I thought that myself when I read that article that talked about a Data Scientist(tm)
Bill Connelly a few years back created a new uberstat for receivers called RYPR (receiving yards/total team plays *Passing S&P+). His description:
Below, you will find a measure that attempts to answer the following questions about a given pass-catcher:
- 1) How much do you produce?
- 2) How important are you to your team's passing game?
- 3) How good is the passing game to which you are important?
- 4) And how much is the forward pass featured in your team's offense?
The idea was to simply multiply the following four factors together: a player's Yards Per Target, his Target Rate, his team's Passing S&P+, and his team's pass rate. Target Rate x Yards Per Target x Passing S&P+ x Pass Rate = RYPR.
Let's skip right to Michigan
I went through several iterations trying to match exactly what Connelly had done, so this may not match the results I reported beforehand. Here's the breakdown of Michigan's targets with NCAA averages in parentheses:
|Target #||Receiver||RYPR (NCAA avg)||Yards/Tgt||Target Rate|
|1||Jeremy Gallon||178.0 (102.6)||10 (8.6)||32% (23%)|
|2||Devin Funchess||97.0 (68.8)||8.1 (8.1)||21% (17%)|
|3||Drew Dileo||22.6 (49.5)||5.8 (7.8)||7% (12%)|
|4||Jake Butt||30.5 (36.6)||8.7 (7.7)||6% (9%)|
|5||Jehu Chesson||28.7 (27.0)||9.2 (7.2)||6% (7%)|
|6||Fitz Toussaint||26.3 (20.9)||10.2 (6.7)||5% (6%)|
|7||Jeremy Jackson||9.2 (17.1)||7.1 (7.0)||2% (5%)|
Funchess's 97.0 was 22nd among teams' second targets though his yards per target were average for No. 2 guys. The max protect stuff in the season's third quarter (Indiana through Nebraska) bore out in the numbers, with that third receiver (Dileo) far under the typical third receiver's usage.
Best Receivers in a Bad B1G
Gallon wasn't the only long term top receiver to graduate last year, and the conference wasn't very deep on receiving talent to begin with. The result is not many wideouts left in-conference to have cracked that 100 (average for a team's best receiver) mark. In 2013 the Big Ten average RYPR for its teams' top three receivers was 69.5, last among major conferences and just ahead of the Sun Belt and Conf USA. When I removed all seniors the Big Ten was behind the MAC (Conference USA was still very worse). Here's the best among those that remain.
|Receivers||Team||Tgt||Rec||Yds||YPT||Tgt Rt||1st Dn%||RYPR|
|Devin Smith||Ohio State||#2||44||660||9.0||20.9%||34%||88.7|
Maryland has lots of receivers. Northwestern's Joneses were pretty productive, and could be more so in a Trevor Siemian offense. The tight ends:
|Tight Ends||Team||Tgt||Rec||Yds||YPT||Tgt Rt||1st Dn%||RYPR|
|Jeff Heuerman||Ohio State||#4||26||466||12.9||10.3%||44%||62.6|
|Jesse James||Penn State||#3||25||333||8.3||10.2%||33%||39.0|
Remind me to draft Heuerman in the draft-o-snark.
No Cam you don't get points for setting up Morgan's one-timer.
Still playing with the big spreadsheet of stats. Sometimes I glom onto something interesting and sometimes, like today, I waste a lot of time to realize a stat they track has no bearing on play at all, and then I have to write my article, and then Comcast manages to make me wish Greg Robinson was my internet provider and, well, that's my excuse.
While Ace was writing the MSU preview for this year's HTTV (you are welcome to pester Brian to start the kickstarter) I was feeding him various kill-me-now defensive stats that showed State was really good at defense last year. One thing we pulled up was a larger percentage of tackles that were assisted, something MSU seemed to share with other teams.
