Yeah, I figured it was taken into account. And you're right that I was thinking just of Coaches / AP rankings not KenPom. But still, man. Can you imagine a top five football team ever being a dog to an unranked team, even on the road? Just surprised about either (a) how much of an effect homecourt has in bball or (b) how much more parity there is across college bball.
Does anyone else think it's nuts that an unranked team can be favored over the #1 team in the country in Kenpom rankings? Not nuts like it's incorrect, but nuts that this actually seems semi-reasonable? Homecourt advantage is ROUGH in college bball.
This is really helpful. Depressing, but informative. I was hoping that the downward slope would be consistent across all the BIG teams. I shudder to think of what your line would look like if we didn't go ham in the last game.
IME this is the issue to consider when debating the future of Gorgeous Al. I'm sure there are mitigating factors. The experience stuff I did suggests that youth probably plays a significant role, especially with our young interior line. Personnel is probably a big part of it too.
My main point:These mitigating factors influence season averages of things like YPP, YPC, YPA, and sack percentage when compared to other BIG teams. They shouldn't, however, play a role in determining the slope of the intraseason progress lines you're producing. If anything, we should expect to improve more than other programs over the course of a year because of our youth, since younger players can be expected to improve more than older players.
If I were Davey B., this is the thing I'd be looking at.* I don't care that our YPC this year was 3.07 and the BIG average was 4.5 (actually no clue). All I want to see is improvement. It's not like we're starting from an elite point where the only place to go is down. Unfortunately, I just don't see it with Alan "Al" Borges. I hope I'm wrong.
* I'm not Davey B., and even as Gandalf I have no control over the affairs of the athletic department. So this really doesn't matter. It's just the 16,000th opinion on the subject.
I'm surprised that "yards per attempt against" and "completion percentage against" isn't a standard metric for DBs. I agree that INTs and PBUs are pretty meaningless. It doesn't seem like it'd take much extra work to calculate these stats. It's like judging a QB just on the number of TDs he throws; it tells you something, but only a fraction of what some slightly more advanced stats tell you.
It's really helpful to have all the one-score games compiled like this. If anything, I think this suggests that our record pretty accurately reflects the way we've been playing. As a guesstimate, I'm assuming we shoud win about 60% of one-score games due to usually having a slight talent advantage (at least based on recruiting star levels).
If we saw Hoke or RR was something like 11-1 or 2-10 in one score games, I'd say that we were particularly lucky or unlucky, and that maybe our overall record doesn't really reflect how well we were playing.
But Hoke is 9-7, and RR was 7-7 in the one-score games compiled here. This seems to be just about right. It's easy to just look at the ones that didn't go our way and think we were a couple plays away from greatness, but this year we were equally close to 2008 levels. Nice to put some numbers behind those gut feelings though. Thanks, Brhino.
From what I've read, plasmas are great for darker areas (e.g., a basement), but you'd probably want to go LED or LCD if it's in a room that gets a lot of light. I'm sure some people have strong feelings about it, but IME I think they all have pretty sweet pictures.
Yeah that's a good point. Talent levels would probably look more like RR's tenure. Not bad, and better than Indiana, but not where they are now. You're right that some of the attraction of Michigan is the pro-style offense.
No problem, glad you enjoyed it! I agree that it'd be useful to look at RB experience/talent, especially with the way our RBs seem to get blown up running and passing the ball.
Fitz gets destroyed in pass protection and he's a senior. But so do the young guys and they're, well, young. I guess that's why the large FBS-wide analyses are useful, since they can help point a general finger towards either youth, talent, or bad coaching being the problem.
Any idea if Indiana's OC is the guy responsible for their up tempo offense? Or is it Kevin Wilson? Either way it looks AWESOME, and if it is their OC, surely we could poach him. Running that kind of offense with our talent (star rating-wise) would be interesting to see.
Yeah, I agree that salary's a pretty poor proxy. Tough to measure "talent" though without it being tautological. I was just trying to think of ways to to explain teams' best OLs playing LT, and I thought one reason might be because that is what preps them to become the best pros (i.e., LT is the most important position in the pros).
I was thinking of the Lawrence Taylor thing too, but any idea if there has ever been a quantitative study on where most QB injuries come from? I was really suprised that sacks tend to correlate the best with the interior of the line, so I wonder if a few anecdotal weak side hits have somewhat entrenched this as a fact. No idea how you'd prove it without some serious work, but it's interesting to think about.
I'm trying to think through this, and I'm still not sure I quite have my head wrapped around it. I mean I get that survivorship exists. But my initial reaction though is that this would actually decrease the r-squared that we get when we look at experience and success, meaning that there are probably a good number of freshman and sophomores in there because they're good and not because they have to be. If they're good they should help produce more YPC or fewer sacks, and if younger people are producing fewer sacks or more YPC, then that should lessen the correlation that we see.
