really cool stuff
I'VE HAD JUST ABOUT ENOUGH OF YOU SONNY
In a recent forum post I put out a request for diary ideas and Brian requested a deeper look at special teams. So here goes.
As any good Michigan fan knows, punting from the opponents 35 yard line is excruciating to watch. But how painful is it?
As you can see above, net punting holds pretty steady in the 37-38 yard range all the way till around the 40 yard line. At that point it drops a couple yards into 35 yards per punt range. It’s when you get to midfield that the averages really start to tank. Over the next 15-20 yards of the field, the average drops from 35 yards per punt all the way down to 20. This is somewhat obvious as the field shrinks the longer punts turn into touchbacks where they would have been 50 yarders on the other side of the field.
But lets look at it another way. On average, where can you expect to start your defensive drive based on where you are punting from.
It’s a pretty linear relationship all the way to midfield and then it stops. On your side of the field, one more yard on offense takes one away from the opposition if you have to punt. You get to the other side of the field and that relationship disappears. At the 50, the other team will, on average, start their next drive at the 16. If you drive all way inside the opponent’s 35 and decide your punter is still the best call, the 16 becomes the 12. The 15-20 yards of offense only translate into about 4 yards of worse field position for the other team. A punt from the 37 should realistically only be expected to cover a net of 25.
All of this is accounted for in my special teams rankings. Punters evaluations on each punt are measured on gross distance, net distance and then compared to punts from that spot on the field. A punt from your own 20 yards that nets 35 is below average but a punt from the opponents 40 that has a net of 30 is above average.
Last year, Georgia led the nation in my measurements in gross punting, it was worth 8.2 points above average on the season. Zoltan came in 7th (1st in the Big 10) at 4.3 points above average.
Punt returners and coverage teams are evaluated in the same manner. A 50 yard punt that isn’t a touchback or out of bounds (but including fair catches and downed punts) will average 6 yards per return but a 30 yard punt should only expect a 2 yard return. Again, teams are only evaluated against the situation they are given, a 4 yard return on a 50 yard punt is a positive play for the coverage team and a negative one for the return team. A 4 yard return on a 30 yard punt is a negative for the coverage team and a positive for the return team. I don’t currently have a way to split the value between the punter and the coverage, so the coverage is a joint metric.
LSU led the nation last year, gaining 9.7 points more than average on their punt coverage. Michigan came in 28th, 3.7 point above average.
What ultimately matters is the total punting rating, the combination of the punt and the cover. Michigan’s combined value of 8.0 points above average was 3rd nationally behind only Oklahoma and Missouri.
Michigan was somewhat unique in that even when adjusting the coverage rating for distance of punt, there is a negative correlation between punt coverage and gross punting. Possibly meaning that the punters getting the most distance on their punts are doing so by kicking more returnable punts than their peers who aren’t kicking as far as consistently.
Michigan did not fare so well on the returning end of the punt game. Michigan averaged a mere 2.35 yards of return per punt (excluding touchbacks and out of bounds) and was 3.9 points below average on the season. LSU again led the nation with 19.3 points above average for their return team.
Kick - Offs
I approached the kick off in the same manner as the punt, with the obvious exception that almost all kick offs are from a fixed spot on the field.
Unlike punting, kick offs have a correlation between good kick offs and good coverage, even when adjusting coverage for kick length. Good kick off specialists provide a more coverable kick than when weaker kickers get a kick of the same distance.
Michigan had another strong showing out of their kick off teams. Ranking 16th nationally at 6.3 points above average. The coverage wasn’t as good, 1.6 paa and 49th nationally. The 7.9 paa was 25th overall in a category that was dominated by Nebraska. Nebraska’s kickoff team was worth 27.9 paa on the season. 50 of their 74 kickoffs either went for touchbacks or were stopped inside of the 20 yardline.
The Wolverine kick return team was a respectable 36th overall, 2.6 paa. Cincinnati dominated the country at 19.7 paa.
Field goal kickers have never really had a good stat with which to measure them by. So much depends on where you are kicking from. Leigh Tiffin from Alabama garnered All-American honors despite missing 4 extra points and making 24 of his 30 field goals from inside 40 yards. Meanwhile in the same conference, Blair Walsh from Georgia makes a nation leading 12 field goals of 40 yards or longer versus only one miss from the same distance and is perfect on extra points and doesn’t even sniff All-American. Walsh’s performance gave Georgia 21.6 paa where Tiffin providing a respectable but not that close 7.4 paa. So how do you evaluate kickers. The easiest way would be to put up a chart.
A nice straight line, right? Look closely at the attempts and something changes around the 30 yard line. With the 30 as a general benchmark, coaches become more and more reluctant to trot the kicker out from that distance or beyond. With this selection bias, the true field goal percentage of field goals from 47 yards and longer is almost certainly overstated. By only getting attempts from the better kickers, the percentage is artificially high.
