Résumé S&P - A Different Way of Ranking Strength of Schedule
October 16th, 2018 at 4:56 AM ^
Stats? Throw a bunch of numbers in a barrel, and my head starts hurting. Its college, teams not only change yearly, but based on injury, they change weekly. The strength of schedule argument changes weekly based on how each team does after you play them, and reputation and name recognition always factor in. How do you determine how much weight to put on that? There is no full proof formula, but football still does a damn good job of figuring out who the most deserving is.
October 16th, 2018 at 7:31 AM ^
If you read the article, that’s sort of what Connelly is saying. True strength of schedule is hard to divine in such a small sample but we can learn something from each week no matter who is played.
October 16th, 2018 at 7:53 AM ^
That's the equivalent of saying recruiting rankings don't matter. Yes, stats mean something. These advanced states mean even more. Weird stuff can happen from week to week but these are just as predictive as anything else. It's not like the poll voters are known for impeccable analysis either.
October 16th, 2018 at 9:07 AM ^
I don't know nothin bout all your fancy numbers but I am interested in these advanced states you speak of.
October 16th, 2018 at 9:40 AM ^
"The strength of schedule argument changes weekly based on how each team does after you play them"
He updates these weekly, based on the cumulative performance of each opponent.
"and reputation and name recognition always factor in."
They shouldn't, and his formula doesn't factor them in.
"There is no full proof formula, but football still does a damn good job of figuring out who the most deserving is."
Fool-proof? And no, it doesn't. As top teams lose, others re-enter the picture. At the end of the season we are usually left with a number of self-appointed 'deserving' teams that is not exactly equal to four. That's why fans, players, media, and coaches tout the high points of their team's resume. (Remember '06 when CBS and Urban Meyer went full-on campaigning for the SEC to get in over Michigan?)
And inevitably, preseason "top" teams who ended up playing a terribly weak schedule (e.g. Clemson, ND) stay at the top by default.
These are our biases in selecting the top four teams. Pundits who go 8-14 against the spread every week telling me who is 'definitely the best team' based on 'the eyeball test'. Bill Connelly is determined to get rid of that stuff.
October 16th, 2018 at 9:42 AM ^
this is the worst take and the article linked does a damn fine job of propping up your straw man and knocking it down.
October 16th, 2018 at 10:06 AM ^
Might want to read the article. He does a really nice job of breaking down some of the BS IN the various rankings.
MSU is way down at number 57 in his adjusted rankings, btw. And he notes that--where SOS is concerned--none of the unbeaten teams have had rough roads at all.
There are reasons this week to be very bullish on Michigan. I have no doubt that Bama is tough, but--get past the MSU hump--and we are looking forward to a very exciting wind-up of the season.
October 16th, 2018 at 10:46 AM ^
Full proof?
Then I should be OK...
October 16th, 2018 at 12:58 PM ^
Yeah sure they do a good job if you believe Alabama deserves infinite shots at the playoff despite not making their conference championship game.
October 16th, 2018 at 5:09 AM ^
MSU at 50
#disrespkt
October 16th, 2018 at 5:55 AM ^
He lost me with PSU at 4. I can't reconcile that one.
October 16th, 2018 at 6:41 AM ^
That was part of the point i was making. They lost to MSU and OSU. OSU is undefeated, but they haven't really played anyone, and MSU lost to 2 bad teams, so where is the logic? Penn St. hasn't beaten a good team. So is this based on the theory that MSU was supposed to be good, and was ranked earlier?
October 16th, 2018 at 7:36 AM ^
Why even consider numbers if you’re just going to punt them and insert your own feelings anyway? Nobody said it’s perfect but it’s an interesting approach.
October 16th, 2018 at 7:57 AM ^
It doesn't really matter is my point. It's just a bunch of numbers based on a persons belief about how good a team should be. When the season is over (regular season) it will sort itself out.
October 16th, 2018 at 8:23 AM ^
It will sort itself out eventually? Well no shit...
October 16th, 2018 at 10:34 AM ^
Why should anyone talk at all about college football? It will all sort itself out anyway so this is all a waste of time.
What really is life?
October 16th, 2018 at 10:55 AM ^
There's also no reason to watch. You can just check the score a few days later.
October 16th, 2018 at 9:17 AM ^
S&P+ is based on how well you played in each game. PSU lost to MSU, but they outplayed them significantly. PSU apparently had 16 passes defensed, 1 of 4 of which would typically turn into turnovers, and put the ball on the ground twice. That means that MSU had the luck of roughly 5 turnovers, which equates to roughly 20 points. Based only on the box score, MSU had a 26% chance to win the game. That means that if MSU played @ PSU 100 times, you'd expect the record to be 74-26 in favor of PSU.