This does make sense if you think of plays that are good for a defense, e.g. a lot of bodies going nowhere at the point of attack, versus how long gains tend to end. Likewise you'd expect the position of the player to make a difference just because of the variance in amount of space between him and the next defender. A typical distribution of tackles was as follows:
|Position Group||% of Total||% Solo|
Noise in the data: I built this from complete game stats, not play-by-play, so I couldn't separate special teams plays, etc. I did re-categorize a bunch of players listed at incorrect positions but I couldn't catch all of them. Tweener positions also throw things off: a WDE to a 4-3 under team is an outside linebacker to a 3-4 squad, 3-3-5 teams call the Spur a safety, Jake Ryan puts his hand down in the nickel, etc. There's tens of thousands of tackles in the above percentages but as we get into teams keep these inconsistencies in mind. FCS teams and stats accumulated against them were removed.
Who's doing the tackling? So in the above table defensive linemen have marginally more assisted tackles than linebackers, and both have significantly more tackles assisted than defensive backs. If tackle assists mean anything other than "more forward players are doing the tackling" we can see that by testing whether the % of tackles accrued by the front 7 or % of tackles assisted have a closer relationship to tempo-free defensive efficiency.
So yeah, it's where the tackle takes place, not some mystical ability of great defenses to get more people to arrive at the ball at the same time. And neither is that strong of a correlation. Sorry, every platitudinal defensive coach ever.
So how'd we do?
The Big Ten ranked by fewest yards ceded per play:
|Team||% by DL/LBs||Rk||% Solo||Rk||Def YPP|
That Illinois and Michigan State are the top two teams at getting assists on their tackles says tackle assists aren't a thing. Rutgers was great at getting linebackers to the ball, but not until lots of yards had been accrued. Northwestern's a good study in this: in 2012 they had safety Ibraheim Campbell racking up Kovacsian solo tackle numbers, but in 2013 they had greater contributions from up front…with little increase in productivity.
I don't even see much in the way of stylistic preferences coming through. Michigan and Nebraska and Ohio State I believe (gleaned from what their coaches say at clinics mostly) are "spill" teams—they try to occupy blockers so a free hitter can make his way to the ball. Michigan State and Wisconsin and Penn State, are the ones I believe were "gap" teams—every defender has a gap he's responsible for closing.
So…okay, this stat means nothing. Good to know I guess.
read option [Fuller]
I am determined this spring to mine every possible stat for every possible insight. This week I delved into quarterback rushes. Not sacks. I wanted to know which offenses tended to have their quarterbacks take off, or planned runs for them into their game plans.
Baseline: here's Michigan and their opponents last year. Sacks and yardage lost to them are not counted, but I couldn't tell from scrambles and QB sneaks, or stuff like if he took off for 10 yards on 3rd and 15 that defenses are happy to give up:
|Season Avg||vs Mich|
|Opponent||QB Rush||Yards||QB Rush||Yards|
Indiana, Nebraska, Northwestern, Ohio State, and Kansas State ran option games. Minnesota's offense was QB power running (thing it is like: Michigan's 2010 offense when Rodriguez gave up on trying to make Denard into a zone reader). According to the UFR database Minnesota quarterback running plays vs Michigan were as follows: 7 QB powers; 2 draws; 2 zone read keepers; a false zone arc sweep thing, a QB sneak, and 7 scrambles.
The stats can't tell the difference between this kind of offense and a dedicated Richrodigan spread 'n shred. There aren't many teams who run this as their base offense, as simple as it may be, but a lot of teams have a mobile change-of-pace quarterback and a small package built around him. Notable teams who deployed a second guy:
|Player (2014 Elig)||Team||% of Snaps||% Will Pass||Rush||Pass|
|Austin Boucher (graduated)||Miami(NTM)||51%||73%||80||211|
|Austin Gearing (So.)||35%||35%||129||70|
|Drew Kummer (Jr.)||14%||71%||22||55|
|Nate Sudfeld (Jr.)||Indiana||61%||94%||22||338|
|Tre Roberson (Jr.)||38%||62%||84||139|
|C.J. Brown (11th year Sr.)||Maryland||73%||72%||119||303|
|Caleb Rowe (Jr.)||26%||91%||14||136|
|Philip Nelson (transferred)||Minnesota||59%||72%||79||200|
|Mitch Leidner (So.)||38%||51%||89||91|
|Gary Nova (Sr.)||Rutgers||68%||93%||25||328|
|Chas Dodd (graduated)||32%||87%||21||143|
|Tommy Armstrong (So.)||Nebraska||39%||68%||63||135|
|Ron Kellogg III (graduated)||31%||90%||16||141|
|Taylor Martinez (graduated)||30%||77%||34||116|
|Trevor Siemian (Sr.)||Northwestern||63%||92%||29||315|
|Kain Colter (graduated)||36%||50%||98||99|
|Braxton Miller (Sr.)||Ohio State||72%||65%||150||276|
|Kenny Guiton (graduated)||25%||74%||39||110|
I included Rutgers to show Chas Dodd wasn't a Drew Henson-ian run threat except in comparison to Gary Nova.
