UniCat

I have cracked the code on why the ESPN blind resume tool seems to hate the SEC. I played around with the blind resume tool a lot last night and got these two results, the 2nd being after I did 30 questions instead of 15. These were my only two runs, both blind. Things I learned on the methodology that explains outcomes: - FPI hates Kentucky (26th) and Mississippi State (41st). That doesn’t affect Kentucky itself (since FPI rating isn’t listed for the team you pick), but it does the teams that *played* them. - Alabama (W) LSU (L) South Carolina (L) and Oklahoma (L) all see Kentucky as “FPI 26-50” which really impacts their resume. - The committee treating Mississippi State as a top-15 team for Bama is also a huge cause of variation from the CFP ranking. - Moving two Alabama “top-15” committee wins to 26-50 FPI wins unsurprisingly makes then often the biggest mover. - The use of “Game Control” and ESPN “Strength of Record” biases in weird ways. Short games have made GC wonky, and it’s too easy to dominate with bad schedules. Using SOR is just inviting the blind ranking to match SOR (which mathematically favors undefeateds heavily). - The “offense” and “defense” rankings use ESPN’s “efficiency” rankings, which aren’t quite FPI and are generally odd. It helps some teams (Bama) and hurts others (Texas) seemingly at random. I say it’s odd because they’re components of FPI but don’t sum to FPI ratings. No one knows exactly what they are or what they measure but they’re likely EPA-based. - I’d rather just see FPI ratings than the three things they gave people as metrics, all of which are weird and hard to decipher. - The way the picker works is also a bit odd. Utah in my set lost every comparison but was still 6th. UCLA was 5th after beating Utah and losing every other comp. So there’s a weird pre-rankings bias at play. You can see how more comps shift things in the 2nd set, where I did 2x the matches and it let teams slide a lot more. Moving OSU up had punished Bama a lot, moving UH down weirdly helped UCLA. In short, you likely need to do 30-45 comparisons, not 15, to approach a meaningful result. - Finally, the number of people with ND, WF, and UCLA very high just confirms most people just say “losses bad” and don’t compare bands fairly. I didn’t value a team that didn’t have as many losses but also didn’t have any wins in upper bands (WF) over teams that had more losses but also some win in those bands (Texas, LSU). - The prior philosophical difference is why a lot of non-P2 stans see the results as so self-affirming. They rank teams with the fewest losses highly most of the time, so it favors teams with weaker schedules greatly.