"Computer Rankings"
I'm sure some of you have a homebrew stat-cruncher. I've got one that determines points, yards gained relative the average allowed of all opponents. Same formula for defense: points and yards allowed vs average of collective opponents' points/yards gained. Calculate by percentage, then multiply Yards per Play and Points Per Game, and out comes an interesting offensive and defensive rating rank.
The most eye-test vs. stat only issue with the ratings are probably the following:
Oklahoma #55
Clemson #41
Offense:
Defense:
Combined:
October 24th, 2021 at 10:00 PM ^
Impressive
Go Blue, beat Sparty!
October 24th, 2021 at 10:17 PM ^
talk about nostalgia
October 24th, 2021 at 10:13 PM ^
Saw the scare quotes and expected this to be a snowflake complaint thread or something, but this is actually quite cool. Nice work!
October 24th, 2021 at 10:17 PM ^
Cool post, and pretty interesting results! The formula to me seems pretty sound. You can rationalize how any particular offense or defense performs relative to their opponents average production against the rest of their schedule would tell you how they stack up against the rest of college football. Of course, there will always be outliers or misleading representation in the numbers, but by and large I think it gives a pretty good overall view of what the college football hierarchy looks like this year.
The Oklahoma rating I am assuming is being anchored by Spencer Rattler's production as the starting QB. Seemed to be like the offense was pretty clunky when he was the starter, and really opened up once Williams got going. They are definitely better than the 55th best offense in the nation. That Clemson ranking... actually feels about right. Their offense is just inexplicably bad this season.
As for Michigan, I am slightly skeptical that they actually have the 11th best defense in the country. I do think we're going to find out a lot about this secondary next Saturday. I'm really impressed with how quickly MacDonald got his scheme installed and also found ways to accentuate what his best players do best.. but we just haven't run into a team that can truly exploit the secondary with excellent QB/receiver play. I think MSU can do that, and then the obvious.
from a macro-level, I'm really impressed with the way these coaches and players have turned around the outlook for the program. We were looking dead in the water to end last year, and to have this team competing as a legitimate top-10 team is pretty remarkable.
I don't think this team quite has what is required to win the B1G and make a run at the CFP, but even the fact that is is vaguely plausible after the 2020 train wreck has me feeling pretty good. If they keep the swag/energy around the program up, we'll start landing the type of talent it takes to get there.
Cool stuff OP!
October 24th, 2021 at 10:22 PM ^
Thanks for the thoughts in the reply :-)
I think the main thing I've gleaned from the stats is that there are 3 teams on a top tier and 4-50 is depressingly similar- I exaggerate in jest but you get my drift.
Optimistically for Halloween weekend, the pace and opponent adjusted matchup predicts:
Mich: 30
MSU: 27
I'm counting on the road team dominance in the last 6 years!
October 24th, 2021 at 11:24 PM ^
Let's hope so! The score and game spread checks out, and feels like a sound prediction based on the variables.
I do think there are data points to suggest Michigan State *could* be a paper tiger this year, but they are absolutely at minimum a solid, competent team and they'll be motivated by the perceived disrespekt as always.
I think Michigan can take them, but they'll have to be ready for Sparty's best shot and also bring the juice on their sideline as well.
October 25th, 2021 at 1:40 AM ^
Yeah, these results mirror pretty closely with FPI and S&P+ for the most part. And that makes sense as S&P+ is largely a yards per play metric (with some adjustments for opponents, explosiveness, etc.). There is a top 3 and everyone else.
October 25th, 2021 at 6:38 AM ^
Looks like Georgia & everyone else. Also, Nebraska really is the best 3-5 team in the country!
October 25th, 2021 at 7:27 AM ^
Michigan will likely get 30, at least. But I don't know if State will get to 27.
October 24th, 2021 at 10:18 PM ^
I like that this random math problem says we're better than MSU. Hopefully that means we'll beat them on Saturday
October 25th, 2021 at 3:38 AM ^
If I understand these math number statisticals correct, the reason we score 120% of the points on offense and on defense, while Staee only score the other 61%. That's why.
October 24th, 2021 at 10:30 PM ^
I assume we're talking about Clemson being #41 on offense, correct? That might be about right, if I am understanding the basic drift of the formula correctly as you described it. They aren't playing offensive powerhouses, so most weeks, they haven't had to work THAT hard on defense, but their offense is a catastrophe, and I wonder if the only thing saving them from being lower is their non-conference schedule (I think they blew out Citadel and SC State, but that might be all).
October 24th, 2021 at 10:34 PM ^
I have Clemson’s offense at #103, that’s how they’ve scraped up to #41 with the #2 defense.
October 24th, 2021 at 11:19 PM ^
A 3-loss ACC team? If it were, say, Boston College, nobody would bat an eyelash. So maybe it’s not as weird as it looks.
