Data-Mining the HenneChart

or

How Tate Stacks Up Against M QBs of 2005-2008

Update 10/22: Lots of changes. Reorganized. New Sheridan charting added. New, better metric proposed. Now fat-free!

Note: links all open in a separate window/tab, in case you were afraid of having to find your place again.

[Editor's note: bumped to front page for holy crap reasons.]

This is a collection of data inspired by Coach Schiano's brilliant new visualization of the Henne/HenneMallett/ThreetSheridammit/UseTheForcier Chart.

First, kudos, thanks, plusses, and a dozen sacrificed bulls go out to Coach Schiano. This is the most excited I've been about some new type of Michigan coverage since I read my first UFR.

I hope he doesn't mind if I go back and try to play around with the same stuff.

I went and made my own Excel spreadsheet (available here) to see if I could visualize some of the things we've been talking about here.

I replicated Coach Schiano's results (tossing out the screen metric).

One thing that stuck out right away, I think, is that the introduction of the "Marginal" metric (first discussed after FOOTBALL ARMAGEDDON in 2006 but not implemented until mid-2008) kind of hurts the QBs functioning under it. Erstwhile "catchables" for Henne/Mallett were discounted for late '08 through '09 QBs. In other words, performances before last year's Penn State game, especially in accuracy, might be kind of overrated.

Also, there were a number of games not charted during this period. These include Delaware State '09 (and high school stats, for much the same reason); Ball State '06 (Comcast was blamed); all bowl games;* and last year's Ohio State game, which was only UFR'ed under a shadow Bolshevik government, and thus cannot be trusted.

* (a UFR for the 2005 Alamo Bowl was provided, but since the entirety of that UFR was, literally, "Screw 2005!", I was unable to get any passing info from it)

Part I: Where We Do It By Downfield Success Rating

The metric we've been using, gleaned from UFRs, is Downfield Success Rating. This divides the Dead-on, Catchable, and Marginal balls against all downfield (read: "Pressure excluced") attempts.

BEST 5 GAMES BY A QB

Season	Quarterback	Game	Opponent	W/L	DO	CA	MA	IN	BR	TA	BA	PR	ATT	DSR
2007	HENNE	7	Purdue	W	3	20	N/A	1	1	1	1	2	29	85.2%
2007	HENNE	6	Eastern Michigan	W	4	16	N/A	1	0	1	2	1	25	83.3%
2005	HENNE	11	Ohio State	L	6	27	N/A	1	1	4	1	N/A	40	82.5%
2008	SHERIDAN	10	Minnesota	W	2	20	4	3	2	1	0	0	32	78.6%
2005	HENNE	5	Michigan State	W	4	25	N/A	1	3	3	1	N/A	37	78.4%

Sheridan's Minnesota miracle looks even more miraculous now! Note, however, there were only 2 Dead-On throws, and 2 bad reads. Also, just one throwaway, no batted balls, and no pressure. That's the O-line and running game's fault.

In other coaching-related-results here, Henne's 2005 Ohio State game was scrumtrulescent by any metric, except apparently not good enough for DeBord to use the passing game to put the game out of reach. Not that I'm bitter...

WORST 5 GAMES

Season	Quarterback	Game	Opponent	W/L	DO	CA	MA	IN	BR	TA	BA	PR	ATT	DSR
2008	SHERIDAN	11	Northwestern	L	0	9	2	9	7	2	0	0	29	33.3%
2008	THREET	9	Purdue	L	1	7	2	7	2	2	2	1	24	38.1%
2008	THREET	4	Wisconsin	W	1	15	N/A	9	3	7	2	1	38	43.2%
2007	MALLETT	11	Wisconsin	L	3	13	N/A	11	6	2	1	2	38	44.4%
2009	FORCIER	6	Iowa	L	1	9	1	5	2	3	2	2	25	45.5%

Sheridan's putrid, I-Survived-That-Game-Patch-Inducing game against Northwestern is, unsurprisingly, at the top. In fact, it may even trump several of my 7th grade sexual experiences in the category of worst-executed-thing-ever-done-by-man.

But I'm getting off topic. Note that Forcier's Iowa game actually made this hall of shame. So RR's pulling him for that game may be excusable. However, it is still my considered position that even with shoulder/weather/concussion/bad game, had Forcier been given that last drive, he would have pulled himself back to respectability.
Though 2008 dominated the Top 5, if you look at the Top 10 we get Threet v. Miami (NTM), three Malletts (Min, PSU, NW) and Henne's all-things-good-in-the-world-destroying 2007 game against Ohio State.

