Is there a "KenPom" for NCAA Football?

Or even a football version of the Blue Ribbon Yearbook?

blue ribbon yearbook equivalent maybe Phil Steele's mag ..  its a preview but will obvi talk alot about LY and give some hints .. best preview since he talks to almost all the coaches directly, others seem more of a popularity contest .. 

there's not really a kenpom but check out bcftoys.com fei ratings (free) and bill connelly does the sp+ ratings (maybe a paywall now) .. if just looking for the power ratings / spreads then find the usa today sagrin ratings .. that said I suspect the models will be really confused again this year, early on for sure .. very few non con games LY.. some teams only had 4 or so games.... lot of teams got hit by covid every which way  .. tons of opt outs .. tons of xfers in the portal for this year / seniors getting extra year .. the football models aren't gunna be nearly as precise as b-ball or baseball anyway and prob much less reliable than usual again this year .. 

good luck!

@Bridge1

 I appreciate the input!

Phil Steele is good to start the season.   I also like Sagararin.com.  Its updated constantly by Jeff Sagarin who started the power rating publishing even before the advent of in the internet in USA Today.   It is not has detailed as KenPom, but its power ratings are good and normally reflect very close to the actual lines.   

Yeah true but recommend looking at sagrin ratings in context of where a team is in their conference  .. the limited interconf play really screwed around with the overall rankings last year .. like E.mich, W.mich, C.mich finish the year in the top 40 ... ahead of LSU and at least 20 better teams ..  FCS spring was like that and because bookies really didn't know better it was kinda easy to look over sagrin and see who was likely to be over/underrated before the lines came out .. FBS prob won't be that kinda season long boondoggle but would say if we see the bookie lines mimicking sagrin early on then good guess it'll be because bookies aren't too sure themselves where the lines go .. 

@Bridge1

The problem with the KenPom approach for football is there is not nearly as much data as hoops.  

12 games vs 30 for hoops

There are also fewer plays in a football game and way more different situations.  3rd and short vs 3rd and long.  In basketball, you are either trying to score, or burn clock and score.

And football also has weather.  A couple games in really bad weather makes the offense look a lot worse and more run-heavy.

And fluke plays in football can have a huge impact on the stats.  There are fluke plays in basketball too, but probably not as many and they are only worth 2 points out of 70.  In football, you could have a pass hit a defender in both hands, carom to the WR who scores an 80 yard TD.  The defender should have had a pick six.  Or the center makes a bad snap, the QB covers the football, and it counts in the stats as a 30 yard rushing loss.

Dr Bob, years ago, was a tout who used to move the lines several points when his picks came out. He used to go through the box scores and adjust the stats to minimize fluke plays. He had some good success for a couple years then the books caught up to him.  I don't know if he is still around. There are probably other groups that do the same, but if you put that much work into ityou won't give it out for free. Ken Pom is all automated, so once it is set up it is no work. He just loads stats and it spits out his rankings.

yeah agree with thorpe ... even if we could customize our model and adjust and carve out things like freak plays, weather, injuries, etc etc there's still tons of variables in the nature of the game w 22 players on the field at one time.. gotta handicap 4 teams: 2 offenses / 2 defenses and special teams sometimes a really underrated element ... by the time we get confident that a team is reliable and worth gambling on we should assume that bookies models figured these things out .. 

Also have to understand the models (like preview mags) are motivated by being the most accurate model over the whole season for all 130 teams .. thus you have alot of abstract predictive elements for all teams that may not be relevant for the team you are betting on .... for ex Bill C explains DL's only account for 6% of the defensive ret pro rating since returning DL's stats have a fairly low predictive value (over all 130 teams) ... for instance only see 1 Wisco DL in the last 7 years who crack the BIG-10's top 50 TFLs/game .. Wisco switched to a 3-4 7 years ago and ever since will use these BIG DL's to take up double teams and fill gaps letting the LB's get all the glory on a big TFL/sack stat play .. I suppose a sharper model could recognize some team's LB production is a function of the DL's but that gets even more abstract and complicated .. end of the day model still accurate (overall) by just diluting the weight of the DL ret production .. 

is kenpom.com really that great or unique? (don't get me wrong, i've been a subscriber at times in the past. and i like it)

is the greatness of kenpom.com that 1) it's done so much work for you, 2) very easy to use; 3) reaches a conclusion (predicts the actual scores.... i assume KP might actually model the game itself)???

massive transfers is new and i agree that they are underanalyzed...

and the analysis i've seen of transfers missed the point of them.......

a transfer doing a horizontal move or a moderate downward move doesn't matter that much.

a transfer doing a significant downward move is significant....... there's group of 5 schools getting former alabama and georgia recruits. they would never get a sniff of that player out of HS. same with lower P5 schools (oregon state, illinois, kansas)... a team UTEP with 6 or 7 power 5 transfers is a BIG BIG DEAL

to the one poster, can you give us more detail on Dr. Bob?

