Nice work, but I think this is just more evidence that BMI is a useless measure for athletes and body builders.
I did not make this headline up
Since there isn’t much sexy in the fall roster updates (so far at least…despite Denard and co. anecdotally touting the gains of the guys upfront in the B1G presser) I thought I’d look at the roster info and see if I could represent the data in a different way.
Body Mass Index (BMI) came to mind – as both an indicator of fitness and body type. In this diary I’m looking at BMI and trying to see what this might indicate for 2011 in the B1G using the results from last year.
Here’s how BMI is calculated along with an explanation (thanks to Wikipedia)…
The body mass index (BMI), or Quetelet index, is a heuristic proxy for human body fat based on an individual's weight and height. BMI does not actually measure the percentage of body fat. It was invented between 1830 and 1850 by the Belgian polymath Adolphe Quetelet during the course of developing "social physics". Body mass index is defined as the individual's body weight divided by the square of his or her height. The formulae universally used in medicine produce a unit of measure of kg/m2.
So what I’m proposing is looking at muscle/fat per square inch. It’s going to be similar to just looking at weight but not quite. In addition I’m going to look at the entire B1G plus ND. Here we go…
I went to each school’s website and pulled their roster and went about computing muscle (or fat as the case may be) per square inch. I thought I was doing everyone a service by pulling directly from the source. It didn’t turn out that way…here’s what I got…
Starting with this variability chart…
You may ask why the heck the schools are ordered like this left to right. I’d like to say it is the leaders on the left, the legends on the right and ND in the middle. But it’s nothing as nonsensical as that. I’ll get back to this. But first…a few comments…
Std Dev. doesn’t play much of a factor here wrt Avg team BMI.
The green diamonds are mean diamonds. They are useful (to me at least) to see if distributions have significant differences. Two distributions are significantly different (p<0.05) when the diamonds do not overlap…i.e. NU is significantly more puerile than Penn State. I say puerile but BMI is not a measure of manliness necessarily…is it? [Ed-M: gracile/robust perhaps? #anthrobionerdiness]
What might the mean differences indicate…it could be team fitness…it could be style of team play…it could be roster management…it’s definitely all these things and more.
So what does Avg. team BMI have to do with being good at football?
BMI in a traditional sense signifies four body types.
Roughly broken out like this chart.
In Football the vast majority of any team is at least overweight on this scale. Per the chart above the average B1G player is nearly obese (29.6). Football players don’t play by these rules however (actually this is sadly not the case –
- but I am not going there right now. Suffice to say – don’t let your BMI get out of control and/or if you must…then lose it when you quit playing or suffer the consequences.)
Since the 60’s at least we have seen larger and larger men take up football. BMI has risen along with these weights.
1969???…I guess Bo started this issue…
Anyway…not all BMI is good for football.
This does not equal This
But then again…one of these guys carried this off the field on his final college game.
While the other guy carried this…long may he not shave…
So what does BMI have to do with being good at football? I guess the best way to look at that would be BMI vs. Wins…
Hmm… let’s regress for a few moments…
OK…so there’s no significant correlation (p<.05) but it’s not like there is nothing going on here. Still it’s not worth extrapolating moments to 2011.
I took a look at some of the names behind the dots…(sorry for the small type…I can’t fix that evidently with this software.)
Ash and Campbell are outliers for Michigan. Relative to other DL types like Nix Jeremiah and Hankins they aren’t so out of control however. Ragoo is simply a very large person. Position group is a factor here however – which is reasonable.
Here’s the breakdown by position (again for the B1G+ND)…
I kept many of the same labels…
Hemingway is a different body type altogether than the rest of the B1G receivers.
Ragoo and Cully are interesting. Both are significant outliers for their position group. Where you might think that Ragoo is a (6’7” 391 lb) Terrence Cody-type lineman (there are no OMG shirtless search happy images of him to compare) Cully is a 5’1.5” 197 lb frosh for Penn State whose team photo is pretty revealing…
Johnathan Ragoo (OL – Minn) Jeff Cully (DB – Penn State)
As stated before Ragoo is huge. Cully is tiny. Both are BMI kings of their position group. Cully is MAC-sized. Cully is stubby legged. Cully is a walk-on (I presume at least.)