That's S&P+. It notably is not meant to tell you who has the best resume. Looking at resume is a historical look at a team while S&P+ is trying to tell you what a team is likely to do going forward. Hence, the new Resume metric.
The article explains the methodology and reasoning in detail. He even criticizes the people like you who say "Watch the game, nerd!". The metric measures what a team would do against the average schedule of the top 5 teams. If you want to learn something, go read the article. If you don't (which seems to be the case), then stay here and bitch about something that you admittedly know nothing about.
October 16th, 2018 at 9:26 AM ^
So by this logic, why is OSU number 2? They've played a bad PSU (as you noted) and a .500 TCU and nobody else. Alabama hasn't played anyone beyond an okay A&M.
I assume this is a bit or something, that you are trying to convey a general distrust for basic mathematics that permeates culture. Becauee otherwise rejecting stats for personal faith is fine but doesn't seem to remotely mesh with this site's ethos.
October 16th, 2018 at 9:43 AM ^
If you read the article, he addresses this. Those teams have been remarkably efficient, but have a lower ranked schedule strength. The efficiency keeps them up despite the lower strength of schedule
October 16th, 2018 at 11:14 AM ^
No, I've read the article. I'm arguing against the post that PSU can't be good because they've not beaten anyone, so then OSU and MSU can't be good because they haven't played anyone, blah blah blah.
The great thing about S&P+ is that it does a decent job at extracting value out of results that otherwise may feel useless, like how OSU took a bit of a tumble because they struggled a bit to beat Minnesota even though, on the ticker, it looks like a solid blowout.
October 16th, 2018 at 11:22 AM ^
Except S&P+ is based on how wel your team performed, not whether you won or lost. S&P+ is there to rank how good your team actually is.
Penn State May have lost to MSU, but they outplayed them... fairly significantly at that. Their S&P ranking isn’t hurt just by losing. Winning and losing is pretty irrelevant in their rankings, and rightfully so. I want a measurement of how good a team is, not a measurement of this team got lucky one week and they had some bad bounces that week.
October 16th, 2018 at 6:48 AM ^
That'll sort itself out a couple weeks from Saturday
October 16th, 2018 at 7:58 AM ^
I’m not sure of how this metric is calculated, but it turns out App State is actually very good. So I’m guessing they get a bump from that.
October 16th, 2018 at 8:57 AM ^
Good for App St. How can you tell? They play Southern Miss, South Ala, Georgia St, Charlotte, Gardner Webb, Ark St, Georgia Southern etc.......How would they be playing the Power 5 or even a MAC schedule? I'm not arguing with you about whether App St is good, I just don't know if you can assume it by them drubbing shitty teams.
October 16th, 2018 at 9:20 AM ^
I don't assume anything about App State. I look at the data that is in front of us for the past 7 weeks. And it says App State destroys weak teams and took Penn State, also a good team on a down to down basis, to the last play. You know what good teams usually always do? They destroy weak teams and play other good teams close. That's how I know App State is a good team. I think you're the one assuming things about teams based on your feelings and very small sample sizes.
October 16th, 2018 at 9:21 AM ^
I don't assume anything about App State. I look at the data that is in front of us for the past 7 weeks. And it says App State destroys weak teams and took Penn State, also a good team on a down to down basis, to the last play. You know what good teams usually always do? They destroy weak teams and play other good teams close. That's how I know App State is a good team. I think you're the one assuming things about teams based on your feelings and very small sample sizes.
October 16th, 2018 at 9:41 AM ^
I'll assume the same thing for App St that I'll assume for UCF. In a one-off game situation, ie: App St-PSU, UCF-Auburn, they are tough to beat and will hang close if nothing else. But on a weekly basis playing the P5, those two teams will show plenty of cracks. Injuries will mount up, and they won't have the depth to keep up. Teams will be better prepared for their personnel and schemes. They'll be taken more seriously as an opponent.
It wouldn't end well on a weekly basis. But one game per year, I like they're chances...
October 16th, 2018 at 9:44 AM ^
Yeah, this is what I was getting at, but he ignored my question about them play a P5 schedule or MAC schedule. It's not a bias or my personal feelings, I just don't think S&P is any better than the eye test.
October 16th, 2018 at 10:08 AM ^
ldevon1: "It's not a bias or my personal feelings"
Narrator: "It is"
October 16th, 2018 at 10:07 AM ^
Have you seen App State play a P5 schedule? I'm guessing not. So you don't know that they would show plenty of cracks. You just feel they would without any data to justify it. At least S&P+ takes actual data from things that actually happened and tries to figure out something about how good the teams are. Is it perfect? No because we can never know exactly what will happen. But it's a hell of a lot better than using only our preconceived assumptions about a situation to determine our opinions or course of action. I just don't get ignoring actual evidence when you have it.