Interesting analogy to the Bill James stuff. I think that's essentially what we see with a lot of the freshmen who are starting. Reflecting on this, I think that perhaps the best metric for experience along the o-line might be the oldest person on the 2-deep at a given position. So if Bosch is backed up by Bryant, the position gets a 2.5 (RS Soph) instead of 1.0 (FR).
Yeah, all else being equal, I think this would say would should improve on the line despite losing some good tackles. There are enough Michigan-specific factors that it's really impossible to say with any certainty, but I'd be the experience on the interior helps us more than losing the tackles hurts us. One can hope, right?
I think I get what you're saying. Basically that older o-linemen are better not because they are older, but because they are just more talented or lucky (and thus fend of younger guys). I guess it's possible that this influences the results to some extent. But shouldn't that survivorship bias show up at the tackle position as well as the interior?
Also is there evidence showing that mutual fund managers don't get better with experience? It seems logical to me that OL would get better with experience, and so if mutal fund managers don't and OL do, then I'm not sure the analogy applies.
And it's certainly possible that outliers drive some of the trends. I would tend to doubt it with the fairly large sample sizes. But it's possible.
Honestly, I've beaten the dead horse about as much as I care to. But I'd be interested in seeing the results if you carry out your proposed study. Not sure how you would measure "how good" an OL was after year 1 or 2 and then "how good" they are now.
Nice work, Mr. Utah. Have you thought about trying to visualize the data in a graph of some sort? Might help your argument come across even stronger. Thanks for the expanded look at youth though. I know it's kinda beating a dead horse, but it's also nice to have some reason for optimism.
This is really interesting! Thanks to everyone who's done the comparative / contextual work as well. That stuff really helps. My gut reaction is that things aren't that bad...except for the last four games where things have COMPLETELY FALLEN APART. With cupcakes excised we've still got more above the break even line.
Is it encouraging that we're still generally above the line? Or discouraging that the second of this season is a tire fire? I don't know. Both?
Does anyone else just wonder why we couldn't have RichRod but with a Mattison DC? Beating a dead horse, I know, but man. I mean there's a decent change RR and BH end up with the same record in year 3, except RR was trending up after getting his own players and BH is trending down after doing the same. Why can't we have nice things?
Interesting AND depressing! Nice combo. A couple thoughts...
Taking each teams YPPA against is a really good idea. Well played.
I think we might expect a downward slope even when teams' YPPA is taken into account, since 3 of our first 4 games are against crappy teams from crappy conferences. We would expect to outpace CMU and Akron's opponents in terms of YPP, while the same expectation might not be there for Nebraska or Northwestern's opponents. This would lead to the early data points being high on the graph causing to a natural downward trend.
Somebody so inclined could test this by looking how many other BIG teams also see a regression in YPP when YPPA is accounted for. If 1/2 the teams regress while 1/2 improve, then we're really looking bad. If 80% of the teams regress, it probably is in part due to scheduling early season cream puffs. Free diary here, get it while it's hot!
Any chance we can see how whether we regressed the past two years as well? You know, as long as you don't have anything better to do. If the regression is an every season thing, that's...suboptimal.
This is the best thing about rooting for the bball team. I mean we lost, and we're probably the better team, and it sucks. But Beilein's history shows that over the time the coaches will adjust, the players will develop, and the team will improve.
Hey Brian from 2006. Somebody linked to this in one of the threads and I noticed it didn't have any comments! Great post and great game. I really liked what the defense was able to accomplish. Tell them to look out for OSU at the end of the year though, that one looks like it might be a shootout.
Also, did you know that four years from when you posted this that our offense will be putting up crazy yards while our defense sits around eating baloney? And three years after that our offense will put up negative rushing yards multiple times in the season, but our defense will be quite solid? Ahh the pendulum of time.
The story of my MGoPoints began long ago. Back in the haloscan days, to be exact, when a one sentence update was all you would get about a 5* DE commitment. Ah, to be young again! Of course I wasn't Gandalf the Maize back then, merely Gandalf the Blue. I wandered far and wide in those days, encountering things more horrifying than one dares to imagine in the current age: the Horror, Armageddon, GERG. It was only when I was able to sublimate these terrors to enhance their antipodal highs - 100th game, UTL I, Denard - that the transformation was complete. But I digress...
Actually, our rushing woes led me put together a study on how o-line experience correlated with run game success. I wanted to post it as a diary but had no points. So I sent a raven to Brian, and he gifted me my first hundred so I could post. Misopogon ended up selecting me as the diarist of the week, and in addition to solidifying the new greatest moment in my life, I think he gave me like 200 points as a reward.
The rest of the story can be found in the Silmarillion, but perhaps those tales are best left for another day...
