BCS Analysis – The Undefeateds

I used to root like hell for “BCS disaster”. Then the #1 team in the country was left out of the NC game, and the BCS survived. Then the undefeated SEC champ was left out of the NC game, and the BCS survived. 1 and 2 loss teams win the NC while undeated teams sit and watch.
Now I don’t really give a shit. It will survive anything, the fucking cock-a-roash.

The true national championship game, to be played in just over two weeks time, will be won by Alabama by a score of 13-10. Bank on it.

College football is completely broken

Never give up the fight, Nordberg!

Please?

I’m with nordberg. Rooting for BCS chaos used to make it fun to watch a lot of games that didn’t really appeal to a UT guy. Now? Nothing short of Congressional action or superconferences seems destined to give us the playoff this sport needs. It IS disheartening.

Very interesting stuff. But computer-generated, mechanical. The games are played by 20-year-olds with an eco mid-term two days away, and instant replays that are called wrong, and odd-bouncing punts.

I think the probability scores are way out of line, in most cases (probably not Boise State’s). . . Stanford 100% over Cal and 91% over Notre Dame? . . . LSU 97% over Auburn? . . . Alabama 99% over Mississippi State, in Starkville? I agree it looks a lot like 2004 — at this snapshot in time. It’ll look different when Clemson and Stanford get beat (they will), which means Wisconsin gets the Auburn ’04 role. Except I think the Badgers spit the bit in mid-November.

Stats are for losers. Every game should be 50%- you either win or you lose. I hope the Hilltoppers blow your ratings out of the water.

Oh, and nice breakdown of all the possibilities. If the above mentioned upset of the century doesn’t happen, here’s hoping for 7 undefeated teams.

Cool stuff. Cal at Stanford…100% win prob. Interesting.

i think the Voodoocats’ chances against OU are a little low. 10%? It’s away from Norman and Bill Snyder is already in Landry’s head.

what the hell am i talking about. the Gooners’ll win big.

bevosbackside,

“Every game should be 50%- you either win or you lose.”

I’m fairly certain that was tongue in cheek (?). But the sad thing is, that’s how way too many people think probabilities work. If there are two possible outcomes, the odds are 50/50. Nevermind the outside forces that influence the probabilities of either outcome.

Great stuff.

I always want BCS chaos. I’m hoping that a 1-loss team from the SEC or BIGXII jumps an unbeaten Wisconsin or Stanford.

Huck if LSU loses a one-point game to Bama and murders everyone else, what are the odds that they stay ahead of tree and badger?

What we have never seen, as I recall, is an undefeated B1G team left out. I wonder what Lord Delaney will have to say about that because it seems highly unlikely that Wiscy will lose and likely that either Bama or LSU will be playing OU.

That would be fun to see. The only thing sweeter would be seeing them square off in the rose Bowl.

So Wyoming has a better chance of beating Boise State in Boise than TCU does? That surprises.

Yeah, I mentioned the game probabilities seemed high to me, so I ran a check against published odds for this week’s games. Taking the juice out of the moneylines leaves us with:

Auburn at LSU – I have 96.7%, odds (1100/-1700) say 91.9%
Tennessee at Alabama – I have 96.1%, odds (3000/-7000) say 96.8%
Texas Tech at Oklahoma – I have 98.2%, odds (3000/-7000) say 96.8%
Oklahoma State at Missouri – I have 79.6%, odds (230/-270) say 70.7%
Kansas St. at Kansas – I have 94.0%, odds (350/-440) say 78.6%
Air Force at Boise State – I have 100% (with rounding of course), odds (5500/-13500) say 98.2%
Wisconsin at Michigan State – I have 82.5%, odds (245/-290) say 72.0%
North Carolina at Clemson – I have 97.0%, odds (315/-380) say 76.7%
Washington at Stanford – I have 96.8%, odds (850/-1300) say 89.8%
Marshall at Houston – I have 88.0%, odds (900/-1375) say 90.3%

Little differences stack up quickly in this kind of analysis. Clearly the computer has a lot more faith in Clemson, Oklahoma State, and Kansas State this week. It’s no coincidence that those three teams have the lowest standard deviations of the teams listed here. Air Force and Stanford also have low standard deviations, which contribute to the discrepancies in those games. Typically speaking, Vegas moneyline numbers are set based on the point spread with very little consideration for relative consistency of the involved teams.

That’s not to say that the computer is more correct than the lines, but an explanation of why the differences exist. I actually agree with those saying that the percentages seem too high. The fact that “the game is played by 20-year-olds” has nothing to do with that, though because those same 20-year-olds’ performances are how the standard deviations to date were calculated.

Phenom -

That is because of the same standard deviation issue. Boise State has a power rating of 57.01 with a stdev of 7.38 so far this year. TCU is 34.97/4.95 while Wyoming is 10.46/18.59.

Basically because all I do is use the sum of two normally distributed variables to arrive at the probabilities, Wyoming has both a slightly better chance of beating Boise and a much better chance of losing by more than 35 points. TCU has a 0.3% chance of winning and an 11.4% chance of losing by more than 35 points. Wyoming has a 0.7% chance of winning and a 75.4% chance of losing by more than 35 points.

nordberg -

That would be up to the humans so it’s hard to predict. But if it happened that would be a little more momentum toward BCS disaster. I see you hoping.

Adding Percentages ! ?

WRONG.

Go look at your math on the ’Undefeated SEC …"

These are discrete sequences and can not be added.

