So exactly three of your home/aways had big swings? Wisconsin in real life last season:

vs. Indiana
Home - win by 11
Away - lose by 3
Home Difference = +14

vs. Illinois
Home - win by 25
Away - win by 12
Home Difference = +13

vs. Minnesota
Home - win by 8
Away - lose by 13
Home Difference = +21

That being said, I also can't recall differences so big in HD, but in real life they are not unrealistic outcomes.

I would never want to play in the badgerdome in HD or real life... Mully got dat on lock

If you want to show a convincing argument, and I think there probably is one to be made, we definitely need a bigger sample.  Honestly, that might be a fun project, I'd like to know about how valuable HCA is.  

Posted by tkimble on 9/26/2014 12:44:00 PM (view original):
If you want to show a convincing argument, and I think there probably is one to be made, we definitely need a bigger sample.  Honestly, that might be a fun project, I'd like to know about how valuable HCA is.  

I looked at this across three of my D2 seasons a while back, and I was seeing a median of about 6 points of HCA swing (meaning going from home to away made a 6 pt difference). But there was a lot of variance in that sample, which as backboy pointed out is exactly what you see in RL too. Also, I remember seeing a study on Prediction Machine's blog a couple of years ago that the swing in RL D1 was 7-8 points, so what's happening in HD seems pretty reasonable.

Three seasons is still not a huge sample, and obviously there are a ton of other factors (player growth during the season, gameplan adjustments by both teams, etc), but at least it's worth something, unlike mully's "data".

not sure how i feel about this, but just for comparison, another team in the opposite division, same conference:

northwestern
home: +15
away: +21
HCA: -6

indiana
home: +25
away: +38
HCA: +13

purdue
home: +8
away: +29
HCA: -21

illinois
home: +6
away: +10
HCA: -4

penn st
home: +10
away: +3
HCA: +7


well, my guess was our case would be less severe, because i know beyond reasonable doubt HCA is not as important as your very small sample size would suggest. i didn't expect it to be a great counter example, but i'll take it? overall, 3/5 we had negative benefit from HCA, and the average of the 5 differences is -2.2

i am stuck saying the same thing i say 75% of the time you make these statements - its not that i necessarily disagree with the premise, but the example shows nothing just on the raw sample size, and beyond that, there is no way it is significant as your example shows. you have to take these over a longer stretch to find something meaningful, picking out the outliers does nothing unless its part of an analysis on variance (standard deviations).

on this particular issue, i agree, HCA in high d1 is too meaningful. but its fairly subtle, i was sort of on an HCA kick for a while, did some research so i could demonstrate how far off it was. but compared to real life, it really didn't come out much different, maybe a point or two points tops. i think the real issue is the HCA spread, HCA figures in d1 are kind of nonsense, too many A+s and such, where there without-question should not be, winning is too big a factor. in the end, i came to the conclusion that while i personally don't like HCA meaning as much in HD as it does in RL, thats just an opinion, but if you try to model them the same, WIS did a pretty good job - on the value of HCA. its how they figure your HCA that is crap. that's just my take... but far more analysis went into that than 5 games :)

So the answer to the original post is "A", engine variability.