This whole post is a wonderful example of how some people don't know what types of graphs (or statisitcs in general) to use to display useful information.

I've never had any formal training on that. If you have suggestions on how I can do a better job with graphs, etc. I'm all ears. I taught myself this shit and am going entirely on instinct and the hope the information gets across. I welcome your suggestions.

I thought this was a super post, but if you want suggestions on things you could do differently (just spit balling here).

Total Season Count of Plays (ex. 16 plays of "other" to focus the discussion)

 # of Receivers in Opponent's Formation UM Defense 4 3 2 1 Total Nickel 121 155 14 1 291 4-3 22 34 195 29 280 Okie 20 32 2 0 54 4-4 1 0 6 11 18 4-6 0 0 10 5 15 3-3-5 5 7 1 0 13 Total 169 228 228 46 671 % of Total 25% 34% 34% 7% 100%

% By Formation (Reads: Given the opponent is in a 4 receiver set, we play nickel 72% of the time.  Interesting: note there really is not a significant difference between 3 and 4 receiver sets.)

 # of Receivers in Opponent's Formation UM Defense 4 3 2 1 Total Nickel 72% 68% 6% 2% 43% 4-3 13% 15% 86% 63% 42% Okie 12% 14% 1% 0% 8% 4-4 1% 0% 3% 24% 3% 4-6 0% 0% 4% 11% 2% 3-3-5 3% 3% 0% 0% 2% Total 100% 100% 100% 100% 100%

For Game by Game analysis I would do an XY Scatter of (X = % of Plays Opponent was in a 3 or 4 receiver set, Y = % of those plays Michigan was in Nickel) in order to see if we played anybody significantly differently than our 70% average, or if its completely explained by how many 3 and 4 receiver set plays the opponent ran.

Well, for starters:

There is no reason why this is a line graph. It implies a relevant relationship over time; that somehow Purdue being played right after MSU is relevant, and the # of Recievers/DBs between those two games somehow has relational relevance more significant than Ohio State vs Western Michigan. Each offense Michigan played had no relationship to any other offense. A scatter plot of Recievers vs DBs with each point representing a game would actually present some sort of coherant data. This? Not so much. Also, making conclusions off of the 'average personel' numbers is dubious, at best, since 1.) the data you're averaging is marginally significant to begin with, and (probably more importantly) 2.) taking an average throws the distribution of plays out. e.g. A team that only uses a 4-3 is going to look the same as a team that 50% of the time uses a nickel and 50% of the time runs a 4-4.

Another example is of poor use of your options graphically, is this:

While this graph has the relevant data hidden in it, it's a poor way of showing it. You're trying to make the point that halfway through the season, Gordon's production increased. You're trying to show the rate of change of this graph (number of tackles per game), so show it. While in this actual graph, it is appropriate to show as a line graph, the data you want shouldn't be. Ideally, this would be a histogram of tackles per game. More significantly, you could create a graph showing percentages of tackles per game. This would have the nice effect of controlling for opponents offensive style/tempo, and show who was really contributing the most, tackles-wise.

What you're really missing here, is how effective each defensive set was verses each offensive set. Was the defense lowering YPP when they added a nickelback against 3 WRs? How well did the offense do with 3-Wide against the 4-3 vs the nickel? This is data you have, and it would actually be interesting to see. This post, however, is not.

these are actually pretty good suggestions, and i'll second that line graphs should really only be used when expressing a change over time. another good rule of thumb to keep in mind is to never use pie charts or 3d anything. both almost always sacrifice clarity for flashiness

Only if there some significance to the order. The fact that Iowa is before or after Purdue has no  bearing on how many recievers each team put up against Michigan, or how many DBs Michigan used to counter it.

There are a few instances when pie charts are usable, but they only show one dimension of data. If you want to show percentages and how they change overtime, for example, an area graph can concisely present the information, whereas you would need to produced multiple pie charts.

Seth, I don't know where you get your information but it clearly isn't accurate. Didn't you know THE SPREAD CAN'T WORK IN THE B1G!!! How could we possibly be spread heavy?

I shouldn't hae to do this but just in case.../s

This post and those like it are why I love this site.  Thanks Seth, I am now measurably smarter!

That post was great. Thanks for taking the time to research all that.

I'm definitely still learning the intricacies of football so I spent most of the time staring the defensive formations and finally making sense of the 3 and 5 techs.

Here comes the stupid question: if the ball is placed on the right hash, does the WDE line up on that side along with the Will? Does the same happen if the ball is on the left side? Thanks for your help gang

Defensive lineman and linebackers mostly align based on strength, which is basically how many bodies the offense has on the line to each side of the center. This is a gross over simplification, but better to look at that usually rather field alignment.

Some coaches will shift techniques rather than flipping players if there's a matchup they favor.

The biggest effect that hash position has on a defense is for corners, field and boundary. But, even they might follow a receiver or formation.