Wisconsin’s projected final SOS is virtually the same as Boise’s. Ha!

I love how the B12-2 enjoys all 10 spots of the top 10 projected final SOS.

“That would be up to the humans so it’s hard to predict. But if it happened that would be a little more momentum toward BCS disaster. I see you hoping.”

I’ve been hurt before.

mitch -

The probabilities of mutually exclusive events are added to determine the probability that at least one of the events occurs. As an undefeated Alabama and an undefeated LSU are mutually exclusive events, those probabilities are added.

If that is incorrect, I’m happy to hear your explanation and revised numbers. Learning is good.

Huck — my only beef with the analysis is that you have South Carolina in the SEC title game.

Georgia is playing better, and South Carolina is rapidly degenerating. Georgia is more likely to play in the Dome.

Your stats have LSU winning that matchup 92.9% of the time… and Bama winning it a staggering 99.0%.

(And I don’t buy that, necessarily, but it does affect your numbers.)

Yeah, I just plugged in South Carolina because they beat Georgia and are in the lead right now. With Lattimore’s injury that will probably change. Main point of this post was to get the analysis out there and it should be updated on a weekly basis. I’ll look more closely at those assumed CCG opponents next week.

Using the Georgia numbers, Bama’s chances to run the table goes up to 48.9%, and LSU improves to 36.6%.

The undefeated SEC Champ scenario goes up to 85.4% — and I believe your other probabilities are augmented as well.

(Don’t quibble with the adding above… it’s a rounding thing.)

Huck, I don’t see odds for a SEC CG. Does the PAC have one?

Oops. Now I see it. Sorry.

Actually, TTR, you caused me to notice another mistake. Stanford’s table has been edited because their CCG would be at home. That’s how the probability was calculated, just a transcription error in the table.

How does your math change if you go ahead and change Kansas’s chance of beating Texas to 100%? We all know that we can’t beat that warlock. I’m actually encouraged by the fact that the computer gives us a 40% chance to win.

I was kind of surprised that Wisconsin’s schedule ranked so poorly, but it makes sense when you actually look at their schedule. Of course, I’m sure they weren’t counting on OSU being so bad. That is the problem with scheduling one decent non con 6-8 years out, as we learned the hard way in 2008. You never know what you are going to get.

I know USC is down, but holy hell. 4-5 years ago I would never have guessed they’d be in a position to have a calculated % chance of beating a team less than 10%.

Great stuff, Huck. Thanks.

Re: USC. Just goes to show you what a combination of lazy recruiting, NCAA sanctions, and hiring Lane Kiffin can accomplish…. Sorry. No need to go down that road. Forget I mentioned it.

Scip – any chance of some post mortems for Okie Lite? or maybe some midseason analysis?

Wisconsin having the same ending SOS as Boise should be shouted daily the rest of the season. Shameful. And ’Fuck Delaney."

what are the odds Texas finishes 10-2 this season ?

so huck, say bama and ou make it to three games left still undefeated.

with 3 games left Alabama has a 1 × 0.978 × 0.925 = 90% chance of going undefeated
And ou has a .948 × .998 × .677 = 64% chance of being undefeated.

Would you say the chances of either ou or Bama being undefeated is 154% ?

K-State will gp 10-2 with respectable losses to OU (30-20) and OSU (40-30) and beating A&M at home.
Does Sherman survive a 7-5 year (losses at KSU, OU, and at home vs. either Mizzou or UT)?

FSU’s about to “rebound” with 4 of next 5 at home (road game is BC) and toughest home game is probably Miami.
OU’s lost 3 in a row in Lubbock but one of the losses was controversial (05) and another came with Bradford injured (07).
Good stuff as usual.

Do the LSU numbers reflect the three suspensions for the Auburn game?

Great stuff, thanks for the analysis.

blacklab -

No, because those are not mutually exclusive events so it is a totally different situation. The probability in that case is 0.9*0.64 + 0.9*(1-0.64) + (1-0.9)*0.64 = 0.964. That’s the probability that both go undefeated plus the probabilities that each goes undefeated while the other does not.

As for the probability that LSU or Alabama is undefeated, those are added in the same way that the odds of rolling a 1, 2, 3, 4, or 5 on a single roll of a die is the sum of each possibility because they are mutually exclusive. Consider LSU the 1 and 2 while Alabama is the 3, 4, and 5. This is the correct approach, actually a shortcut, to take because the same three addends above are different. If LSU goes undefeated then the probability that Alabama goes undefeated is not 45.7%, it’s zero. And if Alabama goes undefeated, LSU’s probability is not 34.7%, it’s zero. So the probability that either one goes undefeated is 0 + 0.457*(1-0) + 0.347*(1-0). Which obviously resolves to simply adding their individual probabilities.

These ’s are way off. Stanford 90+ vs Notre Dame? There’s a good chance that game is lined under a touchdown.

This is completely idiotic… Oklahoma State win percentages are laughable…

98.6 percent likelihood to beat Baylor at Home
97.7 percent likelihood to beat Kansas State at Home
97.6 percent likelihood to beat Texas Tech on the road

It is absurd to think that Oklahoma State goes to Lubbock and wins 98 out of a 100 games this year. Kansas state is playing great football, and somehow they would only manage to win 2 out of 100 games against Oklahoma State?? Get real.

Games that would have lines under 10 points are not 98% likelihood to lose the game straight up. Oklahoma State at home against Kansas State will be around 10… at Tech around 8… and at home against baylor 15.