That brings me to a different topic: I don't think I'm going to be attempting any full-scale charting project this fall. There are a few reasons for this:
1. It's hard. It takes quite a bit of time charting games, and it takes ME quite a bit of time coordinating charter schedules. And since I don't have a way of paying charters, there is predictable, understandable attrition from the number of people who volunteer at the beginning of the season and the number at the end.
2. Nobody's really using the data. We had some issues pulling last season's charting data together into one giant file for use, but people weren't using the data anyway. The goal was to have a giant data set available for people to use, but nobody really used the 2013 version. That certainly dampens motivation to do it again.
3. Pro Football Focus does it much, much better than we could anyway. They pay charters, they undergo more detailed charting, and they began doing college games last year. It costs an exorbitant amount to use their data, which stinks simply from a "some random blogger can't get curious and check something out" perspective, but it makes sense considering the money they make from teams and other media outlets.
So I'm going small-scale in 2015. Intern Chris Brown will still be charting games, and I'll be chatting with others about doing the same. I'll be crafting a Charting Box Score of sorts to use for the week's bigger games, and perhaps that will kindle interest in a more large-scale effort again next season. If you're interested in charting games, drop me a line -- I certainly don't want to stop you, and we still have a pretty good template for doing so. But the effort will be done on a much smaller scale this time around.
Charting definitely is difficult and time-consuming, and PFF definitely does do it more expansively than we ever could, but there is still a level of information to be gleaned from charting that aids in analysis of the game itself. Among others, SaxonRBR proved this with his awesome Charting the Tide series last year at Roll Bama Roll. (Here's an example.)
I try to provide as many Advanced Box Score breakdowns as I can each week for the same reason I just mentioned: it provides an extra layer of analysis. This fall, with the games we do get charted, I wanted to attempt a similar process with this data, as well -- a Charting Box Score, so to speak. It would be awesome to have this data full-scale, but in the absence of that, we can still review the big games of a given week.
Using last year's College Football Playoff finals as a (relevant) example, I tinkered with what some of the tables could look like in a Charting Box Score. Some of the results are below. This could merge with the Advanced Box Score for some huge game breakdowns each week. (Here's that piece from Ohio State-Oregon.)
I haven't set the template yet, and I'm looking for feedback regarding what is or isn't interesting. Here are some of the things a Charting Box Score could provide.
|Backs-Wide||% of Plays||Yds/Play||% of Plays||Yds/Play|
|0 backs, 4 wide||2.3%||1.0|
|0 backs, 5 wide||4.6%||14.0||7.9%||7.3|
|1 back, 2 wide||3.4%||1.0||5.3%||8.3|
|1 back, 3 wide||69.0%||5.2||36.8%||7.3|
|1 back, 4 wide||14.9%||12.1||42.1%||4.9|
|2 backs, 2 wide||1.3%||2.0|
|2 backs, 3 wide||5.7%||1.2||6.6%||5.2|
|Lined up on Hash||% of Plays||Yds/Play||% of Plays||Yds/Play|
|No Huddle?||% of Plays||Yds/Play|
From this, we learn that Ohio State lined up mostly with one back, three wide, and a tight end or H-Back attached to the line. Oregon, meanwhile, operated mostly from one-back/three-wide and one-back/four-wide. (The charting template asks for things in terms of # of backs and #wide, but obviously this could also be listed in football parlance like "10 personnel" (1 back, 0 tight ends).
We also learn that Ohio State dominated from the right hash (likely meaning when they were running from right to left), and while both teams operated from a no-huddle about the same amount of time, Ohio State did it much, much better.
There is a lot of charting data available regarding passing. One of the most important pieces is the Distance of Pass piece, but we have data about pass rushers, the reasons for a given incompletion or interception, individual targeting data, etc. I've left the individual targeting data out for now, but here are some options.
|Distance of pass||Passing||Comp Rt||Yds/Pass||Passing||Comp Rt||Yds/Pass|
|Behind Line||5-6, 14 yards||83.3%||2.3||3-3, 7 yards||100.0%||2.3|
|0 to 4 yards||1-1, 1 yards||100.0%||1.0||7-9, 56 yards||77.8%||6.2|
|5 to 9 yards||4-6, 38 yards||66.7%||6.3||5-9, 46 yards||55.6%||5.1|
|10 to 19 yards||2-3, 42 yards||66.7%||14.0||6-10, 103 yards||60.0%||10.3|
|20 to 29 yards||3-5, 102 yards||60.0%||20.4||2-3, 51 yards||66.7%||17.0|
|30+ yards||1-2, 45 yards||50.0%||22.5||1-4, 70 yards||25.0%||17.5|
|When Ohio St. passed||When Oregon passed|
|Avg. Pass Rushers||4.3||4.1|
|Passing (no blitz)||11-15, 204 yards, 0 sacks,
13.6 yds. per att.
|22-33, 312 yards, 2 sacks,
8.9 yds. per att.
|Passing (blitz)||5-8, 38 yards, 1 sacks,
4.2 yds. per att.
|2-5, 21 yards, 0 sacks,
4.2 yds. per att.
|Reason for INC/INT||Ohio St.||Oregon|
We see that Oregon had a bit of an advantage on passes near the line of scrimmage -- Ohio State was 10-for-13 for 53 yards on passes thrown within 10 yards of the line (4.1 yards per pass), while Oregon was 15-for-21 for 119 (5.7) -- but while both teams struck it rich on longer passes, Ohio State's big passes were bigger. Passes thrown 10 or more yards beyond the line: Ohio State 6-for-10 for 189 yards (18.9), Oregon 9-for-17 for 224 (13.2). Combined with the fact that Oregon couldn't generate much explosiveness on the ground, and you have a big key to the game.
The extra data provided on rushing reflects mostly on direction and QB intent.
Charted rushes are marked as to the right or left and to the edge, toward the tackle, or up the middle. Obviously that is a matter of interpretation, but ... well, all of this is.
We can dial in on direction, too (and yes, the point I made above about moving from right to left bears fruit here: toward left tackle or left end, Ezekiel Elliott carried 19 times for 146 yards, 7.7 yards per carry), but here's a general look at how each team fared in moving
Neither offense fared very well between the guards, and there wasn't a ton of success to be found out wide. But off tackle, Ohio State was quite a bit better.
(Now that I'm thinking about this more, a second table simply documenting rushes to the left, right, and middle might also be helpful.)
Box score stats are limited in part by intent. We see how much a passing play gained and how many passes were completed, but we don't know anything about the distance of the pass itself. And while we know there were sacks or QB rushes, we don't know how much was designed or why the sacks occurred or whether the QB rush was designed, the product of an option, etc.
It is kind of a catch-all, but in our game charting we do try to document and explain some of the QB moves that occur. Here's one way in which it could be displayed.
|Option - speed option|
|Option - triple option|
|Option - zone read||2-0||4-19||1-1|
|Sack - in pocket||1-(-5)|
|Sack - coverage sack|
|Sack - QB fault||1-(-1)|
We see that Ohio State's QB draws were very effective, and Marcus Mariota had a little bit of success (but not a lot) in keeping the ball on zone reads. We also see that, despite mobility for both quarterbacks, neither were intentionally moved out of the pocket very much.
So anyway, let me know what you think in comments. Is this data worthwhile for further breaking down specific games? What could be added/subtracted/clarified?
And if you're interested in helping to create this data for your team's games, let me know that too.