14: 66
2013-14: 148/340 43.5% NBA 2K14 End of season: 74

In any case, clearly the ratings are sporadically updated, unlike the 3 point rating. In addition, when the ratings are updated the ratings is not nearly as tightly correlated to the shooting percentage like 3 point rating.

One possibility for the correlation not being as tight is that 2K is adjusting for shot volume. When I looked at who was above and below the expected rating, high volume medium range shot shooters tended to be above. I am not sure whether that was on purpose, coincidence or star treatment (most high volume medium range shot shooters happen to be stars). Certainly that would make sense, good medium range shot shooters will tend to shoot more and defenses will adjust to it thus requiring a higher medium range shot rating. However, that does not fit the logic 2K uses with their 3 point formula.

Another possibility is that 2K is aware of the data but does not have a formula the way the 3 point rating does, thus the correlation will never be as tight.

Whatever the situation, I think this is an area with an opportunity for improvement. In this era of roster connectivity, continually updating ratings that come directly from statistics should be easy and not done manually.

As I mentioned just after the final graph, I ended up using the 3 year graph, I trimmed up the outliers and settled on a formula of:

112.39 * MedShot% + 34.923

Similar to my other ratings changes, I generally used 2 parts 2013-14 and 1 part 2012-13. I felt that was a good way to temper the results if this season or last season was an outlying year without weighting 2012-13 too much. In addition, if a player performs better in the playoffs, I added the playoff statistics. I say, generally, because I did make exceptions for players with unique situations, such as Derrick Rose.