S&P Rankings: Michigan #10
Caveats about this still being early in the season, which can make for some wacky results, but every week renders the S&P ranking system more accurate, and of course I prefer the ranking system that rates us highest. :P
Here are the Big 10 teams in the top 25:
OSU: #7 overall (#33 offense, #6 defense)
Michigan: #10 overall (#48 offense, #4 defense)
MSU: #15 overall (#19 offense, #26 defense)
Wisconsin: #16 (#42 offense, #14 defense)
Minnesota: #20 (#67 offense, #11 defense)
Penn State: #22 (#68 offense, #13 defense)
Two top 10 and five top 15 defenses--not bad!
September 27th, 2015 at 12:35 PM ^
Wackiest result is Utah #29
September 27th, 2015 at 12:39 PM ^
...but at least it's not some idiot "makin' calls" based on whatever. This is what the data outputs based on the methodology. Unfortunately it isn't too accurate this early in the season. It will be.
...but I still support any ranking system that says we are good. For moral reasons. :)
September 27th, 2015 at 1:51 PM ^
September 27th, 2015 at 4:35 PM ^
don't realize that complex computations are superior at processing a lot of data than their error-prone and biased brains.
The calculations for these things aren't actually that complex - what they are are data-intensive.
I'd put my money on a highly flawed light-on-data algorithim over most people, who only have the capability of watching a handful of games of detail and then maybe checking a bunch of final scores. To wit; the majority of this board couldn't believe that UM was favored by 4 or 5 points against a BYU team that beat Boise, Nebraska and almost beat UCLA. The computers (S&P) meanwhile, said -7.
Yeah sure, you can question the methodology of the models and the (maybe) arbitrary value-judments implicit in them, but humans are going to make even more arbitrary value-judgments in qualitative assessments. More importantly, they can't come close to processing all the available information.
I'd wager the S&P is more accurate than the human polls if you compare week 4 to the final season outcome.
September 27th, 2015 at 7:48 PM ^
I predicted Michigan would beat BYU 24-10, so if S&P is better than "the majority of this board" for getting 2-3 points closer to the final margin of 31, that means, by using a quantitative assessment that looks at the available data as you recommend, my qualitiative, value-based judgment ability was way, way better than S&P's model.
September 27th, 2015 at 8:38 PM ^
The S&P rankings for offense and defense look like a decent estimate, but (correct me if I'm wrong) it looks like they got the overall team rankings by subtracting the defense scores from the offense scores (whatever those numbers are supposed to be) and sorting the results in descending order. That seems really, really unreliable.
September 27th, 2015 at 5:40 PM ^
Great point. Statistical/computational rankings aren't necessarily any less biased than the human rankings, they're just more transparent (usually). They still require the programmer/designer to make weighting decisions that carry some bias toward certain metrics. I think these rankings are generally helpful if you know and understand the system, but they're not any more gospel than a bunch of writers or coaches making ordinal decisions.
September 27th, 2015 at 5:58 PM ^
Because they consider every game equally. Humans can't do that.
You are right that the subjective element isn't removed entirely, because humans decide the methodology, but they at least apply it consistently.
You don't have to understand the computations to appreciate that they are applied to all teams fairly. You don't have to know the individual human to know that they can't do the same thing. Nobody can watch every snap of every game, and each person takes in different media to summarize the rest.
Neither approach (human or computer) is perfect. But one is woefully incapable of taking in all the information and heavily influenced by preconceptions and media coverage.
September 27th, 2015 at 8:28 PM ^
I actually don't have to appreciate that the computations are applied to all teams fairly. As blunt and simple examples, if the algorithm gives too much weight to passing offense or home field advantage, then it treats teams that rely on the run or play less road games unfairly.
I'm honestly just tired of the data worshipping in the national mainstream media. It's pretty clear the excess amount of coverage and consideration is sub-advertising being purchased by the stat companies. It's also gradually training people to accept the output of these algorithms as truisms without questioning the inputs or how they're being processed. I guess one thing we can agree on is that media coverage is heavily influencing.
But quite frankly, the biggest thing that annoys me is that most of these sports-stat writers don't really understand the sports they're writing about that well and over-rely on the number processing without recognizing their overreliance. They also toot their horns whenever data-based predictions correspond to outcomes, but ignore, gloss over, and explain away all the circumstances when they don't.
September 27th, 2015 at 11:16 PM ^
The popularization of analytics makes more people interested in them and learning about them. Instead of going "psssh whatever nerd" people are recognizing the value of these tools.
You're right that the BYU game doesn't prove anything. But every book in Las Vegas using models should tell you something.
Any person who applies models knows that the models have limitations. That's where the "explaining away" comes into play. Know body who makes these things thinks they are the end-all-be-all. But what they are is superior to human judgements sans data.
