Very interesting. I find it reassuring to see statistics detail out that ND and Ohio are not as good as their records. Fortunately Ohio is on probation so we don't have to see the two lucky teams play for the national title.
The Luckiest Teams of 2012
going 12-0 is a often a recipe for this, but especially this year
With the pre-bowl season officially under wraps for 2012, it’s time for my annual review of teams whose record most greatly deviated from what it “should" have been.
To (attempt and fail to) avoid confusion, here is how I define Luck for this exercise.
What I Am Measuring
Luck can mean a lot of things but for this, I am comparing a team’s actual wins this year versus taking their opponent adjusted performance and re-simulating the season with the exact same schedule. Two teams who play a tightly contested game are roughly the same on that Saturday. Over a long horizon these wins and losses tend to even out but over a 12 game season there will always be teams whose final records don’t quite match how they played throughout the year.
What I Am Not Measuring
I am not looking at any preseason expectations. I am not looking at how each team did versus the recruits on their team. Those two would look at over-achieving teams of 2012 more than lucky. I am not going back to individual games or plays to look at if one or two games would have been different. I am also not looking at injuries on personnel changes throughout the year.
Think of this exercise as a sort of Pythagorean Wins for College Football. A lucky season is a great one to have for a fan, because no matter what the expected value is, the end result is all that matters in looking back. But like Pythagorean Wins, “Luck” is a great starting point for looking ahead. There are a lot of different ways to get to the same record. Last year Texas A&M had the most unlucky season in the country and was nearly 4 games below their performance. Kevin Sumlin did a great job this year and having the Heisman Trophy winner certainly helped, but Sumlin’s team was in a much better position than their prior year’s record would have indicated.
Teams with great records are rarely unlucky and vice versa. The formula is [Actual Wins] – [Simulated Wins]. If you win most all of your actual games there is very little room for your simulated wins to be higher. It’s more a factor of math than destiny.
Coach Hoke’s alma mater was 2012’s luckiest team. Ball State was simulated to win 6.4 games this year but pulled out a 9-3 record. Beyond that, three of the four teams following Ball State are of high interest to Wolverine fans.
|Team||Actual Wins||Simulated Wins|
Michigan’s two biggest rivals and bowl opponent all crack the top 5. As noted above, Ohio St and Notre Dame were easy candidates for this list with perfect seasons, but their perfect seasons were the luckiest undefeated seasons in the seven years I have been measuring the luck factor, and by a considerable margin.
Michigan ended the season slightly lucky with 8 wins versus an expected 7.6 based on their total season performance.
Of the teams that finished the year with 2 or fewer losses, Florida State is the only team to finished at least 0.5 games unlucky, thanks to their upset to NC State and an otherwise weak ACC schedule. Their loss to the Wolfpack was the 7th most unlikely outcome of the season based on the simulation but the most likely outcome based on Seminole history. Of the Top 10 biggest upsets looking back, five happened in Week 1 and all by road teams (Youngstown over Pitt, McNeese St over Middle Tennessee, Tennessee-Martin over Memphis, Ohio over Penn St and Iowa over Northern Illinois). Only three of the top 10 happened after the second week of the season with
UMass topping Western Michigan and Florida Atlantic over Western Kentucky joined the NC St upset. The Ohio-Penn St game was an interesting one because people acted like it was at the beginning of the season even though it really wasn’t at the time. By the end of the season Ohio had tailspinned and Penn St turned out to be a much better team.
The unlucky list features some of the same teams from the biggest upsets above
|Team||Actual Wins||Simulated Wins|
Michigan State was a few spots down, as they finished nearly 2 games below their simulated totals, falling on the wrong side a few too many 16-13 totals.
Is This Luck Repeatable?
Almost certainly not. The scatter plot of current year versus prior year luck:
There are a lot of teams in each of those quadrants, each season is its own animal. Notre Dame’s was nearly 2 games above simulated this year but was –5.5 over the last three. Those who remember Northwestern as the team continually defying expectations. The Wildcats continued this year and are one of only two teams (Rice) who have had above average luck for all seven years. With Wake Forest right behind them I began started to draft a “smart schools are more lucky” section until I looked at the rest of the all-time top 10 and saw Middle Tennessee, Kentucky, Auburn and Ball State all on the list.
When you look at the spread of lucky years by
Count of teams by number of lucky seasons from 2006-2012
The twin peaks could mean there is a lucky and unlucky group, each normally distributed. It could also just a be bump in the data or it could be part of the fact that wins by program is somewhat consistent and luck is slanted if you are at one end of the spectrum. My biggest conclusion is that most of it is truly luck but that there is the possibility that teams like Northwestern or coaches like Les Miles have a true ability to consistently win more than they should but also that statistically, teams like that are bound to turn up even if its truly random.
To go undefeated and play for a national title in college football, I feel like many teams have had lucky breaks, that's part of what makes the sport special.
But there's getting a break or two to fall your way and then there's the crap ND and OSU have gotten away with this year.
