Attributes, What's the Deal With? (526 views)

[eric]

So for another project I'm working on I tabulated the last 231 completed skill camps. We only get output in the form of ratings, and some attributes affect more than one rating, but if we add up enough we might be able to get a feel for the functions involved, so here goes.

Overall there were 287 points of improvement over the 231 skill camps, each worth 5 points a piece, so the overall worth of an attribute point is 287 / (231 * 5) = .25 rating points. With 15 attributes and 5 ratings I would have guessed .33, but no. You need 3.36 rating points on average to see a grade change, so you would expect to see a grade change every 3.36 / .25 / 5 = 2.71 skill camps, so in a 231 skill camp sample we would expect 231 / 2.71 = 85 grade changes, and we in fact saw 81 so that's looking pretty good.

231 might sound like a lot but obviously the camps weren't randomly distributed:

n i o h d r attribute
41 59 0 0 0 0 inside scoring
62 0 74.5 0 0 0 jump shot
11 0 10.5 0 0 0 3 point shot
10 0 0 6.5 0 0 handling
12 0 0 6.5 6.5 0 quickness
8 0 0 16.5 0 0 passing
1 0 0 0 0 0 stealing
4 0 0 0 0 0 shot blocking
11 0 0 0 10.5 0 post defense
13 0 0 0 7 0 perimeter defense
3 0 0 0 0 0 drive defense
3 0 0 0 0 6.5 offensive rebounding
6 0 0 0 0 10.5 defensive rebounding
44 33 0 0 20.5 16.5 strength
0 0 0 0 0 0 jumping

Careful observers will not that there are only 229 camps listed here. This is because one Strength and one Quickness camp resulted in a grade improvement without an Attribute listed. Inside, Jump Shot, and Strength made up 64% of the upgrades done, and the sample size for the others are too small to take too seriously. We can however sum everything from "stealing" to "drive defense" under the meta-attribute "defense", as they all affect only the defense rating, and look at the bang for the buck figures for those four:

i o h d r all attribute
0.29 0.29 inside scoring
0.24 0.24 jump shot
0.11 0.11 defense
0.15 0.09 0.08 0.32 strength

Either people had a terrible run of luck on their 32 defense upgrades, or (all) defensive upgrades are really poor return on investment. Strength appears to give about as much to defense PLUS inside scoring PLUS rebounding.

I'm going to tabulate reward camps too but because they're smaller and likely concentrated in the same areas I don't anticipate the conclusions changing much while I'm working on this project. Feedback welcome! :)

[KwYawnza]

Jumping is the loneliest attribute

[kn88]

great stuff

[eric]

Tabulated 725 reward camps, and if the methodology is sound then we should expect similar results. Overall there were 531 grade increases, and 531 / (725 * 3) = .24, ohhhh yeah. :) Now, unlike SC sometimes multiple RCs can be performed on the same player in the same post, and if there's an increased rating that takes into account more than one of the attributes there's no way of knowing which one did what. I can't just give each attribute half of the credit, because we know some attributes factor more heavily than others, but leaving those camps out entirely will introduce a selection bias. I went with the second option, and here's how it looked:

n i o h d r attribute
107 74 0 0 0 0 inside scoring
170 0 139 0 0 0 jump shot
19 0 27 0 0 0 3 point shot
26 0 0 7 0 0 handling
29 0 0 3 10.5 3 quickness
37 0 0 40 0 0 passing
9 0 0 0 0 0 stealing
25 0 0 0 7 0 shot blocking
43 0 0 0 17.5 0 post defense
45 0 0 0 30 0 perimeter defense
21 0 0 0 7 0 drive defense
12 0 0 0 0 0 offensive rebounding
29 0 0 0 0 19.5 defensive rebounding
113 25 0 0 27 26.5 strength
1 0 0 0 0 0 jumping

i o h d r all attribute
0.23 0.23 inside scoring
0.27 0.27 jump shot
0.14 0.14 defense
0.07 0.08 0.08 0.23 strength

Of the 19 multi-attribute +rating camps, 13 involved Strength and 11 involved Inside Scoring, so it makes sense that those are the values depressed compared to the SC table. Let's throw them all together and see what we find:

n i o h d r all attribute
148 0.25 0.25 inside scoring
232 0.26 0.26 jump shot
30 0.33 0.33 3 point shot
36 0.11 0.11 handling
41 0.06 0.12 0.02 0.20 quickness
45 0.37 0.37 passing
10 0.00 stealing
29 0.07 0.07 shot blocking
54 0.15 0.15 post defense
58 0.19 0.19 perimeter defense
24 0.09 0.09 drive defense
15 0.13 0.13 offensive rebounding
35 0.26 0.26 defensive rebounding
157 0.10 0.08 0.08 0.27 strength
1 0.00 jumping

The smaller the n, the less seriously we should take the result. By Occam's Razor I'm guessing every rating adds up to 0.50. Assuming that and estimating, we get...

i o h d r all attribute
0.25 0.25 inside scoring
0.25 0.25 jump shot
0.25 0.25 3 point shot
0.15 0.15 handling
0.10 0.10 0.05 0.25 quickness
0.25 0.25 passing
0.05 0.05 stealing
0.05 0.05 shot blocking
0.10 0.10 post defense
0.10 0.10 perimeter defense
0.10 0.10 drive defense
0.10 0.10 offensive rebounding
0.25 0.25 defensive rebounding
0.10 0.10 0.10 0.30 strength
0.15 0.15 jumping

Defense adds up to 0.80, but if we say only one of the post/perimeter/drive defense is relevant to a player then it adds up to 0.50...? I doubt that's true, that's just my guess at the reasoning.

[eric]

It also occurred to me that we can now calculate how much a full regimen of camps is worth.

3 SC and 3 RC = 3 * 5 + 3 * 3 = 24 attribute points.
Spending one each on Strength and two each on .25 categories gives 8 * .3 + 16 * .25 = 6.4 rating points.
6.4 rating points / 3.36 rating points per grade increase = 1.9 grade increases

A full regimen of camps is worth, on average, about 2 grade increases.

[Heebs]

Is this the part of the program where I bitch about Chalky?

[eric]

Okay, so now that we have a good estimate of the relationship between attributes and ratings, we can look at everyone from the 3010 to 3014 drafts and see how they've increased, subtract off the camps spent on them, split them up based on pre-draft potential, and see if there's any correlation. Five drafts might seem like a big sample but of course we almost never complete a full draft and a significant number of guys aren't on any team in the league anymore. All told we get 135 players: 48 A potential, 55 B, 31 C, 1 F (Phil Lovelette) (no relation). The 3014 guys have had 3 training camps, 3013 have had 4, and so on. When we divide all out, we get...

A potential: +4.0 rating points per year
B potential: +2.9 rating points per year
C potential: +2.0 rating points per year

Looks a lot like 4/3/2 to me! What does it mean? Well, if we take an A potential in 3010 vs. a C with the same grades otherwise, over their 7 training camps the A potential will on average have (4-2) * 7 = 14 more rating points, or about 4 more letter increases.

.

There's probably an aging curve as well, and it's probably the case that not all attributes go up or down in sync (three point shooting ages well? quickness doesn't?). But it's still an interesting result imo! :)

[Buster 5000]

There's your 8th 5 pal.