Two weeks ago, Oklahoma State, Ole Miss, Penn State and UCLA were all undefeated. Tennessee fans had not yet torn down the literal uprights and taken them outside the stadium. And Notre Dame had not yet lost to Stanford (sorry).
Something that has not changed is ZeLO’s continued first-year success. In Week 7, ZeLO went 37-16 compared to FPI’s 36-17 outing. FPI did strike back in Week 8, picking 36-17 again. ZeLO dropped a few close contests, like Pitt at Louisville and Ole Miss at LSU, and went 33-20.
Though the loss was disappointing, ZeLO has been on a hot streak. From Week 4 to Week 7, ZeLO out picked FPI straight up and is tied from Week 4 to 8. Though FPI is 413-125 (76.8%) and ZeLO 396-144 (73.3%), I think the two models are closer than their records might suggest.
As the season has progressed deeper into conference play, with fewer Group of 5 vs. Power 5 matchups, ZeLO has become much more competitive. Early in the season, I wrote that the Power 5 vs. Group 5 problem was one of ZeLOs biggest flaws, and as we have seen fewer of those matchups, ZeLO has been able to close the gap.
However, that is a topic for next week. This week I want to debut a new feature I built into ZeLO and talk about the results I got back.
One of the more silent problems ZeLO has had this season is how highly it touted some of the more substantial Group of 5 teams. This bled through in the model’s propensity to treat Power 5 teams and Group 5 teams as equals in game picks, but it also led to teams like JMU being highly ranked in ZeLOs Power Rankings.
And even when JMU was 5-0, they were, at best, a new young Group of 5 team, certainly not someone who was a peer to teams with legitimate championship aspirations.
To fix this problem, I looked at teams’ strength of schedule (SOS) and current records to reward teams with demanding schedules.
However, I also wanted the new metric to capture current projections to reward teams expected to do well in the future against a strict schedule. Think Georgia, who has one of the toughest remaining SOS but is still expected to win the SEC. Georgia had a low-quality early schedule, but their schedule is good overall, so they should be rewarded for that.
This system served to help lift Power 5 teams, and the initial rankings showed that. To further boost P5 teams, I decided to factor in winning conference divisions and championships, weighting the Power 5 Divisions and Titles as double that of the G5.
I also added a ranking component to help with ZeLO’s win-loss blindspot. The average voter will rank Tennessee higher than Alabama because of a head-to-head result, but ZeLO still has the Tide as the better team, statistically. So, it made sense to factor in outside evaluations of teams to help nudge ZeLO in the right direction.
The top three are likely unsurprising to any college football fan: Ohio State, Georgia and Clemson. All three are undefeated and projected to finish undefeated by ZeLO. As a result, all three should make their conference championship games; and win it. Hard to argue with that.
The Defeated Champs
Fourth is the surprise team of the year, TCU … a surprise to everyone except ZeLO. TCU was one of the teams that ZeLO loved that I could not understand. It worked out pretty well for ZeLO. So why is TCU not in the category above with the powerhouses? Well, ZeLO has them losing at least one game down the stretch. That hurts their case to make the CFP, so they take a bit of a bump.
Also in this category is Oregon, though the Ducks were shellacked at the start of the season when they lost to Georgia, so their margin for error is zero. ZeLO has them running the table, with their main roadblock being Utah. Assuming Georgia wins out, the atrocious loss should at least be excusable.
Outside Looking In
Leading the category are Tennessee and Michigan, who are both projected to lose to their eventual conference champion. The two teams are spared losing in their conference championship and end the season with excellent what-if potential as a one-loss squad.
Next is Alabama, who is hurt by the fact that while they would make their conference championship game, the trip should be a loss, making Alabama a two-loss team. The Tide are still a strong squad, so they are 8th, needing either an upset of Georgia or the CFP Committee to break its tradition of only letting in 1-loss teams.
Following Bama are Oklahoma State and Penn State, two teams who will finish with losses (three for the Cowboys and two for the Nittany Lions) but with solid showings overall. PSU will finish with a loss to Michigan and Oklahoma State. will finish as conference runners-up.
Finally, USC. The Trojans are a good team that should miss their conference championship, taking a loss at some point to either UCLA or Notre Dame. They will be a solid two-loss team without the credit of being Alabama. I completely buy that the Trojans are better than where ZeLO is projecting them, and they could run the table and compete for a conference title. In that case, they would be the underdog in that matchup, making them a two-loss team, so either way, a solid two-loss team.
Week 9 Picks
Notre Dame @ Syracuse
Notre Dame (who ranks 57th in ZeLO’s CFP Projections) is both a Vegas underdog and a ZeLO underdog. Syracuse is a good team, even if they lost to Clemson, and they have home field. ZeLO has Syracuse as one of the top-15 defensive teams in the country, something Notre Dame’s offense-ranked 71st will struggle with. On the flip side, Notre Dame’s defense should be able to limit Syracuse’s offense. I think this one will be a low scoring affair that ND can win, but ZeLO gives them just a 31% chance to do so.
No. 2 Ohio State @ No. 13 Penn State
This game might not seem worth paying attention to (see the -15.5 Ohio State spread). When Michigan beat Penn State 41-17, that felt like the end of the season for the Nittany Lions. But the Lions are hosting this time, and home field advantage makes this one interesting. ZeLO has the Buckeyes with a 73.6% chance of winning, so a far cry from a close match. But one worth watching nonetheless.
Contact Tom Zwiller at email@example.com.