So now it’s time for a new chart, right?
Using the assumed misses from coaches foregoing the field goal, the true field goal percent drops. The straight line out to the 25-30 yard line goes south fast as the distance is stretched.
Last year Michigan’s kicking game came in at 44th with 1.6 paa on the season.
In attempting to determine how much coaches were passing up field goals in “no man’s land” I did also produce one more interesting but not necessarily special teams chart. The 4th down decision chart.
Between the 3 and 25 yard lines its a consistent trend, 80-85% field goal attempt 15-20 % go for the touchdown. It rises to 60/40 at the 2 and flips to 20/80 at the 1. The going for it actually peaks between 30 and 35 as more coaches don’t really know what to do so they just go for it.
Final Thoughts and Notes
There are a couple of things not included in this analysis. Exception plays such as blocked punts and kicks and their returns, fumbled returns (not that those ever happen) and the like are all excluded. These play obviously have huge impacts on the game in which they occur but they are so rare and have little or no impact on other plays of that type that they are excluded . The very best teams in the country may block 4-5 kicks in a season and for all but a few teams, these plays have virtually no net effect.
In general, for any one special team unit, the difference between average and the best and worst is about 2 touchdowns in either direction over the course of the season. Being the best at special teams is worth about a half game a season versus the average team and a full game a season versus the worst team. If there is one unit to excel at, the opportunity is on the kickoff team where last year there was a 53 point differential between the best (Nebraska) and the worst (West Viriginia).
really cool stuff
Could you explain what was done to come up with this adjustment. It makes sense to do something there, it's just not clear what was done.
Using the data from the 4th down decision chart, I assumed that the incremental times the team went for it on 4th above the 15-20% that would do it anyway would be missed FGs. This kept the number of FG makes the same but increased the number of attempts, lowering the FG%.
kinda shows how useless it is to punt inside your opp 40
That's not at all what this shows.
As you get closer, the average net starting position doesn't get any better after around the 40 yard line, but you are still putting them on the ~15 yard line. Rather have them there than on the 40.
Really another chart is needed showing the value of the resulting punt as compared to expectancy of points generated by the FG attempt.
In the same way Mathlete can estimate that a sack of -5 yards is worth -1.4 points (or whatever), he can estimate that a 20 yard punt that puts the opponent on the 15 is worth +1.6 points (or whatever). So the question then becomes, if punting is worth 1.6 points but you only have a 50% chance at making the field goal, or 1.5 points (3 x 0.5), then you should punt because that is the better value.
I'd expect that these two lines (value of an average punt vs. field position; expected value of a field goal by field position) would cross at about the 30-35 yard line. In general, coaches aren't stupid; even the conservative ones.
The expected points is The Mathlete's bread and butter and what sets him apart in my mind. I would love to see the expected points factored in, I love that stuff.
I could be wrong, but wouldn't that basically replicate the David Romer paper? He found that most NFL coaches are indeed too conservative.
I think a more accurate way of saying it would be "Most coaches will play it conservatively, even if it seems stupid." To them, if you have 4th and X at the 37, who cares if you only get 17 yards out of the punt? You got something and that is what matters. (This is primarily for job security; most coaches seem to think, and rightly so, that even if they make decisions designed to pay off better, if the decisions do not pay off and are not "book" decisions, they will be blamed for the failure rather than credited for the decision.)
I definitely agree that the idea of having a QB punt inside the 40 is an excellent one, if for no other reason than to keep the defense on their toes. (A fake punt that involves passing is much more effective when thrown by a QB.)
what it's really saying is almost anyone could be the punter at this point, since the yield in yards from the punt is so small.
In other words, once you pass the opponents 40 and you have to punt, you don't need Zoltan, you can use Tate, and thus increase the chances of a punt out of bounds inside the 10 from the surprise factor of the QB punting.
And even farther into crazy suprise land, turn this into a QB option, so that after it's been called about 5 times, and the defense starts expecting the punt, how easy would it be to pick up the 4th down on a run or short pass?
I wonder if Denard was practicing his pooch punt?
what it's really saying is almost anyone could be the punter at this point, since the yield in yards from the punt is so small.
replacement level is such a sweet concept.
Depends on the distance to go in order to convert on 4th down. Sure, you'd rather have the other team get the ball at the 15 than the 40, but you'd also rather keep the ball in the opponent's territory than give it to the other team at the 15. That distance gained on the punt is minor compared to the possibility of keeping the ball if the 4th down yardage is short enough you'll convert frequently.
not only potential success of making the 4th down (certainly greater chance the shorter the yardage), but also the opponents chance of driving for a score from the position you give them with a 4th down failure.
basically back to the comment with my adds of "(if you're goign to give up the ball) it's better to give it at the 15 than at the 40".
so it's the better chance of scoring from the 40 versus the 15 that you have to compare to the odds that you can make 4th down with 2 yards, or even 5 yards to go. And of course you have to take into account the quality of your defense.
