That is right on! I love the visual representation. Thanks for taking the time to do this.
Visualizing Mgoblog: The Hennechart
To begin the process, I decided to look up Hennechart on google. Oddly enough (and perhaps Brian is now aware of this), the name "Hennechart" is an actual surname. Look it up, and you will find a bunch of french dudes named "Maxime" and "Dominique". And yes, I said they were dudes. Apparently, the male line runs strong in the Hennechart family.
Dominique Hennechart: Friend him on facebook, I dare you
But I digress again. The point of this exercise is to show a visualization of the Hennechart, which I now give you in its fully glory:
Tate's Hennechart (So Far): Pretty snazzy?
So here is the idea: I want to be able to look at the Hennechart and instantly see how well a dude is doing. Here, blue things (dead-on and catchable passes) mean "good"; red things (inaccurate or bad reads) are "bad"; the other things are more in-between: yellow or orange-ish for throw aways, batted balls, and pressures. For each game, I scale the bar to 100% (to make the cross-game comparisons easier), but include the number of "attempts" (the sum of all events) available on the right side. One last thing: Brian has been breaking down the numbers with normal vs. screen passes; here we show screen numbers as diagonally-slashed lines of the same color (e.g., in all games except western, there were "Catchable" screens, which show up as the lighter blue color but diagonally-striped instead of solid).
So there it is. It is a rough and crude first attempt, but frankly I like it better than the chart. Perhaps I am a visual person. Or perhaps I just don't like numbers. But actually, I just like pretty colors. One other thing I like: the color blue. It is no coincidence that the more blue this visualization gets, the better it is.
Suggestions? Comments? Critiques?
Also, if you point me to more Hennecharts, I will make them and put them up for comparison. I would have done Threet from last year but the 5 minutes I spent looking for it were about as fruitful as the football season last year.
Of course mgoblog is full of other numbers and charts. If people have thoughts on which ones might be fun to visualize, drop a note here too in the comments. Would be fun to make this a more visual site...
tremendous effort. The Hennecharts always made me feel a bit dizzy...
I *LOVE* this. But I am easily distracted by shiny things and pretty pictures.
Regardless, a thousand salutations to you, good sir!
This needs to be bumped to the main page so that it can be viewed by the mgomasses. Well done coach, you've made Edward R. Tufte proud!
It would be great if figures like this could accompany all the charts on the front page. Really good stuff.
Scott Mitchell: 45 attempts, 25% batted balls.
Insert requisite Liberace joke.
Fantastic idea. You, sir, have too much time on your hand, but use it wisely!
Why should it be considered "too much" if it's being used wisely? Play on, grapher! Stats geeks unite!
This looks fantastic. Are you producing 'em via Excel or similar? If that's the case, may I request a vertically-oriented (games on the x-axis) non-normalized (bar is as tall as attempts in each game) version?
Also, if you're looking for other numbers to run, going back to the inaugural season of Hennecharts -- back when it meant "chart of Henne" -- would likely give us a decent comparison for Tate, if you're interested in that work.
Still, this is awesome and I'll be looking forward to seeing it as the season progresses.
Games on x axis please - progressing left to right, with both normalized attempts and not (both are interesting.)
Maybe that could be tweaked.
dude, awesome. one suggestion: use the "MA" field as a center and have the good/bad parts sticking out from that, so that it's easier to tell the good/bad ratio.
That is a really good idea.
Now THAT is useful
I'd flip it so %Good passes are on the positive axis and %Bad passes are on the negative. But otherwise, well done!
1. Last year it was the ThreetSheridammit Chart. I think it's time to change names again.
How about the "Use-of-Force Chart"? Good and Bad can be "Light Side" and "Dark Side."
You may want to shift the order a bit, because a TA usually is a LOT better than a BR.
From Good to Bad (screens in-betweens):
- Dead On.
- Bad Read
Also, while we're on this, Brian -- it's time to incorporate rushing into the Force-Chart. The TA metric worked for pocket passers, but because the RR-era QBs are better running options, I'd like to have this charted.
You need three metrics for this. TA, RA, RA+. TA is a throw-away. RA stands for run-away. An ill-advised run-away is still a Bad Read. RA is when nothing's there, the QB takes whatever he can get (from getting back to the LOS to running in space until space runs out, running OOB to avoid contact, taking a slide, etc.). RA+ is for when the quarterback makes something out of nothing, juking someone, spinning past a tackle to get a 1st down, leaping for a TD, etc., where the QB goes above and beyond.
Use Greens for the graph.
So we end up with:
- Dead On.
- Runaway Plus
- Bad Read
How would it look in this format if you didn't normalize and just used attempts instead of %?
Excellent. I can also see this being very helpful when teaching younger kids the sport. It would probably be applicable to basketball shots as well.
I'm very much a numbers / chart guy, but I must admit, every time I look at a Hennechart I come away not entirely certain whether the results were good or bad (until Brian kindly explains it). This is like the Hennechart Rosetta Stone. It is, in fact, so excellent that I had to end years of lurking to congratulate you.
Also, Brian's idea about centering on Marginal and extending outward in both directions to make determination of the good/bad ratio easier is excellent. If it's not particularly difficult, it'd make this better still.