This is good shizz! Really interesting study; it's nice to put Kalis' development in perspective. Even as a five star you're more likely to never start a game than you are to become an All-American. Probably wouldn't have guessed that. Thanks for taking the time to do this.
I think I would give Borges a positive here for QB development considering the increased YPA, despite the fact that completion percentage stays about the same and interception percentage increases slight from pre-Borgesian times. Brian did mention he had a full time QB coach at SDSU, so make of that what you will. Thanks for pointing this out.
The question of whether Gorgeous Al has a track record of developing quarterbacks is a really important one. My first thought is that it should receive a bit more treatment in a post of its own, but with all that bouncing around it sounds like it's difficult to say too much.
The guy's been a D1 OC for 20 years if you include Boise State in the early 1990s (which was actually D1-AA at the time), and we have all of 3 data points for whether his quarterbacks improve. McNown = yes, Cox = no, Paus = maybe. If you include Denard and Devin then that's probably 2 more in the "no" pile.
My 2 cents?
Cent 1 - It's not the smartest thing to hire someone who has bounced around so much because it's really difficult to tell whether they're (a) good at long term development, (b) good when taking people by surprise (i.e., before the rest of the conference gets wise to them), (c) come into good situations, or (d) just aren't very good. I realize this is hard to do in big time college football, because programs constantly take fliers on people who have shown success after two or three years so there aren't many people around at lower level schools who have demonstrated long terms success.
Cent 2 - I'd bring in a dedicated QB coach with a track record of QB development even if it's at lower levels. No clue about the number of assistant coaches you can have, but this seems like a way to potentially improve the situation in future seasons without the risk of mixing things up too much (e.g., if you fired the OC).
I'm with you in "we should be spreading the field out because we can't block" department. My initial thought was wondering whether this trend held up all season, but you mentioned that'd be too time consuming. I was thinking of this as an alternative...
Could you look at YPP, YPA, YPC, and Run% for 2 wides, 3 wides, and 4 wides for the season? I'm thinking you might be able to pull that from the UFR data. I know someone looked at YPP from all the different formations, but in a sense I think that's a bit under-aggregated. This might also help isolate the causal variable as Dileo vs the formation.
Also, have you thought about changing your name to Misopogon in real life? That would be awesome.
I don't think Yeoman's arguing that experience is the only factor that governs offensive success, just that it plays a small role. Do you think this is wrong? And if so, do you have any multivariate data that lacks suppressors, has only invariable interactions, nonspurious effects, and demonstrably reliable independent and dependent variables to back up this assertion? Obviously if there are specific errors or biases those should be corrected, but just calling it useless seems a bit harsh and unproductive.
Interesting stuff. I like the comparison between the deviation from baseline efficiency and experience/inexperience. I'm surprised that cuts the r-squared down so much. Also, I'm glad your basic look at experience and YPC produced the same results! That could have been embarrassing...
Steve Sharik brought this up as well in sort of a different way. His ideas concerned controlling for when an underclassman beats out an upperclassman for a starting spot. Should that still be considered "youth" if your awesome freshman is beating out red shirt juniors and seniors?
It's a really important issue for this type of study, and it's something I'm working on for a revised analysis that I'll try to finish up in the next couple weeks. Thanks for the feedback; I think you're spot on.
Also, yeah man, let's get some databases on this info! It would be awesome just to have some MGoDatabases on the site. I realize why you might not want to put them up there since a ton of work can go into putting them together, and you don't necessarily want to just have someone else go do "your" project or analysis after you've spent all the time collecting the data, but it would be really helpful in facilitating user work. After I'm done doing my little series I'll see if I can get the O-Line experience data up somehow.
Yeah, this is the main takeaway, IME. The data show that youth does matter (at least in a statistical sense), but that it only can account for about 5-10% of the variation in YPC. However, a lot of how us mgopeople have characterized our run game has been based on the last 2 games, which were indeed absymal, rather than on the whole season, which has just been bad. Prior to the Nebraska game (haven't updated the stats yet), our YPC was 3.7, which had us ranked in the 90s (~ 25 percentile among FBS teams). Actually I guess this all just depends how you define bad and absymal.
I haven't redone all the stuff to just look at RB YPC, but as long as there is no systematic bias in the difference between total YPC and RB YPC, you should be able to assume the trend line stays about the same. So you can just compare a RB's YPC to the experience trend line to see whether we are over or underperforming with regard to experience (all else being equal).
Fitz's YPC is still 3.5 (how, I don't know), so Michigan is probably underperforming at that position a little more than the current graphs indicate.
Actually it looks like it does. A brief crunch of the data shows that when the center is the youngest on the interior line (n=23), teams average 4.1 YPC. When a guard is the youngest on the interior (n=77), teams average 4.4 YPC. Oddly, when both the center and a guard are the youngest (i.e., the same amount of experience) (n=25), teams average 4.7 YPC.