I forget which post of Brian's spoke about the Patriots decision to "go for it" on 4th down against the colts, but a huge factor in the probability chart was based on how successful the Colts were at driving for a score in the 4th quarter against anyone's defense.
Which of course also brings in the time clock as a factor. Meaning, in the Patriots case, there weren't going to have anymore chances to score to counter the almost 100% score of the Colts when they got the ball (I know I exaggerate I forget the percentages quantitatively, but in rough qualitative terms it's on target).
In general I love all of this statistical analysis done by Mathlete. But what it doesn't accomplish is the second step in the analytical string, which is to put in the decision matrix for the situation. And that's because trying to factor in all the appropriate situational factors would crush even someone who did this as their only job.
I'll repeat my previous comment. A clear conclusion you can draw just from this analysis is that teams should be attempting pooch punts much more frequently when punting from the opponents 40. Basically anyone should be able to kick the ball25 yards toward one of the sidelines, and the advantage gained by not putting your punt team on, is that the opponent doesn't have their punt return team on the field. Add in that the offensive team can still go for it from their best offensive team alignment.
Considering that Zoltan had the ability to decide to go for it if he thought he had a chance, I think it's highly probable that you could expect your QB to make this kind of heady decision.
I think what it shows is that it's not worth going for it on 4th down from around the 45 until you get into field goal range. If you miss, they have excellent field position. If you make it, they only have about 4-5 yards worse field position than if you had just punted from where you were on 4th down within the 45. This wouldn't be true if the extra ten yards that could be attempted after getting a first down would make a field goal likely, but in that range of the field (30-45 yard line), punting position should not be a factor when deciding whether to go for 4th down.
you are a diary wizard. Great work, man.
I wonder, do you have any insight into the Arkansas H.S. coach that always goes for it on 4th down? I read a few articles about him and he justified this with averages and percentages that I don't exactly recall but that I thought were very interesting.
Great job, again.
How come nobody does this anymore? I used to see punters try and angle a punt out of bounds to avoid a returner growing up, but you never see it these days. Is it because a shanked kick that's aimed out is really, really bad? I would venture a guess that even a badly kicked punt that goes out of bounds can be on average better than a high kick that gets a decent return. When it comes down to it, if the coffin corner was such a good idea people all over the nation would do it - and they don't. Any coaches out there with a good explanation of this?
It's thoughtful and well written articles like this that stop me from even trying to do one of my own.
Keep up the good work.
Any chance of posting a "Best of" so I can catch up? (naive perhaps, but I still ask...)
I'm never going to attempt to make a Diary. For one thing, I refuse to take the time to put together cohesive thoughts. I also guarantee that mine wouldn't reach the level of quality that people like Mathlete put up consistently. I figure anything I think of can be either put in a forum topic, or will be covered by somebody else at some point.
Very good evaluation of the data, a true metrician. It is funny how often after the analysis you look at what it means and it reveals a truism.
"The going for it actually peaks between 30 and 35 as more coaches don’t really know what to do so they just go for it."
I think this is a coaches 'gut' feeling on a given situation. Good stuff man.
Is there no separate data on kickoffs returned from inside the end zone? Your data starts with returns from the goal line, so is it safe to assume that that data point includes all kickoffs returned from the goal line or beyond? I'm guessing yes, since the slope changes so dramatically when comparing the 0-1 and 1-2 datapoints.
The NCAA play by play data classifies everything at the goalline or in the end zone the same way so I have no way of directly pulling it out. With that said, the relationship is pretty linear for the first 5-10 yards so you could pretty safely assume that it would hold into the endzone as well. This would but the break even point of kneel or return at 3-4 yards deep in the endzone.
unlike the NFL, no official yardage is marked in the end zone (thus the record books show 100-yard kickoff, punt, interception returns). I don't know if this is because of the lack of television coverage back in the day (and for some other conferences, lol, even today) thus giving them no good way to review the exact spot where a play started, the crazy end-zone patterns some schools have (that would make it difficult even on TV), or what ...
I'll really be excited for the returning part of our special teams if Dorsey makes it. I can see it now, Darryl Stonum and Demar Dorsey causing absolute havoc on teams with their speed.
you are awesome
"Hold on to the damn ball" is a pretty important aspect of any return game. Our punting/kicking and coverage teams did well, but our returns were difficult to watch. Its nice that we did well when we dont consider dropping the ball, but I just think thats more important than actually gaining yards on the return because last year our defense sucked, so it risked giving up one of the few stops we were likely to get (which was of disproportionate value to Michigan as opposed to other teams with, you know, competent defenses).
