I think it does a good job generally at both predicting and tying together joint probabilities.
I actually think it underestimates the power of Alabama as a brand. For example, if you pick Ohio State, Clemson, Oklahoma, LSU win out:
I think you'd see a lot of pundits clamoring for Bama there, especially if they lose to LSU without Tua.
OU may end up being us the last two years, especially if K-State falls apart down the stretch. That L will look worse and worse.
"I find your lack of kalua disturbing" - Darth Lebowski
The best thing for OU is that they made a comeback and didn't lose by 30. That might benefit them in the end, even if KState collapses down the stretch.
This is 100% right. Ohio State's bad losses were by 31 points in 2017 and by 29 points 2018. One loss by 7 points will not keep Oklahoma out, unless that spot is given to Oregon.
That 4% Alabama probability includes the chance they lose one of their final 3 games
I think all depends on how did Bama lose - close 3 point loss to LSU or more of a blowout
I don't buy one goddam drop of gas in the state of Michigan!
Nah; OU is a "best of both worlds" pick for the pundits: they'll appear to be unbiased while the talking heads refer to Hurts as the "amazing Alabama transfer".
At one point they gave an overview of how their formula works; if I recall correctly it doesn't take brand name into account whatsoever. It does weight probability of winning games that I believe it pulls from another calculator, and it weights the historic record of the selection committee (But based on their liklihood to choose a team based on winning it's conference championship, or having 1-loss, or etc... not WHO the team is).
Just poking around with that, seems the best % would be LSU winning out and Bama taking the loss in regards to us winning it all. (39% chance)
Not that I'd put any stock into this myself. Fun toy though.
Interesting tool, would have liked to know how it stacked up to reality the last couple years. Never the less, it seems to like OSU's chances.
It had us getting the #4 spot over OU and UGA after the CCG's. So, it was logical, it just wasn't accurate.
Well it gives probabilities, not certainties of anything happening. If it predicts something to happen with a 60% probability and it ends up actually happening 100% of the time, the model is wrong and is basically worthless. A prediction of 60% means it happens 60/100 and there's still many opportunities for it NOT to happen.
I'd say the numbers almost always make sense from that standpoint.
If you take everything I’ve accomplished in my life and condense it down to one day, it looks decent!
Yeah, they obviously do a lot of election forecasting, and if you read it enough, you know that even a 75% chance means 1 in 4 times the 'predicted' outcome doesn't happen. People misread this kind of thing all the time, because as a species, we suck at statistics. We see 75% and our lizard brain translates that into a near certainty. Then when it doesn't happen, we assume everything about the analysis is dumb and flawed, and ignore it ever-more.
The psychology of that kind of thing is fascinating from a lay perspective.
just an FYI 538 had Ohio State as the 4th best % to make the playoffs last year..and they came in a distant 6th
Right, because it used actual data and not feelings.
Exactly, which is precisely why I'd favor the 4 (or 8) team playoff being selected by computers, a la the BCS. The committee is just far too inconsistent and uses dubious logic.
We can't stop here; this is bat country...
If Clemson, OSU, Alabama and Oklahoma all win out, here are their odds of each team winning the Nat'l title:
I did not expect that, but I like those odds!
Born a Badger, Raised a Buckeye
I was surprised to see that Ohio State only has a 46% chance of winning out. Also interesting to see that if either Bama or LSU wins out, the other is left out.
It's a really tough three game stretch, including B1G championship game.
Success - it's what you do with what you've got. - Woody Hayes
5 game stretch, probably ;)
If Oklahoma makes the playoffs then the last 2 years the whole system is hypocritical! Oklahoma lost to an Unranked Kansas State Team, they not only lost they were blown out. Just like Ohio State was. Yea they made it close but they were down almost 20 points late! They are DONE!!! The Alabama LSU winner is the 4th team to make the playoff and should be as long as the winner wins the SEC Championship game.
I really wish that they would go back to the BCS formula and just take the top 4 from that. I know it would have been the same as what the committee has picked each year so far so lets just take the committee crap out of it and get rid the possible bias stuff.
The only issue I had with the BCS was that it only took the top 2 teams.
If OU, OSU, and Clemson win out, I still think OU gets in. At that point everyone else would have one loss and OU would have the conference champion "tie-breaker". Last year Oklahoma got in in a somewhat similar way. They were at #5 and UGA was #4 before UGA lost to Alabama in the SEC CG and OU jumped them to get the #4, presumably with the conference championship being the tie-breaker. The loser of the LSU-Alabama game won't even get to play in the SEC CG.
Unfortunately, I could also see that as a good thing for either LSU/Bama since they won't have to lose the day before the final playoff seedings come out.
The difference is Ohio State completely gave up against Purdue and Iowa, Oklahoma fought and was an onside kick away from possibly winning. Also, the committee ONLY cares about their rankings and Kansas State, unless they lose to Kansas, will be a Top 25 loss.