This is supported when you look at how well each position serves as a predictor of running success. The center spot has a correlation coefficient of 0.20, an r-squared of 0.04, and p-value of 0.02. When you use the youngest guard as a predictor, you get a coefficient of 0.14, an r-squared of 0.02, and a p-value of 0.11.
Probably could look at this a bit more, but preliminary returns show that yeah, center appears to be the most important.
That's a good point, and I think it's the type of info I should be able to get without too much work. What measure of experience do you think would work best for each position along the o-line? Total or average of the two-deep experience at each position? Or do you think the best measure might just be the most experienced person on the 2-deep at each position?
For example, let's say we've got Bosch (true freshman = 1.0) as the starter at LG and Bryant (RS sophomore = 2.5) as the backup. Do you think a total (3.5) or average (1.75) (these are essentially the same measure) would would work better than either taking the experience of just the starter (Bosch = 1.0) or just the most experienced person (Bryant = 2.5)?
The more I'm thinking about it, the more I'm thinking the most experienced person on the two-deep should serve as the experience measure for that position. You don't want lower Michigan's experience metric at LT or RT just because they've got young backups, which would happen if you take a total or average. But you're right that younger guys should get some type of credit for beating out a more experienced teammate.
Thanks, Yeoman. I look forward to reading your work, especially what you have to say on the bust percentages. Were you able to come up with a way to quantitatively determine those for each level of experience? Or is it more of an estimation? Either way, sounds interesting. I'm sure your data on experience is better than mine if you've pulled it from team websites. I was kinda going for the path of least resistance there. I might hit you up for that to use for part two of this study.
Getting some time depth on this study would be worthwhile, and it's something I might look at in the future. You're right, of course, that Michigan had a very experienced line last year, and that the RB production was pretty grim. This is why I tried to be pretty guarded with my conclusions regarding the indictment/absolvement of the coaching staff. Even though we fall pretty close to the trend line in terms of YPC once experience is accounted for, I still think there must be problems on the coaching side of things. It's a big data set, and there are definitely experienced teams that suck at running and young teams that are quite good at it, and this variance could be due to any number of things (e.g., RB skill, scheme, O-Line talent) in addition to player development. If we looked at the data over the course of the Hoke/Borges/Funk tenure, we'd have a better idea of whether we're consistently underperforming given our experience.
You may want to include overall defensive efficiency rank (something like defensive FEI) in order to test whether we tend to over/under perform against good/bad defenses. For example, is putting up 751 yards against Indiana (overperforming against a bad defense) and putting up -48 rushing yards against MSU (underperforming against a good defense) a trend? Or is it anomalous? Could be interesting to look at over a multi-year span.
I definitely wanted changes to be made after the Akron and UConn games. Now I imagine most of us wish we hadn't messed with it. Hindsight's 20/20, but it sure seems like the gains in terms of cohesion along an unchanging line would have outweighed whatever benefits (hmmm) we've gained from mixing it up. Here's hoping this widespread experience pays off in future years...
Yeah, this is essentially my line of thinking as well. A young line suggests you're putting up like 3.5 - 4 YPC instead of 4.5-5 YPC, all else being equal. It doesn't mean you're stringing together multiple negative rushing yardage games.
Somebody suggested this at least semi-seriously in another thread. At the time I thought it was ridiculous, but the data here suggest it might not have been crazy. Obviously you can't do it with 3 games to go, but it's at least an interesting thought experiment to image Braden and Mags on the outside with Lewan-Glasgow-Schofield on the inside.
You'd have to wonder how much this would cripple the passing game. After all of MSU's stunts and double A-gap blitzes, I wonder if it's less problematic than we'd assume.
Thanks, Ron! Glad to hear you enjoyed it, even if it doesn't make you feel much better about our eleventy first ranking in YPC. And the data I used were pre-Nebraska, so the current picture is even less rosey.
Thanks for the feedback, Gob. As I mentioned, I'm a humanities guy, so I'm just trying my best with limited stats knowledge. It seems like most of your critiques come down to the fact that there are a lot of other factors that are influential in determining a teams ypc (e.g., schedule, talent, coaching, etc.), and, of course, you're totally right. Even if we take the results as significant, it only goes a little bit of the way towards explaining YPC.
My main hope was that the large sample size (at least relatively large compared to just looking at the Big 10) would help alleviate some of these concerns. We'll get young teams that have played tough schedules and young teams that have played easy schedules, experienced teams with talented linemen and experienced teams with low-profile linemen. Basically operating under the assumption that these things would balance themselves out, and so if the trend holds up with a low p-value we can actually put some faith in it.
I'm planning on looking at pass protection next, but after that I might play around with doing some multi-variate analysis to see if we can increase that r-squared value. Also, I like the idea of looking to see if non-linear models might better fit the data. Thanks again for the input.
