and also, NFL VoA github: github.com/gshelor/NFL_...

In the #NFLVoA this is the first week that only current season stats are included in the model so it's interesting to see it doing the same "expand the range of overall ratings values" thing that the CFB VoA did in its first week using only current season data. Hopefully the NFL VoA avoids ths yo-yo

and of course, github link: github.com/gshelor/CFB_...

Anyway, Resume #VoA doesn't really seem to have any issues with this same weirdness, so that's nice, go #WIndiana and here are the Week 13 Projections

For the life of me I can't figure out why #VoA loves Missouri so much more than Bama, Tennessee, and Georgia. Please advise. It's not the only way the rankings don't make sense, but I probably wouldn't notice much other weirdness with the ranking order if this weirdness wasn't there.

This usually feels like less of an issue in conference play, but in non-con games, like LSU-WKU (I know they're not a MAC team but still) this will result in #VoA projections of ~5 point margins when the spread is somewhere in the 20s. Sure, VoA's been good ATS but that's scarily big disagreement

The #MAC is another example of an issue my model is having, where it just doesn't seem to properly rate how (sorry) bad the weaker teams in FBS are. Admittedly I mostly say this because I'll look at @espn.com's #FPI and Bill Connelly's SP+ as benchmarks for what is "reasonable" for rating values.

On the other hand, look at Wazzu as FBS's most ironic example of #VoA stability

The *rankings* seem mostly stable or at least reasonably responsive to performance but the instability of the ratings seems like a pretty huge problem and I have no idea how to fix it.

Also poor OKState. League of our own sadly.

The yo-yo effect is officially a #VoA thing and I hate it and have no idea how to get rid of it. It's one thing for teams to maybe have one new horrendously poor performance and the model adjusts but another for all the teams to mostly move up or down in unison.