i remember him. written up on cover page of WSJ............ but i don't remember details on what he actually did to predict games. you say, "adjust for funny scores". anything else come to mind?... "adjusting for funny scores" makes total sense to me. and doesn't happen that often in NFL.

what was the service that moved G5 football games a huge amount about 10-15 years ago? people waited breathlessly for their picks.

i like analyzing G5. teams can change alot. look at MAC. ball, kent and buffalo were the best teams last year. perennial doormats or close to it... northern illinois was the worst. was the best not that long ago. same for CM.

but for the P5, it seems "same old" to a large degree............ in the big 10, indiana and minnesota make mini-runs from lower-mid pack. but that's about it.... FSU falls on hard times. .... but not much more than that. seems like the general within-conference rankings are similar every year.

There were some really good responses here and I would add a few things to it.

TBH I don't find Sagarin's (not sagararin) #s helpful at all, at least not for fantasy play (which for me is fantasy picks on spreads and SU play).

If you were to take Sagarin's ratings and say that a team with a higher rating would beat a team with a lower rating, you will quickly find yourself losing. For example, I see ohio st. above michigan. Well, Michigan won this year and won decisively. Another example is Baylor. He has them at #13, with Oklahoma and Oklahoma Lite ahead of them and 5-6 spots at that. Well, Baylor beat both of those teams this year (lost to ok lite in regular season and beat them in their big 12 champ game). Does that make sense to rate teams above teams they lost to? No, it does not. If you read his instructions it says to add 3 points to a team for home field advantage. Even that still fails because if you applied that to michigan vs. ohio state, michigan barely moves ahead of ohio state on ranking. 94.92 vs 94.18. So the ass whooping Ohio state got is only worth a .74 difference? Again, you can see its just not a system that is really worth anything.

I disagree with other people in here, I think there is enough data, even with fewer games compared to basketball. The problem is the time it takes to mine the data that is meaningful. I have yet to find a site that puts all the data I want into 1 place; I have to get it from several locations. I use teamrankings.com and footballoutsiders.com, as well as injury (covid mostly) lists.

Another problem I run into is most sites don't separate OOC game data from conference game data. I wish they did because I don't care if Georgia beats on the virginia school of the blind and texas southern when I'm looking at its first SEC matchup in a given season. The prior stats against those inferior programs is irrelevant. But for some reason every site dumps all of the data together.

Bowl games are much harder to predict. There are so many moving parts (opt outs, covid problems, tons of coaching and staff changes) that a lot of stats can just get tossed. Plus, some teams have weeks off from their last season game to their bowl game. That has to negatively impact their play, especially precision and timing offenses. Plus, motivation - what motivation do these kids have to play hard and win? The games are meaningless and despite what you hear in the media, a lot of the kids, while enjoying the games, they do find the games meaningless - because they are. They are just exhibition games created to make people money. Oh wow, we won! We are the cheeze-it bowl champs! yay! (sarcasm). For bowl games I find myself going with A). the # of games I saw a team play B) my gut. So far this season I only played 3 bowl games. Yep, only 3. I felt Oregon was going to lose big to Oklahoma, I felt N. Carolina would lose to S. Carolina and I felt AZ St would lose to Wisconsin but I wasn't sure by how much, since Wisconsin is so slow and plodding.

Someone mentioned Phil Steele. I have been buying his preview for 20+ years, every year, as a tradition. He also has a subscription service which is good for some weekly updated numbers and injuries, etc but as far as a Ken Pom approach, capping solely, or even predominantly off numbers, as some of the Bros mentioned, there just isn't those types of data points in football. Baseball, on the other hand...

I would add , I grew up watching the P10 (now 12) but I find myself watching a lot of B12 football. The reason is they are the only conference that plays a true round robin. There are 10 teams and they all play each other. That is 9 conference games. This is great for data purposes - you really get the data you need.

Most CFB ranking sites fail in that they try to rank teams from different conferences. Statistically you cannot do that and say the data is legitimate. This is where the "not enough games played" argument gets legs. There are so few data points in intra conference matchups that's its silly to try and rank the P5's.  Be reasonable when you read my statements. I'm not saying the SWAC is better than the SEC. Now, you can say the P12 is better than the MAC and I wouldn't argue with you, even though there are almost no data points (actual games played) to support it.

Another site that is somewhat useful for quick data to look at is oddsshark. I'm not big into spread trends but they have decent data for that. One thing you can do is pull a matchup and adjust the data under edge finder, which is ok I guess but it uses OOC data, which again I'm not a fan of. Mostly I use oddsshark for the easily found injury lists and "last 10" records.

Another problem I run into is most sites don't separate OOC game data from conference game data.

@wonglar

 I run my own numbers omitting factors that I feel skew the numbers, including games against The School for the Blind, second halves of blowouts where a team gets into the third string, etc.