Thanks akim. Yeah, I focused more on the "Michigan is pretty close to the trend line" aspect in the interpretation, but you're right, no matter how you slice it, we're falling a bit short even when experience is accounted for.
Recent Comments
Yeah, I figured it was taken into account. And you're right that I was thinking just of Coaches / AP rankings not KenPom. But still, man. Can you imagine a top five football team ever being a dog to an unranked team, even on the road? Just surprised about either (a) how much of an effect homecourt has in bball or (b) how much more parity there is across college bball.
Does anyone else think it's nuts that an unranked team can be favored over the #1 team in the country in Kenpom rankings? Not nuts like it's incorrect, but nuts that this actually seems semi-reasonable? Homecourt advantage is ROUGH in college bball.
DNak attack!
This is really helpful. Depressing, but informative. I was hoping that the downward slope would be consistent across all the BIG teams. I shudder to think of what your line would look like if we didn't go ham in the last game.
IME this is the issue to consider when debating the future of Gorgeous Al. I'm sure there are mitigating factors. The experience stuff I did suggests that youth probably plays a significant role, especially with our young interior line. Personnel is probably a big part of it too.
My main point: These mitigating factors influence season averages of things like YPP, YPC, YPA, and sack percentage when compared to other BIG teams. They shouldn't, however, play a role in determining the slope of the intraseason progress lines you're producing. If anything, we should expect to improve more than other programs over the course of a year because of our youth, since younger players can be expected to improve more than older players.
If I were Davey B., this is the thing I'd be looking at.* I don't care that our YPC this year was 3.07 and the BIG average was 4.5 (actually no clue). All I want to see is improvement. It's not like we're starting from an elite point where the only place to go is down. Unfortunately, I just don't see it with Alan "Al" Borges. I hope I'm wrong.
* I'm not Davey B., and even as Gandalf I have no control over the affairs of the athletic department. So this really doesn't matter. It's just the 16,000th opinion on the subject.
I'm surprised that "yards per attempt against" and "completion percentage against" isn't a standard metric for DBs. I agree that INTs and PBUs are pretty meaningless. It doesn't seem like it'd take much extra work to calculate these stats. It's like judging a QB just on the number of TDs he throws; it tells you something, but only a fraction of what some slightly more advanced stats tell you.
It's really helpful to have all the one-score games compiled like this. If anything, I think this suggests that our record pretty accurately reflects the way we've been playing. As a guesstimate, I'm assuming we shoud win about 60% of one-score games due to usually having a slight talent advantage (at least based on recruiting star levels).
If we saw Hoke or RR was something like 11-1 or 2-10 in one score games, I'd say that we were particularly lucky or unlucky, and that maybe our overall record doesn't really reflect how well we were playing.
But Hoke is 9-7, and RR was 7-7 in the one-score games compiled here. This seems to be just about right. It's easy to just look at the ones that didn't go our way and think we were a couple plays away from greatness, but this year we were equally close to 2008 levels. Nice to put some numbers behind those gut feelings though. Thanks, Brhino.
From what I've read, plasmas are great for darker areas (e.g., a basement), but you'd probably want to go LED or LCD if it's in a room that gets a lot of light. I'm sure some people have strong feelings about it, but IME I think they all have pretty sweet pictures.
Yeah that's a good point. Talent levels would probably look more like RR's tenure. Not bad, and better than Indiana, but not where they are now. You're right that some of the attraction of Michigan is the pro-style offense.
No problem, glad you enjoyed it! I agree that it'd be useful to look at RB experience/talent, especially with the way our RBs seem to get blown up running and passing the ball.
Fitz gets destroyed in pass protection and he's a senior. But so do the young guys and they're, well, young. I guess that's why the large FBS-wide analyses are useful, since they can help point a general finger towards either youth, talent, or bad coaching being the problem.
Any idea if Indiana's OC is the guy responsible for their up tempo offense? Or is it Kevin Wilson? Either way it looks AWESOME, and if it is their OC, surely we could poach him. Running that kind of offense with our talent (star rating-wise) would be interesting to see.
Yeah, I agree that salary's a pretty poor proxy. Tough to measure "talent" though without it being tautological. I was just trying to think of ways to to explain teams' best OLs playing LT, and I thought one reason might be because that is what preps them to become the best pros (i.e., LT is the most important position in the pros).
I was thinking of the Lawrence Taylor thing too, but any idea if there has ever been a quantitative study on where most QB injuries come from? I was really suprised that sacks tend to correlate the best with the interior of the line, so I wonder if a few anecdotal weak side hits have somewhat entrenched this as a fact. No idea how you'd prove it without some serious work, but it's interesting to think about.