There isn't a rule. In general the offense will align its "strong" side (ie the side with more players on it) to the wide side of the field, since that gives them more room to run to the sideline. In the 4-3 Under the SAM is kind of a hybrid DE/OLB -- he's basically the WDE for the other side, and has to be a bit more athletic and coverage oriented because all of the space that's out there.

I'm pretty sure that "Under" shift basically means that the DL is shifted away from the strong side of the formation, the concept being the linebackers are then shifted "Over" to compensate. The WDE's job then is to hold the weakside if the play goes that way, and otherwise pass rush.

Fronts can shift as quickly as a back crosses the formation, so it's not like the "Weakside" defensive end will always be on the weak side. Spread teams value the extra horizontal space to the wide side so much they'll rarely flip the strength of their offense pre-snap to capitalize on whatever difference there is between the WDE and SDE in edge protection. They make you pull the SAM for a nickel unless your SAM can cover the slot receiver wherever he goes, and then they run at the nickelback. For a converse, think of the DeBord zone-left offense, when Michigan would often line up on the left hash with the fullback and tight end to the wide side of the field, then shuffle the FB to the left and run left, the idea being the WDE is constantly having to hold up to double-teams instead of the 5-tech.

it's, it's mesmerizing.

Your description for that chart is;

I pointed out the two extremes on the schedule with boldation: Northwestern used about twice as many receivers in their formations as Iowa did, and Michigan got to spend four times as many plays in the 4-3 because of it.

But you highlighted Minnesota and Iowa.  Not Northwestern and Iowa.  The logic of the second sentence seems flipped.  Iowa had less receivers than Northwestern, yet the sentence says Michigan was in the 4-3 because of "it" and the it for me refers to the Northwestern situation of more receivers.  What I can see from the chart is that for Iowa Michigan was in the 4-3 more than 60% of the time, and for Northwestern was in the 4-3 just a hair under 15% which is ballpark for "being in the 4-3 for Iowa, 4x more than being in the 4-3 for Northwestern.

If I'm correct in reading the chart, then it does support you basic premise that Michigan responds to 3+ receivers (which is also an indication of "spread" formations) with a Nickel package.

I would like to point out that responding to 3+ WR's with a Nickel is not a new development strategically.  Go back and watch the Rose Bowl against WSU and you'll see a Michigan defense that almost never put their Mike LB on the field, and sometimes only had one LB with a four man front (maybe this is Dime?)  That was in response to WSU having an injury to their main RB and playing almost the entire game in 4 WR sets.  Note this has nothing to do with a Run Option QB, which is the main Strategy of the Rich Rodriguez offense. It's more in response to the threat to pass.  I bring this up in counter to the next chart by season where you highlight 2010 and characterize the defensive response as "fearing the run".  Maybe I'm reading too much into this, but of course opposing teams feared the run, because they were facing a Run Option Offense that rarely if ever used a TE.  So of course the differential as measured by WR to DBs is going to be low.  But I'm hypothesizing that if you examined 2010 against the other seasons for "passing down situations" that the defensive response to the Michigan offense is more similar between 2010 and 2011 than different. And that is without throwing in other qualitative points of view like, 2010 Denard was not a QB that demanded tight coverage, nor was Rich Rodriguez a coach who used complex crossing and isolation routes for the WR.  The strategy was pretty much "run Denard and streak the WR's, if the Safety sucks down Roundtree will be open"

Although that did not seem to be the case for MSU, Wisconsin and OSU.  Or really Mississippi State either.  Maybe all those defenses didn't really fear the passing game, rather than they did fear the running game.

That .gif is incredibly informative... once I downloaded it and took it apart piece by piece in GIMP.  Up at the top of the page it flashes by too quickly - which I guess just shows I can't play quarterback.  I like the format, but maybe next time slow down the frame-rate?

One question - I'm guessing this is me not knowing enough - but frame 10 is labeled "Offense:  122" when it seems like it should be 212 (I-form with HB and FB, Y on right end, 2 wide).

Anyway, cool post.

that gif doesn't have a 3-3-5!!

If I had a dollar for every time my photo of the field was used without permission, I'd be a wealthy man. I've never seen someone put another brand on it though (i.e., the mgologo). That seems just a tad bit unprofessional, a little below the level of what I expect to see here. http://www.hillhaus.com/index.php/2006/12/29/michigan_stadium_the_big_house_1

Several schools have used that photo in their media kits (mostly before the stadium expansion). All of them were professional enough to ask. Combine all of those media kits together and the readership is probably less than the daily visitors to this blog. Just emphasizing the fact that you need to follow higher standards. This isn't Ohio, fergodsakes.

When did you get that shot with no one and nothing in the Stadium?

Seth, you questioned why SDSU averaged 4.89 DBs despite Michigan averaging 2.44 WRs against them; this is because SDSU runs a 3-3-5. You might recall that Hoke was replaced as head coach by Rocky Long, who runs the 3-3-5 almost exclusively.