I defer to your analysis since you've actually done the proverbial math, but it still seems like when your defense is bad, turnovers (of any kind) are even more damaging than they are on teams with otherwise competent units.
Thank you for the fantastic analysis. Very interesting.
Where do you get the play by play data to conduct this type of research? I hunted around the NCAA site but couldn't find it - I certainly may have missed it.
Plus a lot of work to get all the play by play data organized.
Boom! +1 for empirical research!
I want to know who the hell punts from the 29 yard line.
Freaking our beaks since 2008.
In the last 3 years there have been 13 punts from inside the 30 yard line, Virginia Tech and Temple have both done it twice. Last year Southern Miss was the worst offender, punting from the 29 on 4th and 6. The punt was of course a touchback.
And I'm not being sarcastic, I am too dumb to see the value here that everyone else sees. When I read most of his well researched, well written and well reasoned posts I usually see a lot of data that normally confirms what we know or what we think we know already. I'm sure someone will point out to me where I'm wrong, and I will be happy to hear it as well as take the negs.
Mathlete, do you still have eligibility? I favor using a scholarship to put you on the sidelines during the game.
It's a fair question. My take on what he's doing is showing WHY it is common opinion. The math is to show how that common opinion came to be. For example, everyone knows not to punt inside the 35 yard line. A cursory glance would say "Because you can at most punt 35 yards, and some of the time it will be 15 after a touchback." The analysis of the data confirms it. Football has been around long enough that, yes, what he's doing isn't new science, but a description of why the game is played the way it is.
It takes a lot of time and effort to sift through all the data, prepare an assertion, and prove that assertion. Being able to present the conclusion in a logical, organized format takes skill.
...misopogon interview in which we are treated to:
How, if all, has your scientific approach on ‘Decimated Defense’ changed your perception of the team? And if the actual coaching staff inexplicably asked if you could offer any analysis or advice, what would you say?
The thing about the Decimated Defense -- and I think this is really really important for understanding that series -- is that it doesn't change perceptions one bit. It's all just a great big confirmation of what we already know or suspected from watching recruiting and attrition over the years.
Do you need me to tell you that Michigan recruits better players than MSU? Or that attrition made Michigan's depth chart ludicrously thin in 2009? Or that recruiting a lot of players of high rating will yield a better team? No. It's just an affirmation, collecting all the data that you've already internalized into an argument.
Not that I think such things are useless. On the contrary, I think things that we take as givens are those that are most in need of re-proof. Otherwise, when we debate them, we're just giving our party line, and they're giving their party line. If we're right, we oughtta be the ones with the facts to back it up. The best you can accomplish with the DD series is to use it as a response to every asswipe poster after a Rittenberg column, minus the Summer Glau sig: "Oh, you think that? Well read <link>this</a>."
I think that answers why the research is useful, even if it's just to confirm conventional wisdom.
I think that Mispogon's work was a flash of light to many that follow M football--although we believed that depth and talent had fallen off, I don't think many people knew just how badly. The fact that we had only 59% of all the players recruited during '06-'09 was a stat that was like a bolt of lightning (to me anyway) that really crystalized the problem. But anyway, I hasten to add that I admire the work that the Mathlete does, I was only questioning why so many seemed to find it illuminating. But your point is well taken.
because it's the beginnings of a follow-up to David Romer's paper. all he'd need to do is incorporate all the expected values to spit out a yes/no answer to punt given down, distance, expected remaining possessions, etc. after you get that, you can compare to what coaches in fact do. and hopefully it enlightens people to the beauty of generating expected value tables.
Every time I read his posts I raise my index finger and say that very loud. But not out loud.
Any change you will post the team-by-team rankings? I'd love to see where specific teams I follow ended up.
The "go for it/punt/fg" decision must be impacted by distance to achieve a new set of downs (or score a touchdown). Obviously, anytime coaches go for it on the 2, they only have 2 yards to get. While on the 32, they could need significantly more yardage for a first down. So part of the reason to go for it more often from the 2 or 1 might be that it is, on average, easier to achieve success because the average yardage needed is less. It would be interesting to see how distance to go impacts the decision. For example, between the 20 and 30, how often do coaches go for it with only 1 or 2 yards to go vs. same distance, but goal to go or vs. same part of the field, but longer "to go" distance.
Also, the incidence of going for it (15-20%) was higher than I would have thought. I wonder how much of this is driven by "crunch time" decisions -- ie down 4 points at the opponents 25 with 45 seconds to go, a coach would obviously go for it. Another way to look at the stats for only quarters 1-3, when the "crunch time" factor shouldn't be present.
According to the results, 80% go for it from the 1. Just curious what the percentage is of touchdowns versus attempts, though I imagine it's similar. I couldn't help but think of the 2009 Illinois game. Minor(RAGE) unable to punch in 4 straight times. Painful memory.