I'm trying to think through this, and I'm still not sure I quite have my head wrapped around it. I mean I get that survivorship exists. But my initial reaction though is that this would actually decrease the r-squared that we get when we look at experience and success, meaning that there are probably a good number of freshman and sophomores in there because they're good and not because they have to be. If they're good they should help produce more YPC or fewer sacks, and if younger people are producing fewer sacks or more YPC, then that should lessen the correlation that we see.
Interesting analogy to the Bill James stuff. I think that's essentially what we see with a lot of the freshmen who are starting. Reflecting on this, I think that perhaps the best metric for experience along the o-line might be the oldest person on the 2-deep at a given position. So if Bosch is backed up by Bryant, the position gets a 2.5 (RS Soph) instead of 1.0 (FR).
Yeah, all else being equal, I think this would say would should improve on the line despite losing some good tackles. There are enough Michigan-specific factors that it's really impossible to say with any certainty, but I'd be the experience on the interior helps us more than losing the tackles hurts us. One can hope, right?
I think I get what you're saying. Basically that older o-linemen are better not because they are older, but because they are just more talented or lucky (and thus fend of younger guys). I guess it's possible that this influences the results to some extent. But shouldn't that survivorship bias show up at the tackle position as well as the interior?
Also is there evidence showing that mutual fund managers don't get better with experience? It seems logical to me that OL would get better with experience, and so if mutal fund managers don't and OL do, then I'm not sure the analogy applies.
And it's certainly possible that outliers drive some of the trends. I would tend to doubt it with the fairly large sample sizes. But it's possible.
Honestly, I've beaten the dead horse about as much as I care to. But I'd be interested in seeing the results if you carry out your proposed study. Not sure how you would measure "how good" an OL was after year 1 or 2 and then "how good" they are now.
Nice work, Mr. Utah. Have you thought about trying to visualize the data in a graph of some sort? Might help your argument come across even stronger. Thanks for the expanded look at youth though. I know it's kinda beating a dead horse, but it's also nice to have some reason for optimism.
This is really interesting! Thanks to everyone who's done the comparative / contextual work as well. That stuff really helps. My gut reaction is that things aren't that bad...except for the last four games where things have COMPLETELY FALLEN APART. With cupcakes excised we've still got more above the break even line.
Is it encouraging that we're still generally above the line? Or discouraging that the second of this season is a tire fire? I don't know. Both?
Does anyone else just wonder why we couldn't have RichRod but with a Mattison DC? Beating a dead horse, I know, but man. I mean there's a decent change RR and BH end up with the same record in year 3, except RR was trending up after getting his own players and BH is trending down after doing the same. Why can't we have nice things?
Interesting AND depressing! Nice combo. A couple thoughts...
This is the best thing about rooting for the bball team. I mean we lost, and we're probably the better team, and it sucks. But Beilein's history shows that over the time the coaches will adjust, the players will develop, and the team will improve.
What do you think?
The very first Upon Further Review. (Offense vs. NIU 2005)
Most recent offensive UFR (vs Northwestern)
Robohenne indeed.
Hey Brian from 2006. Somebody linked to this in one of the threads and I noticed it didn't have any comments! Great post and great game. I really liked what the defense was able to accomplish. Tell them to look out for OSU at the end of the year though, that one looks like it might be a shootout.
Do not take me for some conjurer of cheap tricks!
The story of my MGoPoints began long ago. Back in the haloscan days, to be exact, when a one sentence update was all you would get about a 5* DE commitment. Ah, to be young again! Of course I wasn't Gandalf the Maize back then, merely Gandalf the Blue. I wandered far and wide in those days, encountering things more horrifying than one dares to imagine in the current age: the Horror, Armageddon, GERG. It was only when I was able to sublimate these terrors to enhance their antipodal highs - 100th game, UTL I, Denard - that the transformation was complete. But I digress...
Actually, our rushing woes led me put together a study on how o-line experience correlated with run game success. I wanted to post it as a diary but had no points. So I sent a raven to Brian, and he gifted me my first hundred so I could post. Misopogon ended up selecting me as the diarist of the week, and in addition to solidifying the new greatest moment in my life, I think he gave me like 200 points as a reward.
The rest of the story can be found in the Silmarillion, but perhaps those tales are best left for another day...
+1 insightful
Thanks Mr. Carson! The winged hat helps people identify me when they look to my coming on the first light of the fifth day.
This is good shizz! Really interesting study; it's nice to put Kalis' development in perspective. Even as a five star you're more likely to never start a game than you are to become an All-American. Probably wouldn't have guessed that. Thanks for taking the time to do this.
Lindley Stats
I think I would give Borges a positive here for QB development considering the increased YPA, despite the fact that completion percentage stays about the same and interception percentage increases slight from pre-Borgesian times. Brian did mention he had a full time QB coach at SDSU, so make of that what you will. Thanks for pointing this out.
The question of whether Gorgeous Al has a track record of developing quarterbacks is a really important one. My first thought is that it should receive a bit more treatment in a post of its own, but with all that bouncing around it sounds like it's difficult to say too much.