[Jump: Okay spread zealots, do teams with running QBs have an advantage?]
Gardner's implied question is the same we're all asking [Fuller]
The 2014 football season hinges on whether the offensive line can go from one of the worst in the country to just mediocre. We've mentioned the downsides: it has to replace two NFL tackles. The upside is an offensive coordinator who plans to simplify the things they'll have to do, a ton of talent, and rather good excuses for why the bulk of guys weren't so good (youth compounded by panicky/insane coaching decisions). The competence of coaches replaced, arriving, or remaining can't be determined until they play, so guesses at their 2014 performance have to be extrapolated from what we know of the current players and the typical progression of men like them.
When Michigan was still putting together those 2012 and 2013 classes I looked over the history of our offensive linemen going back to the mid-'90s, because my memory before that is weak.
|Year in program|
|Not on team||1||6||13||17||29|
|% Solid +||1%||11%||21%||29%||37%|
The results were the growth chart below. I've reproduced it with updated data from 2013:
Really it's more specific than the above. If you're the backup to Steve Hutchinson in 2000 you could be pretty solid or terrible, but if you were an interior lineman on the 2013 team and hale and still couldn't crack the depth chart, you were obviously not good at that point. One thing working in our favor is Michigan has historically brought in offensive line classes rated about as highly as the recent crops. If you tried this with MSU over the same period there would be stretches of 2-stars (and, um, personal issues) to throw off the numbers.
A more precise way to show where our OL are at this point is to find closer comparisons to historic players at this point in their careers. I couldn't figure out a good way to show "tracks" before, but I think I've learned enough about table html now to make a crude flow chart. Sample sizes are way too small to say "Kalis will be X good by Y season," but if you can read it to say "At that age, Steve Schilling and Patrick Omameh were both about where Kalis is now." Usefulness is better at capping expectations: you can always say so-and-so was a backup at this point, but Miller's not going to be Molk.
|Not on team (x)||TransferRS||Backup||Solid||Star||Jonathan Goodwin|
|Solid||Star||Star||Jansen, Hutchinson, Backus, Long, Lewan|
|Star||Star||David Brandt, David Baas|
|Solid||Star||Tony Pape, Adam Kraus, Schofield|
|Liability||Solid||Frazier, Petruziello, Bihl, Ortmann|
|Liability||David Moosman, Perry Dorrestein|
|Backup||Ben Mast, Courtney Morgan|
|Backup||Solid||Kurt Anderson, Leo Henige|
|Backup||N. Parker, Denay, Kolodziej, McAvoy|
|Unrenewed||Partchenko, Potts, Christopfel, Gaston, DeBenedictis, Ciulla, Gallimore, Khoury|
|Injuries||Zirbel, Mossa, Sharrow, Brooks, Schifano, C. Bryant, Tannous, A.Brown, Simelis, Berishaj, C.Pace|
|Transfers||Ries, Moltane, Zuttah, Wermers, O'Neill, Posada|
[Discussion after the jump]
More fun with stats! CFBStats helpfully grabs every play off the NCAA's box scores and turns lines like "Devin Gardner pass complete to Jeremy Gallon for 14 yards" into downloadable data on receiver targeting. Here's where Gardner's passes went last year by down:
|Receiver||Target(%)||1st Dn||2nd Dn||3rd Dn|
|Total passes||395 (n/a)||142||144||105|
|Jeremy Gallon||137 (35%)||43%||28%||34%|
|Devin Funchess||92 (23%)||25%||18%||28%|
|Drew Dileo||30 (8%)||6%||5%||12%|
|Jake Butt||27 (7%)||3%||13%||4%|
|Jehu Chesson||24 (6%)||4%||8%||6%|
|Jeremy Jackson||10 (3%)||3%||3%||1%|
|Joe Reynolds||7 (2%)||2%||3%||-|
|A.J. Williams||2 (1%)||-||1%||-|
|Fitz Toussaint||20 (5%)||4%||8%||3%|
|Other backs||23 (6%)||6%||6%||6%|
There were four passes on 4th down: two that Funchess converted and two that Dileo didn't. For our purposes I'm going to count them with 3rd downs because they're functionally the same (i.e. not converting is a failure). When every preview this year says defenses will be focused on taking away Funchess, you can see why: most every other target from last year is graduated or not immediately available (Butt). The data also show whether each reception ended up in a 1st down:
|Receiver||1st/2nd Dn||Conv%||3rd/4th Dn||Conv%|
I don't know if the conversion rate for 1st and 2nd down will be that valuable except as a measure of team dink-and-dunk-iness. The numbers for conversion downs show tendency and success. Again, nothing surprising here. Gallon and Funchess remained equal targets, with Dileo the only other likely 3rd down destination.