October 25th, 2021 at 1:49 AM ^
Yeah, I feel like #41 is completely reasonable for them.
Oklahoma is a weird case and perhaps it's not surprising. They have really exemplified the concept of "playing to the level of their competition with one score Houdini escapes over such powerhouses as Tulane, Nebraska, WVU and Kansas. And none of those games were fluky. They were close with those teams in total yards each game.
October 25th, 2021 at 6:32 AM ^
I’ve sat through a couple Clemson games and that offense is as bad as it gets. So that ranking is more than right
October 25th, 2021 at 7:35 AM ^
Northwestern says ‘Hi!’
October 25th, 2021 at 10:20 AM ^
Clue me in, Cobra - I haven't watched any Clemson this year. Why (re: how) is their offense so bad? It can't be just the loss of Lawrence at QB, can it? That's a big cog, sure, but that offense had talent all over the field! Any idea what gives?
October 25th, 2021 at 11:37 AM ^
For reasons that are unclear to me based on my general awareness of their recent recruiting successes, their offensive line seemingly cannot block anything and they can't accomplish anything running or passing as a result.
October 24th, 2021 at 10:36 PM ^
Very interesting. Let's hope your home-brew computer rankings hold for this week. Beat MSU !!
October 25th, 2021 at 2:28 PM ^
FWIW, Vegas has M by 4.5. This might not sway me too much, but the Vegas line has been pretty accurate regarding M this year.
BTW- Pt Loma here. SD represent!
October 24th, 2021 at 10:38 PM ^
Colonel Mustard in the library with the candlestick....
Oklahoma would probably beat about 52 teams ahead of them. Their BIG12 defense is throwing it off, along with changing QBs.
But it does have Michigan's defense at #11, and so do the stats at the NCAA website.
October 24th, 2021 at 11:03 PM ^
Nice thesis defense. You passed.
October 24th, 2021 at 11:08 PM ^
Clemson’s offense is bad bad.
October 24th, 2021 at 11:18 PM ^
It looks like that lightening in a bottle that Dabo Swinney captured went back out of the bottle.
October 25th, 2021 at 5:58 AM ^
You're describing Cajun Hoke here, not Dabo.
October 25th, 2021 at 7:21 AM ^
Both are possible.
October 24th, 2021 at 11:22 PM ^
Was expecting something way different because of the scare quotes. This is cool
October 25th, 2021 at 12:06 AM ^
I can buy these rankings. It's pretty much Georgia, OSU, and Alabama in the elite tier, then everyone else fighting for the rest. Interestingly, two of the non-elite could very well make the playoffs if Georgia runs the table and hands Alabama it's (at least) 2nd loss.
I think one thing that your model lacks is the impact of special teams. These greatly impact the amount of yards and points that offenses get and defenses allow. Also, I think turnovers is incorporated in special teams. I think getting turnovers is largely random and luck, and losing turnovers is incorporated in a team's inability to punt (which I consider as yards gained by a team) or score.
That said, I did very much enjoy this post and your contribution.
October 25th, 2021 at 12:26 AM ^
Well, Sagarin says we win by 5. So yeah, the computers are in our corner.
October 25th, 2021 at 12:40 AM ^
I never thought or heard of a software program like this for a layman like me. Pretty cool. It fits the eye test well. #5 vs #24 makes me feel a lot better.
October 25th, 2021 at 1:02 AM ^
Cool program. One thing I've learned from looking at various advanced stats / predictive formulas is that ratings are the most illustrative, not rankings. This year there are 3 top tier teams while #5 through #40 are a relative logjam of good-ish but flawed teams. Your percentages reflect that as well.
Unfortunately, we are trained to think of college football hierarchy in terms of rankings and poll logic ("can't have OSU ahead of Oregon" e.g.) and with a somewhat arbitrary cutoff of 25. That's why sometimes these computer results cause cognitive dissonance
October 25th, 2021 at 1:07 AM ^
Very cool. This basically shows how relatively well, or poorly, Team A performed relative to their opponents' schedules. The only thing I might add is some kind of SOS variable. If Team A scored 100% and Team B scored 80%, then the thought would be that Team A is better than Team B (which, if I'm understanding correctly, is what you're showing above). But if Team A plays worse competition than Team B, then that conclusion may be flawed.
I'm not sure how to handle the SOS variable. Perhaps this would be more accurate with some control... intra-conference play? Maybe some multiplier could be applied to final intra-conference numbers to represent the conferences' relative strengths (SEC = 1, B1G = .95, PAC 12 =.01, etc. Just kidding PAC 12), and then at the end of the season that formula could produce a ranking?
Again, VERY COOL.