Ohio I don't know got an excessive about of luck; only the MSU fumble really stands out to me. ND, however, has gotten every break and every call. They barely beat a terrible Purdue team, they only beat us by a TD even though we gave them 6 turnovers, they needed two gifts from the officials to beat Stanford and Pitt, and they managed to get USC one week after their starting quarterback went down with a shoulder injury. That is more than luck, that is backing into a title game, but I guess that's okay because you need a lot more than luck to beat Alabama.
Like the #1 defense? Cornholing Oklahoma in Norman? Question the offense all you like, but you cannot question the defense.
Which brings up an interesting point. In baseball, the pythagereon therom becomes less accurate for teams with exceptional bullpens. This results in a team winning a lot of close games (1 run games, extra inning games, etc). I suspect there is a similar issue with teams with great defenses, especially coupled with so-so offenses.
Alabama had a bit of luck of their own. Two undefeated teams had to fall in the polls ahead of them and they did. Georgia couldn't figure out how to manage the clock on the final drive of the SEC championship game and undefeated Ohio State can't go to the NC game... I'd say they're some lucky SOB's as well.
I think you've completely missed the point of the whole damn post.
Sorry if I'm missing it, but can you elaborate a little more on how "simulated wins" is calculated? Not to give away your secret sauce, but are you essentially taking an opponent's expected performance (based on the body of the season) and assuming they perform at that level that particular saturday that they lost?
Also, what has our luck looked like over the last 4 years? By observation, I have to say last year's team was fairly lucky, this year's team was lucky at times (N'western) and unlucky at other times (Losing Denard, not having Devin ready for Nebraska).
Not sure where you're getting your data, but Western Michigan defeated UMass (in a blowout actually, 52-14). Might want to double check some of the results in your data set in order to get a more accurate metric.
WMU had 4 wins on the season, not 3. Perhaps just the UMass - WMU result was wrong or maybe there are other games with the wrong result.
I wonder what it'd be like to compare the points in each game where a given team was most unlikely to win (kind of like Michigan's 10 greatest comebacks a while back), and then calculate the discrepancy between them winning and losing over a season. I bet Nebraska would be ranked near/at the top.
...and Iowa over Northern Illinois)
A humbling moment for Hawkeyes everywhere
Yeah, I actually had to read that part twice, stop, think and remember that Iowa sucked and NIU was good.
to not see Arkansas and Auburn on the unluckiest list. Were they actually projected to do that poorly in the first place? I would've thought they deviated from their predicted wins quite significantly.
Arkansas...well, going from Petrino to John L. Smith probably dropped their expected wins by about 4, but, yeah, Auburn was much worse than I thought they would be. I expected them to be pretty mediocre (say...7-5) but they were just god-awful.
I am a little surprised not to see Nebraska on the "lucky" list. Jeebus, it seems like those guys pulled one win after another out of their a** all year long. Personally, I was thrilled to see them get pummeled in the B1G championship game even though it meant a 7-5 team would represent the B1G in the Rose Bowl.
I would've thought Oregon State would've made the list, too. Every other year they have a losing season. Last year they went 3-9 losing all out of conference games, this year they went 9-3 and finished the year ranked 13th
I believe the Mathlete calculates how good a team is 'projected' to be not only based upon their previous year's record, but a bunch of factors such as passing yards and receiving yards returning from last year. Maybe Oregon St returned like everybody on the team and thus had a projected win in the Mathlete's calculations that were far above the conventional 'look at the team and guess how they'll do' type projections we see in magazines all the time.
How "lucky" was Michigan's team last year?
Not lucky, they just had a favorable schedule. They outperformed everyones expectations but I think that was coaching more than anything else. This years 8-4 was more like what this team was capavbel of...maybe another win or two would be about right since this years schedule was brutal; the polar opposite of last years.
so I'm a believer of the notion that you can largely create your own luck, however there are some aspects of last year (unlucky - no PI in Iowa/lucky - VT "catch" was not a catch) that made me wonder about it
Though this isn't what the Mathlete is measuring in this analysis, so "luck' as it is defined is roughly (as I interpret it) a team overperforming against good competition, and so I wanted to know how the team did last year.
Couldn't these teams also just be the ones for which your method is the least accurate - for whatever reason? You're essentially saying that teams that won more than you expected got "lucky", but what if your model is just plain wrong for some teams?
Nobody has made a luck of the irish joke yet? wow.
Am I missing something? But how did you get the "simulated wins"?
A bunch of architecture students in Muncie are shaking their fists at the "smart schools" comment.
Or at least they would be if they were allowed to recreationally use the internet, or allowed to leave the Architecture building.
this is really interesting! would it be possible to see the entire list? also I would love to know how you evaluate teams' strength to decide the expected outcome of each game. I only ask because I am looking forward to sharing this with some ND friends and would like to be able to explain how the teams were evaluated. great work, thanks!!
also where did Nebraska end up? They seemed incredibly lucky