September 27th, 2015 at 8:02 PM ^
In addition to not being verifiable in terms of whether the available data is being considered properly, data's limited in what it's able capture, and algorithms are limited in what they're able to consider. There are many things that aren't easily representable in numbers and classifications, and there are many things that aren't properly represented by numbers and classifications.
As another example, when BYU was punting on their second possession with UM up 7-0, I knew this was going to be a blowout (this was after BYU had "driven" into Michigan territory twice and before their 3-and-out streak). The computer models had no clue a blowout was on the horizon, but it was pretty clear to me using my special human intuition power.
September 28th, 2015 at 12:09 AM ^
All that's true (well the first paragraph anyway). Yet - The list of human limitations is far longer. Imperfectly considering all the information is better than VERY imperfectly considering a small fraction of the information.
Excellence is attainable, perfection is not.
September 27th, 2015 at 12:52 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 27th, 2015 at 1:35 PM ^
have great numbers against Fresno State. Only reasong they aren't higher.
September 27th, 2015 at 3:29 PM ^
for Utah and it was the best game that USU has played this year.
September 27th, 2015 at 4:26 PM ^
but if you discount the Oregon victory (and Oregon has done nothing this year to earn anyone's respect) then all Utah has to it's resume is a defeat over Michigan (at home, by 1 score) and a couple middling results against Utah State and Fresno State.
The S&P rank isn't going to give Utah credit for the fact that Wilson was out for those 2 games.
September 27th, 2015 at 1:45 PM ^
RIght up there with:
September 27th, 2015 at 8:08 PM ^
I was really annoyed by all the comparisons with Beckham Jr. in the media without there being any mention of Marquise Walker against Iowa.
There's also nothing that compare's with Woodson's snag in Sparti Stadium.
September 27th, 2015 at 12:45 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 27th, 2015 at 12:46 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 27th, 2015 at 2:29 PM ^
It's science, bitches.
September 27th, 2015 at 12:42 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 27th, 2015 at 12:50 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 27th, 2015 at 12:50 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 27th, 2015 at 12:51 PM ^
It's good that there is a playoff system now, because I think every team in the country is beatable. I wonder if we will see a 2 loss National Champion this year?
September 27th, 2015 at 12:52 PM ^
September 27th, 2015 at 12:55 PM ^
September 27th, 2015 at 12:59 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 27th, 2015 at 12:55 PM ^
September 27th, 2015 at 1:00 PM ^
September 27th, 2015 at 1:25 PM ^
That's a great gif.
September 27th, 2015 at 1:35 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 27th, 2015 at 1:42 PM ^
I've waited for YEARS for Michigan to field a back who could teleport.
September 27th, 2015 at 1:20 PM ^
September 27th, 2015 at 1:28 PM ^
Sent from MGoBlog HD for iPhone & iPad
September 27th, 2015 at 1:37 PM ^
Sagarin thats almost all a result of the Michigan/Utah/Oregon/Michigan State comparison scores.
September 27th, 2015 at 1:38 PM ^
September 27th, 2015 at 1:59 PM ^
I expect them to get better as the season progresses. But this is not a the same defense of the last 3 years.
September 27th, 2015 at 2:56 PM ^
MSU just does not have the same kind of corners they have had in recent years.
This means that they either (1) have to play zone coverage more and commit fewer players to blitzing and completely shutting down the run (the formula that they used against Michigan for the last several years), or (2) play man coverage, blitz and stack the line, but then pay for it by giving up chunk pass yardage.
At this point, the tables are turned. All of a sudden, MSU hasa suspect set of DBs and M's corners are playing great.
Therefore, M can play more man coverage and commit an extra guy to going after the QB. That formula was lethal against BYU yesterday. Though I expect Conner Cook to handle it better than Mangum, this at least makes M v MSU a toss up game this year.
M still has three more weeks to get better every day.
September 27th, 2015 at 9:23 PM ^
September 27th, 2015 at 2:15 PM ^
It has not been updated yet, but before today's game, it looks like we were #36 in overall FEI with an estimated mean wins at 7.3 and estimated remaining wins at 5.1. It will be interesting to see where those same numbers sit after, say, tomorrow when I believe it will update.
Actually, come to think of it - I think the probability matrix for the remaining games might be an interesting exercise at this point. Off to Massey...
September 27th, 2015 at 2:39 PM ^
Yeah, you have to imagine that those numbers are going to tick up.
September 27th, 2015 at 2:31 PM ^
It overrates sparty.
September 27th, 2015 at 4:59 PM ^
the computer polls are better at predicting future outcomes than the human polls, even this early in the year.
They're far from perfect, especially this early in the year (i.e., this light on data), but the numbers have been more in line with reality than human polls since FEI and S&P started being published.
For now, it's interesting to compare and contrast. By week 7, the human polls should be ignored entirely in favor of these computer polls.