The guy's been a D1 OC for 20 years if you include Boise State in the early 1990s (which was actually D1-AA at the time), and we have all of 3 data points for whether his quarterbacks improve. McNown = yes, Cox = no, Paus = maybe. If you include Denard and Devin then that's probably 2 more in the "no" pile.
My 2 cents?
Cent 1 - It's not the smartest thing to hire someone who has bounced around so much because it's really difficult to tell whether they're (a) good at long term development, (b) good when taking people by surprise (i.e., before the rest of the conference gets wise to them), (c) come into good situations, or (d) just aren't very good. I realize this is hard to do in big time college football, because programs constantly take fliers on people who have shown success after two or three years so there aren't many people around at lower level schools who have demonstrated long terms success.
Cent 2 - I'd bring in a dedicated QB coach with a track record of QB development even if it's at lower levels. No clue about the number of assistant coaches you can have, but this seems like a way to potentially improve the situation in future seasons without the risk of mixing things up too much (e.g., if you fired the OC).
I hope you're referring to changing your name. But either way I think the results should prove interesting.
I'm with you in "we should be spreading the field out because we can't block" department. My initial thought was wondering whether this trend held up all season, but you mentioned that'd be too time consuming. I was thinking of this as an alternative...
Could you look at YPP, YPA, YPC, and Run% for 2 wides, 3 wides, and 4 wides for the season? I'm thinking you might be able to pull that from the UFR data. I know someone looked at YPP from all the different formations, but in a sense I think that's a bit under-aggregated. This might also help isolate the causal variable as Dileo vs the formation.
Also, have you thought about changing your name to Misopogon in real life? That would be awesome.
-1 unfortunate
+1 humorous
-1 unfortunate
I don't think Yeoman's arguing that experience is the only factor that governs offensive success, just that it plays a small role. Do you think this is wrong? And if so, do you have any multivariate data that lacks suppressors, has only invariable interactions, nonspurious effects, and demonstrably reliable independent and dependent variables to back up this assertion? Obviously if there are specific errors or biases those should be corrected, but just calling it useless seems a bit harsh and unproductive.
Interesting stuff. I like the comparison between the deviation from baseline efficiency and experience/inexperience. I'm surprised that cuts the r-squared down so much. Also, I'm glad your basic look at experience and YPC produced the same results! That could have been embarrassing...
Or fumbles that are returned for touchdowns.
Steve Sharik brought this up as well in sort of a different way. His ideas concerned controlling for when an underclassman beats out an upperclassman for a starting spot. Should that still be considered "youth" if your awesome freshman is beating out red shirt juniors and seniors?
It's a really important issue for this type of study, and it's something I'm working on for a revised analysis that I'll try to finish up in the next couple weeks. Thanks for the feedback; I think you're spot on.
Also, yeah man, let's get some databases on this info! It would be awesome just to have some MGoDatabases on the site. I realize why you might not want to put them up there since a ton of work can go into putting them together, and you don't necessarily want to just have someone else go do "your" project or analysis after you've spent all the time collecting the data, but it would be really helpful in facilitating user work. After I'm done doing my little series I'll see if I can get the O-Line experience data up somehow.
Yeah, this is the main takeaway, IME. The data show that youth does matter (at least in a statistical sense), but that it only can account for about 5-10% of the variation in YPC. However, a lot of how us mgopeople have characterized our run game has been based on the last 2 games, which were indeed absymal, rather than on the whole season, which has just been bad. Prior to the Nebraska game (haven't updated the stats yet), our YPC was 3.7, which had us ranked in the 90s (~ 25 percentile among FBS teams). Actually I guess this all just depends how you define bad and absymal.
I haven't redone all the stuff to just look at RB YPC, but as long as there is no systematic bias in the difference between total YPC and RB YPC, you should be able to assume the trend line stays about the same. So you can just compare a RB's YPC to the experience trend line to see whether we are over or underperforming with regard to experience (all else being equal).
Fitz's YPC is still 3.5 (how, I don't know), so Michigan is probably underperforming at that position a little more than the current graphs indicate.
That sounds good. I'll take this into account when I revise this in the next couple weeks. Thanks for the feedback.
Actually it looks like it does. A brief crunch of the data shows that when the center is the youngest on the interior line (n=23), teams average 4.1 YPC. When a guard is the youngest on the interior (n=77), teams average 4.4 YPC. Oddly, when both the center and a guard are the youngest (i.e., the same amount of experience) (n=25), teams average 4.7 YPC.
This is supported when you look at how well each position serves as a predictor of running success. The center spot has a correlation coefficient of 0.20, an r-squared of 0.04, and p-value of 0.02. When you use the youngest guard as a predictor, you get a coefficient of 0.14, an r-squared of 0.02, and a p-value of 0.11.