October 25th, 2021 at 6:51 AM ^
I like the 'green screen' look ... is your program written in COBOL? ;-)
While I agree the top three looks to be Georgia, Alabama, and Ohio State, I'm reserving judgment on Georgia until they demonstrate they can beat a legitimate good team. I give half-credit for their whomping on Arkansas (37-0), but I'm not sure their Auburn or Kentucky victories tell us much. I had hoped for Florida to be in a place where the Georgia/Florida game was going to be a blockbuster, but I'm not sure of Florida any more either.
The SEC championship will be a good game.
October 25th, 2021 at 8:41 AM ^
That absolutely looks like a mainframe screen. If so, and the program is in COBOL the chances are the OP is an old fart. We COBOL programmers are a dying breed.
October 25th, 2021 at 9:03 AM ^
They don't like to hear it, but I tell young tech guys that if they want to make instant big bucks: Learn COBOL.
That shit is still everywhere. Systems that we all thought would be replaced by now are still humming along with no end in sight. It stood the test of time, it's not going anywhere soon.
Yet there are fewer and fewer people left each year that know COBOL.
To paraphrase an old golf saying: Java for show, COBOL for dough.
October 25th, 2021 at 3:26 PM ^
I started out in 1979, back when it was Data Processing instead of IT. I progressed through my career in the corporate world and ended up in the MS stack.
But whenever I wanted to make real money I'd drop back into the mainframe world with COBOL, VSAM and DB2. There are millions of lines of existing code. The youngsters all say convert it, but it would cost companies billions of dollars to do it. And it's code that is stable and passed the test of time. All it needs is tweaked occasionally.
I'm retired now, but I get headhunters banging on my door everyday waving obscene amounts of money. Maybe COBOL ain't sexy but it can make you a shitload of cash.
October 25th, 2021 at 9:44 AM ^
It’s Google sheets “in the style of” old school mainframe text ;-)
October 25th, 2021 at 8:02 AM ^
Pretty sure we won't be taking MSU lightly.
Win the game.
October 25th, 2021 at 8:23 AM ^
Spartan stats are based on big plays against a weak schedule. (Kind of like the Rich Rod teams.) The game this Saturday should reveal the coaching acumen on both sides. If State performs on Saturday in the way they have this season up to now then I will be very impressed with Mel Tucker and his staff. Rivalry voodoo aside, I am looking forward to seeing what Mike Macdonald puts on the field.
October 25th, 2021 at 8:55 AM ^
I'm not a stats guy so it's entirely possible that I'm completely misinterpreting things, but what's the ultimate meaning of a metric that puts 5-2 Iowa St, 4-4 Tennessee, and 4-3 Florida in the top 10?
October 25th, 2021 at 9:07 AM ^
I cant speak for the SEC teams, but based on their description of the stat they are using I am not surprised by Iowa State. They severely outgained their opponents in both losses (even a decent margin on the yards per play it appears). The Iowa game was lost on a 4-0 turnover comparison which I don't believe is accounted for here and the Baylor game they gave up a 98 yard kickoff return and a 41 yard punt return which again I don't believe is accounted for here. So to the formula those games both likely look positive for Iowa State.
October 25th, 2021 at 9:07 AM ^
That not enough variables are being measured!
It's like the guy that drowned in the river that averaged only 2 feet deep. His stats didn't quite tell him what he thought they were telling him.
October 25th, 2021 at 9:39 AM ^
I’m not sure there is an ultimate meaning. It doesn’t factor in wins and losses, and only counts FBS vs FBS games from the 2021 season.
It’s just an objective metric to use in comparison with the eye test. Some weeks it goes 60+% ATS on all the games, some weeks it’s right around 50%, so nothing magic, just a “stat-test” to be used to corroborate the “eye-test”.
October 25th, 2021 at 9:47 AM ^
It means that some of their losses were flukes. Some teams have a bad day and an unusual amount of turnovers. Sometimes there are big plays that will not be allowed a second time. Iowa State might win all those games if they played them again. As we know, the best team doesn't always win. And then again, every team has weaknesses and sometimes there is a team/coach that can exploit them.
The stats tell us who has a good football team. Only the score tells us who won the game.
October 25th, 2021 at 9:29 AM ^
Skynet began to learn rapidly and eventually became self-aware at 2:14 a.m., EST, on August 29, 1997.
October 25th, 2021 at 9:55 AM ^
According to your program, the Michigan vs OSU game may be the closes since 2016. May be home field advantage will help us win this season.
I think if we can stay ahead of MSU at the half and our offensive running game can wear down their defense, we will win the game by 10+ points. My guess is UM 31, MSU 20.
October 25th, 2021 at 9:56 AM ^
I realize you aren't proposing this as a good ranking method, just one that produces interesting results, given its simplicity. Clearly, it does not directly consider SOS and win pct. It rewards teams that may have losses but seek to impress via massive yardage totals and blowout wins of weak opponents with tough schedules. According to a broad cross-section of 25 different computer polls, OSU isn’t #2 in the nation---It’s #2 in Ohio.
https://masseyratings.com/cf/compare.htm