Warrior General Games

EDIT: I was close, here is the final version.

0.4 * blocking + 0.2 * strength + 0.1 * vision + 0.1 * agility + 0.1 * confidence + 0.1 * weight/8

-------------------------




Proposed formula:



For the blocking bar, something along the lines of...

0.2*STR + 0.42*BLK + 0.1*AGI + 0.1*CONF + 0.07*VIS + 3
or
0.21*STR + 0.46*BLK + 0.1*AGI + 0.1*CONF + 0.07*VIS



I'm still looking for data for the overall bars, and over the next week I'll be using marketplace to get data from other positions as well to determine a few more bars at other positions.



--------


NO EQUIPMENT STATS PLEASE


----

Our goal? To determine the formulas for the overall and blocking bars for the OT, G and C positions. If this works, I'll extend the project to all positions.


Assumption: The scouting bars are a linear function of your player attributes.


We'll probably only need a handful of data points - nothing too extreme, but data across both low and high levels will help.

The MOST helpful data will be what happens EXACTLY when the scouting bars moved. But anything well help for now.

If you want me to keep your build private, you can PM me also, I'll just put it in excel. I have no personal motives knowing your builds - as you can see, I have a bazillion o-linemen already, but I'll respect your privacy wishes if you want.


I've got a few statistical programs that should do this really easily. I'm just going to enter your attributes as vectors and run a multivariate least-squares regression on it, and it'll spit out some coeffecients.

I've got like 10 guys already, but I'll just post mine


OT

Overall 51
Blocking 57

Strength 66.3
Blocking 64.63
Agility 58.49
Vision 44.41
Confidence 34.97
Stamina 41.07
Speed 40.34

Jumping 8
Tackling 25.88
Throwing 9
Kicking 9
Punting 9
Catching 9
Carrying 8

This seems interesting.

The assumption has to be a weighted combination of base attributes. If the weights are different for each attribute it's going to be difficult to tie them down.

It is possible that major- and minor- attributes have different weights only, I'm going to build a high confidence G next season so I will keep you posted with what happens to his bars.

Also when my lvl 40 G increases his overall and blocking bars I'll PM you his exact stats at the time he reaches the new threshold. Hope this helps.

http://goallineblitz.com/game/player.pl?player_id=931483
His build will stay open for a while.

Oh, don't worry, all the weights are different and will be calculated accordingly. Software does it for me, but it seems I'll need about 30+ data points for each position for it to mean anything.


I'm guessing its based on major/minor as well, but its not that simple for all positions. For example, confidence doesn't move the punting bar nearly like punting does, despite being a major attribute as well. And stamina moves the overall C bar more than vision, etc.


Hopefully if we get good participation, this will be pretty easy to crack.

http://goallineblitz.com/game/player.pl?player_id=687021
(Build Open, here's naturals - bars are 45/54)
Strength: 65.9 Blocking: 66.03
Speed: 19 Tackling: 21.52
Agility: 60.65 Throwing: 9
Jumping: 8 Catching: 9
Stamina: 29.52 Carrying: 9
Vision: 26.44 Kicking: 9
Confidence: 26.2 Punting: 9

Here's my Center's naturals (bars are 42/57):
Strength: 75 Blocking: 73
Speed: 8 Tackling: 14
Agility: 49 Throwing: 9
Jumping: 9 Catching: 8
Stamina: 25 Carrying: 9
Vision: 22 Kicking: 8
Confidence: 18 Punting: 9


And my Guard's naturals (bars are 45/54):
Strength: 68.94 Blocking: 66.06
Speed: 10 Tackling: 15
Agility: 59 Throwing: 9
Jumping: 9 Catching: 8
Stamina: 26 Carrying: 9
Vision: 21 Kicking: 10
Confidence: 28.4 Punting: 9

OT - 54/60


C 51/63

Guard 45overall, 54 blocking
Strength: 67.94 Blocking: 62.97
Speed: 23.17 Tackling: 15.57
Agility: 48.7 Throwing: 8
Jumping: 8 Catching: 8
Stamina: 29.92 Carrying: 8
Vision: 22.1 Kicking: 8
Confidence: 34.57 Punting: 9

http://goallineblitz.com/game/player.pl?player_id=681949
cleared

Got everybody so far.

I think it'll take a while to get the amount of data I need, so if you boost and level and put 20 points in and something moves, let me know. Though too much dependent data will screw stuff up, I'd like to keep "same build" data down.


Anyway yeah, I'll be sure to publish results in the GLB forum and here, but expect it'll take a while to get the volume I need. I'll post the formulas to let players test and whatnot, and if it worked I'll run a similar project GLB-wide.

Originally posted by Octowned
Oh, don't worry, all the weights are different and will be calculated accordingly. Software does it for me, but it seems I'll need about 30+ data points for each position for it to mean anything.


I'm guessing its based on major/minor as well, but its not that simple for all positions. For example, confidence doesn't move the punting bar nearly like punting does, despite being a major attribute as well. And stamina moves the overall C bar more than vision, etc.


Hopefully if we get good participation, this will be pretty easy to crack.

Sounds good. I imagine you'll need 100+ data points as the scouting bars are only to the nearest 3, but it should be possible.

One thing that is worth thinking about is how the weighted sum of attributes changes the scouting bar. For example, is a bar of 60 from 57-60, 58.5-61.5, or 60-63? One thing that does intrigue me is that the blocking bar goes to a light green 60 first, then it changes to a dark green 60. This would indicate to me that the bar is light green 60 when the weighted sum is 58.5-60, and dark green 60 when the weighted sum is 60-61.5.

If you ever need a hand interpreting the dataset I could put together some sort of monte carlo method quite quickly to get suitable ranges for the weights. Of course, the more data there is, the acceptable weights will have smaller ranges. This method would work but would be messy - I'm struggling to think or a nicer regression method at the moment due to the number of degrees of freedom and also the 'to the nearest 3' nature of the bars. If you already know what software to use though that is great.

Once we crack one or two positions I expect we can use this to quickly crack the others.

I'm not doing anything fancy. I'm a statistics masters student now, but haven't really learned much applied yet, testing my knowledge.

I'm doing a multiple linear regression with the intercept at 0.

The bar colors I think you nailed.

One problem I'm having is that the data is quite dependent. With only 4 points and 8 parameters I'm using right now (tossing catching, etc., for now), I guess any initial results will be unreliable

C
Overall 51
Blocking 60

(without equipment)
Strength 77 Speed 25.54
Agility 62.81 Jumping 8
Mental Attributes
Stamina 34.89 Vision 31.11
Confidence 30.19
Football Skills
Blocking 74 Tackling 15.2
Throwing 8 Catching 8
Carrying 8 Kicking 8
Punting 8

C
overall 48
blocking 60
(without equipment)
Strength 77.24 Speed 18
Agility 54.45 Jumping 9 (+0) +-
Mental Attributes
Stamina 35.56 Vision 19.45
Confidence 19.45
Football Skills
Blocking 76.29 Tackling 14.45
Throwing 9 Catching 8
Carrying 8 Kicking 8
Punting 8

This only works if you get attributes without equipment. Equipment does not factor into scouting bars.