I realize the Hennechart's sole purpose is to evaluate the QB performance, but it might be good to have a line or something within each bar (where relevant) indicating the percentage of each category that was actually caught. It seems that would be a fairly trivial addition and would add significantly more data (i.e., the performance of our WRs). But, these look great and do really help to understand the performance.
A great idea, though I think that would be a more appropriate visualization for the receiver chart Brian produces. I think the two could happily live side-by-side, and would love to see it.
best diary...or post...or whatever it is. ever.
a billion cocktails to you, sir.
that is fantastic. good work, sir!great way to track trends, improvement, etc.
I'm not sure if this is now a pre-requisite for a complementary name for your creation, but I see from the google that Hennegraph is also a surname. At least for some theologian named Hank.
Awesome chart. Very easy to comprehend. Thanks for taking the time to do this.
You get a Wolvy Plus!
This is awesome. Now I regret the half hour I spend learning to fully grasp the Hennechart.
You are my hero Coach Schiano!
(well ... after Brian)
well done sir. A stacked line chart might also be pretty cool.
For us color-blind folk, this chart is useless. HAHA! But thanks for the effort.
Brian couldn't bring himself to UFR the "Strawberry Fields" game vs. OSU last year (that job fell to the defenders of the proletariat). So the NW game has the totals you can use for last year.
I waited and waited for Brian to UFR the Citrus Bowl, but alas, it was all OMG PRYOR MAKE A DECISION and a Who's Who of departing players and lots of entropy until the move. There was a UFR of the WMU/Rutgers game.
Again, no post-season UFR, so OSU charts it is.
Here too no post-season UFR. Brian explains:
I hope you'll forgive me: I ain't watching it again. Here's your UFR: screw 2005. The one redeeming feature of the last play of that game was I got to say "it's over" and put Michigan football, 2005, from my mind forever. In retrospect, even the run of success, or at least non-incompetence, that got us to 7-3 only served to raise our hopes just in time for the final two games to dash said hopes broken upon the rocks. Every step in the season seemed to raise the factor of cruel mockery to yet another level.
So: here lies 2005, killed by its own incompetence and that of others. Its gift to future generations is the phrase "well, at least it wasn't 2005." Try it: "well, we may have lost to MSU, but at least it wasn't 2005." "Well, I may have inoperable pancreatic cancer, but at least it isn't 2005." "Well, that rapture thing happened, my bet on Hinduism came up craps, and now I'm faced with hell on earth during Armageddon, but at least it isn't 2005."
By the way, Cook, I tried that mantra all last summer. Now I have a new one: "It can always get worse."
First, Coach Schiano, this is a great idea that was fantastically executed. Bravissimo.
Second, I strongly encourage re-ordering the events, such that what we consider to be the most strongly negative events are further to the right (this already seems to be the case for the positive events). That would take this chart from an A to an A+.
I think the reordering is a good idea. Placing the negatives all the way on the right makes it easier to compare their relative percentage because they all start at an equal spot (all the way on the right).
Well done Coach Schiano.
That is an excellent job of converting this into a visual graph. I know all the other commenters have said it already, but I love it!
I think you could do the same thing for the Receiver chart and get the same nice result. Heck, you could even do it for the defensive metrics!
I also like the idea of doing this historically. Imagine a link on the page where you could go for archive UFR data and charts. It would be like surfing MLB.com for historical hitting statistics! Then you can truly see how various Michigan quarterbacks compare to each other.
Then we can get Brian to hire an intern and have that intern do historical UFRs for the Griese/Brady/Henson days. My mind is spinning!!
Very nice! I think I would try adding one more wrinkle though: scale the height of the bar for each game by attempts. Then, the width of each color will represent the proportion of the throws that falls into that category (as it does now) and the area of the color will show the number of the throws that fall into that category. Then not only can you see in a glance in which game the QB was more efficient, but also how many he threw in absolute terms without having to do some division in your head.
Somebody should create a little app to generate these things from the raw numbers so Brian can easily incorporate them in the UFR. Hell, if I find a bit of time I might do just that.
It still suffers from the biggest criticism of the Hennechart: even with the (useful) categorization of the attempts, it doesn't accurately capture a player's performance. To wit, suppose Tater has 44 attempts consisting of 40 dead-on accurate passes and 4 bad reads. Your visualized chart would look VERY good (regardless of whether or not you normalize the data or use pure attempts). However, say those 4 bad reads all resulted in INTs. You've got a positive looking chart that describes what most would think of as a disastrous day.
So I like what you've done with the data you've been given, but I think in general the data fail to capture a perfect picture. Then again, that's the nature of reductionist statistics: information loss is effectively inevitable.
What you say is correct, but I don't think the result of the performance is something the Hennechart was ever supposed to represent. Rather, it is the expected performance of the QB. Whether a BR results in an INT or gets lucky and dropped by the D greatly impacts the results, of course, but aren't really different in terms of expected outcome - in both cases there was a high probability of an INT and thus were equally bad decisions. The QB just got luckier on one result than the other. The same could be said for dropped passes or circus catches by the receivers; the actual good or bad result is separate from whether the QB did his part at maximizing the chances of success.
I always thought it was interesting that the Engineer in Brian would "chart" aka table the results and not actually "chart" them with a graph...so well done.
I like this. It's nice to look at. Thanks.
Great idea, and nice execution.
Wow, I really like what you did with this charting. Makes a lot more sense than grinding the numbers on the sheet. This is well deserving of the pluses.