This brings up a dataset issue. I pulled everyone on each teams' roster when I imported the data off the Web. I wanted the best comparison possible, but when I went to each team's website roster I found them in various formats and protocols. Penn State does not list redshirts. The position groups have different nomenclature school to school. Nebraska (don’t get me started on Nebraska) had 152 players on their roster. Notre Dame only had 78.
WhiskeyTangoFoxtrot Nebraska. They are an orange to the B1G apple…
I leveled the rosters by position and threw all redshirts with their graduating classes. I did not ferret out the walk-ons or separate the projected starters (it was simply too much work.)
Here’s the offensive line data for the B1G. This is clearly the most significant of position groups.
Taylor Lewan is bottom of the group for Michigan. This is likely a sign of fitness on his part (and Schofield and Gunderson for that matter) [Ed-M: Lewan was also a project recruit who started football later than most][Ed-TSS: Glasgow is a project as well yet is above the mean for the subset of Michigan OL].
Frederick for Wisconsin is in the top of his group. This is likely an equal if opposite sign of fitness on his part. This is clear when looking at the same chart versus weight. Here’s that data:
When you look at weight alone Lewan/Schofield and Gunderson pull up to middle of the pack. Frederick falls down into the mid distribution. This isn’t true for other laggards like Flavin and Tansey or for Ragoo who are still kings of both charts.
Sorry again for the labels …
Though overall BMI for OL correlates well for 2010 Wins (p<0.06) – the Std Dev. of the BMI data correlates even better (p<.016). Using both with a weighted leverage to model the wins brings the p<.0155.
Again the labeling issue. Mich and Penn State are overwriting each other here.
Here’s a graph of the residuals; this doesn’t overwrite if I stretch it enough.
Teams above the line outperformed the model; teams below did worse than expected.
In 2010 the biggest OLs with the tightest distribution of BMI won more games. Not a huge insight but insight non the less. Basketball on grass was not working last year.
The good news for Michigan is that the incoming class of OL is the BMIiest in the entire conference. Adding those numbers to the previous charts gives Michigan a considerable bump in beef.
Posada and Bryant have identical BMI. Yerden and Williams at too close to label simultaneously but the difference is pretty stark vs the previous chart. This new class is a different kind of beast.
I’m not saying this new class of OL are going to be the difference between winning or not this coming year (I would like them all to red shirt if possible), but our offensive line is looking good with respect to muscle fitness and incoming beef.
Thought I’d share some data. First time diary. Must go to sleep now. Go Blue!
Nice work, but I think this is just more evidence that BMI is a useless measure for athletes and body builders.
I've always hated BMI. It tells you absolutely nothing. Muscle weighs more than fat, therefore if you're very muscular with 5% body fat you can still be considered "overweight" according to BMI.
Nevertheless, interesting to see how we stack up. Thanks to OP for taking the time to do this.
I hate to pile onto the OP since he did a lot of work here, but as others have said BMI isn't really consider scientifically valid anymore. It's still used grudingly in online surveys since you can't measure body fat over the internet, but thart is about it.
With BMI if you are an average sized human male (so 5' 10" and in the 160-170 lbs range) and you lift weights for 2 hours a week over a 2 month period, you'll end up repeorting on the high end of the obese scale. If you do seriousily weight training you'll report at morbidly obese within 10 weeks (normally less).
The whole problem though is BMI can't tell the difference between good weight (Jake Long) and bad weight (Mr. Plow's love handles and gut).
I also hate to pile on, but BMI is entirely useless when it comes to sports. BMI is, at best, a screening tool that you can use in a primary care setting to screen average NON-athletes for obesity.
