setherick has given some great info, looking "behind the scenes" and sharing with the rest of us. He didn't want his thread getting "bogged" down with details, but rather than flood him with PM's, I figure I'll just give a very simplified version of what he is talking about.

Linear relationship: This basically means a 1:1 or a diagonal line. So the difference between a skill of 1 and 10 is the same as the difference between 91 and 100, net of 9. So the skill "check" of an opposed attribute between 100 and 75, 75 and 50, 50 and 25 would all have the same "result" in the difference or how often it would succeed. This is the "old way" (pre 0.4.5) for most attributes

Parabolic relationship: This is a "U" shaped curved line, the basic formula is y=x^2. What this means is there is a MUCH greater difference between 1-10 than there is between 91-100. To give some easy example points: An attribute (our y axis) of 1 would have an effect (x axis) of 1, an attribute of 4 would have an effect of 2, an attribute of 81 would have an effect of 9, and an attribute of 100 would have an effect of 10. This results in integer changes at the following skill levels: 1, 4, 9, 16, 25, 36, 49, 64, 81, and 100. The result of this is less difference between the top skill points and more difference at the low end. (Going from a 2-3 is seven times more impactful than going from 92-93 for example)

What the above change means to you:
Starting in 0.4.5 leagues you can "bracket" scout players, so if you can only get a 65 speed defender on a 75 speed receiver, you will have less "failures" than before, likewise the difference between having a 100 skill and a 90 skill is akin to about 2 point difference under the old linear system. Bigger is always better, but the bigger the number the less better it is that the number preceding it.

Weighting: Weighting is exactly what it says it is, a weighted average. What this means is whatever numbers you put in end up getting converted to a percentile, with all percentiles equaling 100%. What this means is that not all 100 weights are "equal". Some examples using TE's and OG's (from setherick's 4.5 weights):

Pass catch TE:
100 Speed, 71 B&R, 71 RR, 58 Str, 58 Run B, 45 Courage, 35 Pass B, 32 Acc, 20 Int
this equates to the following formula to give you the OVERALL rating:
.20*Speed+.145*B&R+.145*RR+.12*Str+.12*Run B+.09*Courage+.07*Pass B+.07*Acc+.04*Int
Blocking TE:
100 Speed, 71 B&R, 71 RR, 71 Pass B, 58 Str, 58 Run B, 45 Courage, 32 Acc, 20 Int
this equates to the following formula to give you OVERALL rating:
.19*Speed+.133*B&R+.133*RR+.133*Pass B+.11*Str+.11*Run B+.09*Courage+.06*Acc+.04*Int

LOG
100 Str, 71 Pass B, 45 Run B, 45 Acc, 20 Int
this equates to the following formula to give you OVERALL rating:
.36*Str+.25*Pass B+.16*Run B+.16*Acc+.07*Int

ROG
100 Str, 71 Pass B, 58 Run B, 45 Acc, 20 Int, 12 Speed
this equates to the following formula to give you OVERALL rating:
.32*Str+.23*Pass B+.19*Run B+.15*Acc+.07*Int+.04*Speed

What this means to you:
The more weights you put up, the less effect each weight has overall is one take. Another is that if all your weights equal 100 it would end up being the same, just less "flexibility" in tweaking. If you want to have a "true relative" OVERALL rating of ALL your players, the TOTAL of all weights would need to be the same. Whether that is a total of 100, 300, 500 or 1200 it wouldn't matter. Now if having that data point is important or not is a very subjective discussion.

I think I hear MFN 1 calling you.

here is MFN 1 calling for him = https://mfn1.myfootballnow.com/team/1

Great post!

For all of you interested, and don't want to do the math yourself, here's the very basic formula I used to come up with my weights if you want to make your own.

The thought here is that I want one or two primary attributes to be equal to 100, and then each other attribute to be a weight based on the effectiveness of the secondary attribute compared to the primary attribute.

The formula then breaks down like this:

Attribute Weight = 100/(100-Effectiveness)^0.5

So if I want a secondary attribute to we weighted at 90% effectiveness to the primary attribute, you get this.

Attribute Weight = 100/(100 - 90)^0.5 = 32

Or if I want it to be weighted at 98% effectiveness as an extreme example:

Attribute Weight = 100/(100 - 98)^0.5 = 71

That 98% effectiveness really shows off how a parabolic weight system differs from a linear scale.

Are all ratings parabolic ? Or just certain ones?

The only two that I know for sure are parabolic in 0.4.5 are Accuracy and Speed. Basically, < 50 Accuracy is unplayable. And < 50 SP, especially for RBs, is really slow. You can get away with <= 50 SP at positions like TE, but they will not run away from anyone.