October 16th, 2018 at 9:43 AM ^
If App State were in the MAC this year they would probably be undefeated in league play. Some of those teams are hot garbage this year.
October 16th, 2018 at 9:50 AM ^
"He lost me with PSU at 4. I can't reconcile that one."
This is a new formula, and could use some tweaking. Penn St might benefit too much from running up the score on Pitt, and maybe he doesn't penalize a team enough for losing. (A one-point loss to #2 OSU puts them right behind OSU.)
However, Penn St -- assuming they don't mentally collapse -- is still a Top 10 team at worst by advanced metrics, and you should bet on them accordingly.
October 16th, 2018 at 7:17 AM ^
this is a good 1st step, but PSU being that high and ND being that low (LOL) don't pass the smell test.
Probably too much weighting on MoV - that metric should probably get binned into 1 score wins (7 or less), 2 score wins (8-15), 3 score wins (16-24), and blow outs.*
* didn't get a chance to read in detail to see if he actually did this
October 16th, 2018 at 9:25 AM ^
I agree that Bill C's resume S&P isn't yet a great measure of accomplishment. ESPN's strength of record is a much better measure of the value of a team's record.
October 16th, 2018 at 10:08 AM ^
What does ESPN do?
October 16th, 2018 at 11:02 AM ^
It ignores MoV entirely. ESPN's strength of record assesses the quality of opponent using their own advanced analytics formula, but assesses the resume only using wins and losses. Their strength of record Top Ten is:
1. Ohio State
2. Notre Dame
3. LSU
8. Michigan
33. Penn State
October 16th, 2018 at 1:19 PM ^
See that screams of "pre-season" rankings still influencing how they view the difficulty of a schedule.
In strength of record, OSU is getting "difficulty" credit for wins against a week 2 #15 on a neutral site who is now unranked, and a week 5 #9 on the road who is now #18. (#33 in their own strength of record) Plus, ND is getting credit for a Stanford victory when they were #7, now unranked plus a victory over a #24 VTech, who had just lost to Old Dominon and is now unranked.
In S&P resume, Bama and OSU get credit for the impressive way they are winning in relation to the quality of teams they've played. That's why UofM is in the top4 since they "hung in" impressively in their road loss to a top team, and they have beat up the lessor teams. PSU's resume is high only because they have out played the teams they have lost to, I guess there's no metric in there for the "Franklin Factor".
October 16th, 2018 at 11:10 AM ^
They do strength of record.
October 16th, 2018 at 11:49 AM ^
Agreed. I was surprised MoV is capped at 50! Garbage time starts at 28 or so, I'd say, and the relative margin starts to matter less. Did Maryland or Wisconsin's late TDs really mean Michigan is a worse team?
October 16th, 2018 at 7:44 AM ^
I realize data analytics are all the buzz now, but when you get this result - especially PSU at #4 - it’s time to junk the results and figure out what’s wrong with the math.
October 16th, 2018 at 8:14 AM ^
Or, maybe, just maybe, he’s already taken his hundreds of thousands of data points, done the math lots of different ways, and found this is still the best way to run the numbers.
Or you can take the rational route and say the process is just wrong because you don’t like the answer.
October 16th, 2018 at 6:37 PM ^
Penn State has what I would chalk up as 2 1/2 losses (OSU, MSU and App St). All at home.
None of those three teams has another impressive win.
It is literally impossible that PSU could have accomplished enough in their 3 1/2 total wins to end up at #4.
It isn’t not liking the result.
It is telling you that if you have a small scale for weighing rabbits and a large one for weighing elephants and after doing each you tell me rabbits weigh more than elephants, I tell you your scales are fucked up.
October 16th, 2018 at 9:53 AM ^
"Junk the results"
Or maybe tweak them? And/or reevaluate your own assumptions about how good PSU is?
I think people who can't pick a game to save their lives are telling me about their airtight 'smell tests' and 'eyeball tests'. We see it with blathering pundits every week.
October 16th, 2018 at 7:48 AM ^
PSU at #4 basically invalidated his method
October 16th, 2018 at 7:52 AM ^
Or Penn State is, down to down, that good, but Franklin shits the bed when it comes to game theory and decision making
Most likely some combination of both; overvalued by the metric but we underestimate their ability b/c coaching
October 16th, 2018 at 8:11 AM ^
Hooray! Someone who understands S&P and math with a reasoned take!
October 16th, 2018 at 9:38 AM ^
Believe they were 90%+ win expectancy in 4th quarter if both losses as well.
October 16th, 2018 at 5:27 PM ^
PSU's defenders dropping like 5 picks probably had some to do with the lose to sparty as well.