The OP has certainly put in a lot of effort, but this isn't really even worthy of much discussion MGOBoard, let alone a Diary. I guess it was bumped there for effort, and perhaps quantity of content.
Here, from the Mathematical Association of America:
if you have a pound of fat and a pound of muscle. The muscle weighs more? Muscle is more dense dude!
Yes, if you have the same volume of fat and volume of muscle, the volume of muscle will weigh more. You're correcting something that doesn't need to be corrected because the sentiment is identical.
would differentiate the witches from the warlocks. The reality is...there is no "Good at Football" metric - especially not available to the fans. Tony Mandarich sure looked good didn't he? Even the experts miss with all the data.
I dipped into this when I said football players don't live by the same rules wrt obesity that the common population does. The fact that the highest third of BMI inflated NFLplayers have a 600% higher mortality says that muscle or fat - there is a problem there. I don't think Frederick is stronger than Lewan - but I don't have access to that data.
As stated at the beginning this is just another way to look at the height and weight data.
There are several take homes here shown and not.
We all really appreciate the work, it obviously took a lot of time/effort. We were just discussing the merits of BMI, and why people should take this data with a hefty grain of salt.
Also, with regards to the higher mortality rate, this is my thought on the reasoning behind it: the highest BMIs are commonly linemen, and the style of play for linemen makes them more prone to head injury. The constant banging of heads is more detrimental to the brain/body than I think their weight problems are-- there's a reason footbal players have a higher incidence of these problems later in life. I think attributing it to BMI is a little premature. Correlation does not equal causation.
Again I would like to reiterate: thank you for the work, and it is a very interesting read.
I can't find the 1992 NIOSH report on BMI and mortality of NFL Vets online at least. I have followed the popular press in its examination of Post Concussion Syndrome (PCS) fairly close and I don't recall a discussion of CV disease either symptomatic of PCS or as a contributing factor - but that admittedly means nothing.
Note I misquoted the stat in the study above...the highest BMI group (not the upper third) have 6X mortality as compared to the lowest BMI group of NFL veterans studied. It's not clear what that means without seeing the study - but it's not the upper third. Here's a section of the article linked in the diary - that I misquoted...
...when players were stratified according to level of BMI, those in the highest BMI category had a 50% increase in mortality , versus middle and lowest BMI categoryplayers. Also, players in the highest BMI group had a six fold increased risk of CV death compared to lowest BMI player groups, which most closely resemble the general population. In this study, no controlled health testing were performed and CV risk analysis was not obtained.
I'll give the sincere disclaimer that I am really not trying to be/sound like a jerk here, but would you mind mentioning what the take-homes are here? How is Lewan's lower BMI a "sign of fitness" but other's higher BMI correlating to better performance? What is BMI adding here that simply looking at the OL's weight does not?
P.S. I think that putting this much work into evaluating BMI was probably a waste, and had initially down voted based on frustration with BMI/assessment of the content here. I switched over to an up vote, however, because even though it seems this turned out to be a long way of saying nothing, I do appreciate the effort.
Wow...that's a lot of information to work through! I'm glad you narrowed the focus down to the OL. As we all know, OL/DL results have everything to do with team’s success.
I find when working on a project like this, the guy compiling the information comes away with more insight than can be relayed with the graphs and data summations. I think that has to do with the mind organizing the data (in the background) as you are working with it, to support a theory you believe in.
The last paragraph you write is case in point. Your gut feeling is our OL looks good for future results.
Thank you for pulling this together and sharing it with us.
there is a story there as well...but turning hips and back pedaling is much less intuitive a leap from the height and weight data. I can't wait to see Avery in action this year. He is a lowball BMI DB who is going to make an impact this year with his agility IMO.
It's really too bad we don't have uniform speed data for the guys off the ball. For that matter it's too bad we don't have uniform height and weight info.
I looked at 3-4 and 4-3 LB weight and BMI deltas - but at a certain point you start to make up the data. I tried to keep this as level and simple as possible.