2x account suspension survivor
They could lose to Kansas, Texas or Iowa State or all of them. They are average by averaging about 18 points a game vs conference opponents but gashed Oklahoma by scoring 48. That’s Oklahoma’s fault. Bad loss!
56 - 27 = 8 consecutive
Not that these are not fun to see, but it is hard for them to be right. Because the committee is a bit of a black box. We know the criteria for comparing similar teams, but that is not always clear.
When you check the box for Clemson to lose, it tells me that the likelihood of that happening is <1% so it is almost inconceivable. ACC! ACC! ACC!
Same thing with the Oklahoma - Iowa St and Utah - West Virginia games.
Though the conference is weak, this has everything to do with them playing Wofford and nothing to do with the ACC.
I wonder if 538 predicted the Purdue game last year.
Doubt it, but if you play the probability and odds game, there is always a small chance of an upset happening, but you are better off going with the high probability team than the low in most cases.
I don't think anybody predicted the Purdue game last year. Even the worry people thought the game would be closer or that OSU could get upset if they were not careful. I am not aware of anybody that predicted Purdue would blow OSU out. Or Iowa for that matter. Maybe they would have tough games but not sure there was a soul that called a blowout.
No thanks - this is the same group that predicted HRC would win something like 38 states in 2016. This dude is not a football guy
He is not. He is a statistician. The whole point is this is what the numbers look like, when the human element is removed.
I mean, it said she had about a 70% chance of winning. So that’s not really out of the realm of possibility that the opponent would win.
The same group in 2018 predicted with an 80% chance that the Dems would gain between 21 to 59 seats with a mean of 234-201 and they won 235. They also did very well in the 2012 and 2008 presidential elections.
Additionally, Nate Silver is VERY MUCH a sports guy. His background is baseball not football but the dude knows his stats. He creates forecast models and then shares them. Like all forecasting, it’s not an exact science, but it’s still interesting.
Yeah, it's baffling to me how something that 538 predicted to have a one in three chance of happening ends up happening and people point to it and say they were so wrong. That's why it only gave her a 68% chance at the end.
We as a species suck at statistics.
* but still beat any other species.
**This space to be filled with an inspirational quote from a famous Buckeye**
“For instance, on the planet Earth, man had always assumed that he was more intelligent than dolphins because he had achieved so much—the wheel, New York, wars and so on—whilst all the dolphins had ever done was muck about in the water having a good time. But conversely, the dolphins had always believed that they were far more intelligent than man—for precisely the same reasons.”
The late, Great, Douglas Adams
Nah, only about 87.5% of us suck at statistics. The other 18% are great.
If I remember correctly, FiveThirtyEight gave Trump one of the best chances to win among statistical models....
Anyone who thinks 538 is not good because of that one instance (where something that they gave a 30% chance of happening actually happened) clearly doesn't understand how statistics or probabilities work.
When I walked in this morning and saw the flag was at half mast I thought, "Alright, another bureaucrat ate it." but then I saw it was Li'l Sebastian. Half mast is too high. Show some damn respect.
Yeah, a whole lot of people were mad at Nate in advance, calling him an idiot for saying there was such a high chance for Trump.
And his model was closest to success because he baked in the possibility of small polling error, and the reality that if there was modest error in one state, it was probably true in neighboring states too.
"I don't think you necessarily have to get a trophy to be a winner." - Nick Saban 1/2/15.
IIRC Nate's model was one of the lowest - some of the "start-up" polls were in the 90s.
538 had Clinton winning 24 states and D.C.... They gave her a 71.4% chance to win based on the electoral college. For a college football comparison; last year their final odds to make it into the playoffs were as follows:
Alabama - > 99%
Clemson - > 99%
Oklahoma - > 72%
Ohio State - > 66%
Notre Dame - > 61%
All others - < 1%
IIRC he mentioned that a big failure of the model is that it valued the conference championship game/13th data point as compared to an undefeated ND.
Oh man I wonder if people won't think its biased since it's favorable to Ohio State.
Proudly dispensing unbridled arrogance since 1983.
Sounds right. Although if OSU keeps winning, it is all that matters. The probabilities and such are nice to see but are estimates for such things. Particularly since the whole selection has aspects that are not data driven. I mean you can estimate the "eye test" but it can mean whatever somebody wants it to mean.
Given the numerous values and stats for every aspect of performance at this point it is a bit ironic that it boils down to a committee making the call. Would have figured the reverse than them taking over for computers. Ironic to hear pundits demand one and then gradually start to want the other or some modifications of the system.
An aside Nate Silver was born in East Lansing and re: 2016 if you read 538/Silver's articles, he/they had great reservations about the election not the least of which was the lack of polling in PA/MI/WI. Plus both candidates were/are:
Whenever you find yourself on the side of the majority, it is time to pause and reflect. ~ Mark Twain
Keep Purdue out of this.
Funny, when I saw that GIF, I immediately saw the reference to our performance in West Lafayette...
Stats are based on what did happen and not on what will happen. i will wait for the facts to come out,