Was it common for teams to be so focused on a few guys? Well those 3rd down targeting numbers are high. Gallon was the recipient of just over a third of Michigan's 3rd/4th down attempts; that's 7th in the nation at go-to-guyness. The rest:
|Receiver||School||Tm Att||Tgts||Conv %|
|Alex Amidon||Boston College||106||43 (41%)||42%|
|Jordan Matthews||Vanderbilt||104||39 (38%)||38%|
|Shaun Joplin||Bowling Green||114||41 (36%)||49%|
|Willie Snead||Ball State||131||47 (36%)||55%|
|Allen Robinson||Penn State||129||46 (36%)||43%|
|Ryan Grant||Tulane||133||46 (35%)||46%|
|Jeremy Gallon||Michigan||109||36 (33%)||42%|
|Ty Montgomery||Stanford||100||33 (33%)||55%|
|Titus Davis||Central Michigan||98||32 (33%)||56%|
|Quincy Enunwa||Nebraska||112||36 (32%)||33%|
Gallon was as important of a chain-mover for Michigan as A-Rob was to Penn State. What's weird is Michigan's 2nd guy was also really high on the list. Funchess (29% of 3rd/4th down targets, 39% conversion rate) also appears on the national leaderboard, at 19th, right behind Jared Abbrederis.
[After the jump: Michigan was the most obvious team in the country, finding Dileo-like objects, target types.]
This hurt. [Fuller]
Longtime readers will know the MGoBlog policy on sacking: sacks and sack yardage should be counted as passing, because they are pass plays, not rushing, as the NCAA and thus everybody else is wont to do. Counting sacks as passing leads to a better understanding of success and where yards come from, and prevents problems like the computer in the NCAA videogames passing every play because the sacks that generates keep making the rushing numbers look progressively more awful.
For the Hail to the Victors preview books (kickstarter coming soon) each year we put these "At-a-Glance" boxes into the opponent previews, complete with offensive and defensive stats that we've adjusted for this. Having done the calculations for that, I thought I'd share them with you.
First, the difference it makes to passing stats:
|Team||Pass Att||Pass Yds||YPA||Rk||Sacks||Sack Yds||YPA||Rk|
By counting sacks as passing Michigan drops from 8.15 yards per attempt (good for the best passing team in the conference last year) to a more realistic 6.85 YPA, dropping them to fourth. Minnesota's passing game dropped from middling to awful, Iowa's climbed from the bottom to the middle.
And the difference to running stats:
|Team||Rushes||Rush Yds||YPC||Rk||Sacks||Sack Yds||YPA||Rk|
Michigan's awful running game is still awful, but it no longer looks like the Scheelhaase option-running game was a disaster. Ohio State's 7.27 YPC isn't just first among the conference; OSU and Wisconsin were the #1 and #2 rushing offenses in the country. Michigan: 115th out of 125 teams.
This isn't perfect since quarterback scrambles still can't be pulled out of rushing stats, but that's not so big of a deal considering a running QB should be contributing to your rushing success.
[Jump for Devin Garder's passing season and profiles of next year's opponents]