Wait a minute...BMI has a unit of kg/m^2??? These are the units for area densit, hmmm I did not know that.
It is a measure of density, but not like you would think: it does not differentiate fat from muscle. A body builder could have a much higher BMI, but have significantly less fat. Therefore without a real body fat percentage measure (caliper, water dispacement, etc) BMI gives you no information at all.
BMI is interresting but it isn't really all that useful a measurement. Using BMI as a stand alone measurement doesn't get you very far.
A team of all upperclassmen will have significantly more body mass than a team of underclassmen. They have an extra year or two of natural HGH/chronological growth and a year or two more of having been in the training program.
improvement in BMI class to class. The players who gained were drowned out by the players who lost though this is just my feeling man. I didn't take the time to look at the signal too much. The delta by position was in the limited scope more interesting to me. There is considerable drop out in the raw data going from under to upperclassman.
Nebraska's huge walkon numbers denigrate the comparison and the overall signal as well. All the work is in getting good numbers to begin with.
That BMI for most players measures how long they've been in a university-quality weight training program.
So as a surrogate for another couple of measures--physical maturity (whatever that might be) and experience (which we know is important), it might be telling.
It would be interesting, if someone wanted could sort out the data, to compare S & C programs. It might be that someone who employed Barwis could legitimately say to a guy, "You'll be lighter and leaner when you graduate," would have an advantage to someone who doesn't want to play OL in the NFL. But that's a real reach.
Final thought: the last year Michael Jordan won the MVP, his BMI, per the Bulls roster, was 27.
So, what you're saying is that smaller countries will unite and take down the superpower that was America?
Just like Europe, amirite?
Interesting data, there are drawbacks to the BMI as others said above; the position breakdown is the most interesting to me. I would be really interested to see where ND players stack up in each category, would there e an easy way for you to color code a specific teams players?
I imagine UM fans would be interested to see the same
Thanks, I appreciate it, I look forward to it
The comments about BMI have already been made, so I won't make them again.
I would, however, like the thank the OP for the work he put into this one.
Just becuase an experiement doesn't work out how you wanted it to, does not make it a failure. We learn from every result.
Meaning that we learned something by finding out that BMI is not a great predictor for football success. At least not at the college level.
I wonder if it may be different in the NFL due to players being professional athletes who have spent several years in weight training programs?
Excellent chart. I'd be interested to see an adjusted predicted wins chart using your information, taking into account the possible number of conference wins in a season. It would be possible by evaluating the head to head match ups.
Good work by TSS. I'm not sure if proves all that much beyond Wiscy likes big guys and BMI is a very blunt tool for measuring fitness of players. I know it is not readily available, but if we are going for the Cody-versus-Long fitness comparison, stats like VO2 Max would be a more telling barometer. But I am not going to hate on the effort put in - sometimes it helps confirm a bit of common perception with some numbers, even if the numbers aren't the best example.
BMI is most useful as a rough proxy for body fat percentage for "normal" people, that is, those with average levels of muscle mass. It's utterly useless when it comes to analyzing some of the most muscle laden athletes in all of sports. An analysis based on penis size would have equal usefulness.
at the 2006 OSU game i noticed that one of OSU's linebackers had not so great size when he was standing captain morgan style.
That's a great idea. Who has a ruler?
That National Championship Trophy is pretty cool, where can I see it? I've been to Schembechler Hall and didn't see it? Is it kept somewhere else at the University, maybe the Presidents office or something? I guess that would make sense.
Any data on 40 yrd dash speeds to add to this data since mass x velocity = force.
Sorry to ask a non-football question but what program did you use to make these charts? I'm currently doing a project where I need to be able to graph something similar (testing various political factors and seeing if it determines, and hopefully can project, how a congressman will vote) but I haven't found a great program yet.
Besides that... amazing amount of research. Congrats.
I used JMP 7.0 and R...I'm transitioning to mostly R based analysis as that is where it's at IMO. At work I currently use JMP 8.0.