Warrior General Games

I'll be careful when I next assign my 15 points to note the attributes just before and after a bar jumps. This should give greater constraints.

Although then it would be necessary to worry about the hidden decimals ...

I noticed with my guard that when my confidence went from 26 to 27 that my blocking bar jumped 3, and then when it went from 27 to 28 it jumped again by 3. It made me think there was some sort of floor that had a large impact on the bars.

That would be interesting, and totally destroy any analysis. If a linear regression model, or any other sort of summed, exponential, standardized, I dunno any simple models don't work then we'll have to think outside the box a bit.


I've got about 10-15 data sets for each position, if not more. I'm not entering in the catching, kicking, throwing, punting, carrying, jumping. My argument is that even if these DO have an impact, everybody has the same in them anyway (between 8-10), so including them in the formula will be of no statistical significance.

I assume it's a pretty simple model because a lot of other things in this game are driven by a pretty simple model (like XP).

One BIG problem is the high correlation between variables. High str = high blk, meaning a lot of the value of knowing strength already comes from knowing blocking. Anybody who has high disparity between their attributes would help.. a lot!

As it is, I'm finding that for the blocking bar, it is mostly blocking and confidence (not much strength at all) and for the overall bar it is mostly strength blocking agility and stamina. These are at the OT position, from only 12 data points and the ranges are *very* high.

Like I said, the dependence is what's killer for now, so I'm likely to try a "sum" approach where I sum your str/blk, then run it, and try to decompose from there. Once I have more data!

Perhaps you could expand your study to include Fullbacks who have a blocking bar and usually very high STR and low/mediocre blocking.

Well, if we think the blocking bar is the same for all positions... then yes, that would make a VERY large data set for blocking and definitely be worthwhile to include fullbacks.


The overall bar, I'm quite certain, is different for all positions though. Maybe that one is a major/minor attribute formula, and the specialized bars are other formulas that carry across position? That'd be pretty easy to analyze, actually.


I'll keep collecting data for now, thanks for the suggestion.

The fun part about regression is that you'll always gets an answer that fits your data. Then you try to bring in somebody else, and it doesn't work by a mile. As in right now, I have VERY low residuals, its predicting basically perfectly, however some coefficients are negative which is laughable (aka, adding to a stat will lower the bar). Luckily these negative values are all very very small, so I'm trying to just set them to zero and assume they do nothing, but then the fit gets worse (of course).


I'll combine all data for the blocking bar and see if something comes up.

Sounds good.

You have to assume things like kicking/punting are weighted to 0 to progress with this.

I'll play. Though I can't really help with the high blk/low str question.
These are without eq

G 48/60
Strength 73.38 Speed 12.07
Agility 49.28 Jumping 8
Mental Attributes
Stamina 23.93 Vision 29.32
Confidence 52.16
Blocking 73.97 Tackling 16.13

C 36/54
Strength 72.95 Speed 15
Agility 26.4 Jumping 9
Stamina 17.4 Vision 18.4
Confidence 14.4
Blocking 73.04 Tackling 13.4

C 39/57
Strength 74.85 Speed 12
Agility 18 Jumping 8
Stamina 22 Vision 22
Confidence 25
Blocking 74.95 Tackling 14

I'm not even typing in that garbage, including tackling. Tackling is just a function of level, which is also nearly 100% correlated with the sum of your other points. A few things aside (good building, SAs), of course, but I'd rather not have "level" play a part in this.

Tackling MIGHT have a part in the bar, who knows, so I'll give it a try, but anyway.. yeah..


So I'm going to progress later today with the blocking bar, using all G/OT/C data. I've got about 50 points then.

I'd leave tackling in if I were you.

There are some 1st season O-Linemen who added to tackling. I'll try to find one for you.

Originally posted by mandyross
I'd leave tackling in if I were you.

There are some 1st season O-Linemen who added to tackling. I'll try to find one for you.

The reason is OL used to have two bars, blocking and tackling. So, folks tried to raise their tackling bars to be better OL. Bort finally came to his senses and said tackling doesn't help and removed it.

Yeah, that was hilarious. He should get rid of it for P/K imo.

One thing that will help a lot is getting lower-leveled players.


So far, for the blocking bar, we have...

57
42
51
60
21
21
57
57
54
54
51
51
57
54
57
54
57
57
57
54
54
54
54
60
63
54
48
57
60
57
57
57
63
60
60
54
57
57
57
57




Yet the bar goes from like 21 to 75+ in the current market of players, and it would help a LOT to have more data points in the 20s and 30s.

If you have a slowbuild team or something and want to post/PM bars of players who are in their teen levels now, that'd be nice. We all know how slowbuild players build o-line, anyways :-D