Probably could look at this a bit more, but preliminary returns show that yeah, center appears to be the most important.
That's a good point, and I think it's the type of info I should be able to get without too much work. What measure of experience do you think would work best for each position along the o-line? Total or average of the two-deep experience at each position? Or do you think the best measure might just be the most experienced person on the 2-deep at each position?
For example, let's say we've got Bosch (true freshman = 1.0) as the starter at LG and Bryant (RS sophomore = 2.5) as the backup. Do you think a total (3.5) or average (1.75) (these are essentially the same measure) would would work better than either taking the experience of just the starter (Bosch = 1.0) or just the most experienced person (Bryant = 2.5)?
The more I'm thinking about it, the more I'm thinking the most experienced person on the two-deep should serve as the experience measure for that position. You don't want lower Michigan's experience metric at LT or RT just because they've got young backups, which would happen if you take a total or average. But you're right that younger guys should get some type of credit for beating out a more experienced teammate.
Thanks, Yeoman. I look forward to reading your work, especially what you have to say on the bust percentages. Were you able to come up with a way to quantitatively determine those for each level of experience? Or is it more of an estimation? Either way, sounds interesting. I'm sure your data on experience is better than mine if you've pulled it from team websites. I was kinda going for the path of least resistance there. I might hit you up for that to use for part two of this study.
Getting some time depth on this study would be worthwhile, and it's something I might look at in the future. You're right, of course, that Michigan had a very experienced line last year, and that the RB production was pretty grim. This is why I tried to be pretty guarded with my conclusions regarding the indictment/absolvement of the coaching staff. Even though we fall pretty close to the trend line in terms of YPC once experience is accounted for, I still think there must be problems on the coaching side of things. It's a big data set, and there are definitely experienced teams that suck at running and young teams that are quite good at it, and this variance could be due to any number of things (e.g., RB skill, scheme, O-Line talent) in addition to player development. If we looked at the data over the course of the Hoke/Borges/Funk tenure, we'd have a better idea of whether we're consistently underperforming given our experience.
You may want to include overall defensive efficiency rank (something like defensive FEI) in order to test whether we tend to over/under perform against good/bad defenses. For example, is putting up 751 yards against Indiana (overperforming against a bad defense) and putting up -48 rushing yards against MSU (underperforming against a good defense) a trend? Or is it anomalous? Could be interesting to look at over a multi-year span.
I definitely wanted changes to be made after the Akron and UConn games. Now I imagine most of us wish we hadn't messed with it. Hindsight's 20/20, but it sure seems like the gains in terms of cohesion along an unchanging line would have outweighed whatever benefits (hmmm) we've gained from mixing it up. Here's hoping this widespread experience pays off in future years...
Yeah, this is essentially my line of thinking as well. A young line suggests you're putting up like 3.5 - 4 YPC instead of 4.5-5 YPC, all else being equal. It doesn't mean you're stringing together multiple negative rushing yardage games.
Somebody suggested this at least semi-seriously in another thread. At the time I thought it was ridiculous, but the data here suggest it might not have been crazy. Obviously you can't do it with 3 games to go, but it's at least an interesting thought experiment to image Braden and Mags on the outside with Lewan-Glasgow-Schofield on the inside.
You'd have to wonder how much this would cripple the passing game. After all of MSU's stunts and double A-gap blitzes, I wonder if it's less problematic than we'd assume.
Thanks, Ron! Glad to hear you enjoyed it, even if it doesn't make you feel much better about our eleventy first ranking in YPC. And the data I used were pre-Nebraska, so the current picture is even less rosey.
Thanks for the feedback, Gob. As I mentioned, I'm a humanities guy, so I'm just trying my best with limited stats knowledge. It seems like most of your critiques come down to the fact that there are a lot of other factors that are influential in determining a teams ypc (e.g., schedule, talent, coaching, etc.), and, of course, you're totally right. Even if we take the results as significant, it only goes a little bit of the way towards explaining YPC.
My main hope was that the large sample size (at least relatively large compared to just looking at the Big 10) would help alleviate some of these concerns. We'll get young teams that have played tough schedules and young teams that have played easy schedules, experienced teams with talented linemen and experienced teams with low-profile linemen. Basically operating under the assumption that these things would balance themselves out, and so if the trend holds up with a low p-value we can actually put some faith in it.
I'm planning on looking at pass protection next, but after that I might play around with doing some multi-variate analysis to see if we can increase that r-squared value. Also, I like the idea of looking to see if non-linear models might better fit the data. Thanks again for the input.
Thanks akim. Yeah, I focused more on the "Michigan is pretty close to the trend line" aspect in the interpretation, but you're right, no matter how you slice it, we're falling a bit short even when experience is accounted for.
Drinkin coffee, writing a dissertation, and wearing sweatpants!
Ahhhhh....grad school.