BEST CAREER PERFORMANCE

Rank	Quarterback	DO	CA	MA	IN	BR	TA	BA	PR	ATT	DSR
1	HENNE	163	417	N/A	139	55	49	35	41	899	67.6%
2	FORCIER	17	97	9	19	15	19	2	16	194	67.5%
3	SHERIDAN	2	29	6	12	9	3	0	0	61	56.4%
4	THREET	16	97	7	45	21	24	11	11	232	52.8%
5	MALLETT	16	58	N/A	38	17	13	10	7	159	48.7%
6	ROBINSON	1	5	5	4	3	3	0	0	21	37.5%

Umm...waitaminute, Sheridan better than Mallett?!? Okay, I can see that since Mallett played like an idiot and Sheridan's Minnesota outlier was the bulk of his stats. But Threet? No way, right?

Right.

I think we're finding that Downfield Success Rate isn't all it's cracked up to be.

We're also finding out, as suggested in earlier posts, that the ability to run downfield is actually a hindrance to these stats, since it charts "Throw-aways" essentially the same as it does "Innacurates." Except a Denard Robinson "throwaway" could be a 20-yard juketacular.

The other problem I'm seeing is that it doesn't make a difference between catchable and Dead-Ons. Threet had 16 Dead-On balls thrown; Sheridan had two.

Lesson: DSR is flawed.

Part II: Weighting DSR

So...new metric. I don't have data to fix the run-away/throw-away problem. But I can rate each figure, thus:

DO: x2
CA: x1
MA: /2
IN: x1
BR: x2
TA: /3
BA: x1
PR: 0

Marginals get counted for you, but only half. Throw-aways get counted against you, but only a third.* Bad Reads and Dead Ons double. This, by the way, sucks for RR-era QBs, who have a lot more reading to do. Then again, they're in a better offense, so I'm letting that offset.

* I thought of removing them altogether, but think about it: when we run a passing play, and Denard jets rather than sit in the pocket to let the play develop, is that a good thing? TA's are a mini-neg, but a neg nonetheless.

So, weighting things...

BEST SEASON BY A QB (by weighted DSR)

Quarterback	DO	CA	MA	IN	BR	TA	BA	ATT	DSR	DSR (w)
2006 Henne	50	168	0	36	21	17	12	327	71.7%	73.69%
2009 Forcier	17	97	9	19	15	19	2	194	67.5%	70.27%
2005 Henne	91	114	0	56	23	24	17	325	63.1%	69.98%
2007 Henne	22	135	0	47	11	8	6	247	68.6%	69.74%
2008 Sheridan	5	57	6	19	11	9	2	110	60.2%	60.34%
2008 Threet	16	97	6	45	21	24	11	231	52.8%	55.46%
2007 Mallett	16	58	0	38	17	13	10	159	48.7%	51.04%
2009 Robinson	1	5	5	4	3	3	0	21	37.5%	46.34%

TOP TEN PERFORMANCES

Season	Quarterback	Game	Opponent	Result	DO	CA	MA	IN	BR	TA	BA	ATT	DSR (w)
2006	HENNE	5	Minnesota	W	5	13	0	2	0	3	0	24	88.46%
2005	HENNE	11	Ohio State	L	6	27	0	1	1	4	1	40	87.97%
2007	HENNE	6	Eastern Michigan	W	4	16	0	1	0	1	2	25	87.80%
2005	HENNE	3	Eastern Michigan	W	12	2	0	4	0	0	0	18	86.67%
2007	HENNE	7	Purdue	W	3	20	0	1	1	1	1	29	85.71%
2005	HENNE	10	Indiana	W	7	11	0	3	0	2	1	24	84.27%
2007	HENNE	8	Illinois	W	5	15	0	4	0	1	1	29	82.42%
2006	HENNE	4	Wisconsin	W	5	16	0	2	1	3	1	29	81.25%
2009	FORCIER	1	Western Michigan	W	2	14	1	2	1	2	0	25	79.86%
2005	HENNE	1	Northern Illinois	W	19	4	0	5	2	2	1	33	79.75%

Sheridan's dink-tacular Toledo game is omitted. Sheridan basically threw little outs and things got adventurous anytime the throw went more than 10 yards, so little dinky passes for negative yardage stacked up and gave him a ludicrously good-looking day (good for 9th best in UFR history) against Toledo, which we all know was not anything of the sort.

Lots of Henne. It speaks well of Forcier that his debut performance was somewhere in between what Henne did to MAC teams as a sophomore, and what he did to Wisconsin in a Big Ten Opener. Of course, Henne got two of these by beating up on EMU, and another was a particularly bad Indiana team. But that sophomore year performance against Ohio State is now second.

Let's go back to that UFR. Reading the first few drives gives us a few exempli gratia as to why Henne fills the top of the charts:

1st drive, 3rd and 5:

"A wobbly, ill thrown route that does make it to Manningham. Danielson points out a solid blitz pickup from Hart. (CA)"

1st drive, 4th and 1:

"The slant is there. Henne throws a bullet that's too high to be caught…no reason to fire this so hard. Just get it complete. Later, on replay, Danielson highlights the coverage from Youboty, which is uncalled holding. Beaten by Avant's outside feint, Youboty is forced to grab him outside the shoulder and give a heavy yank that throws the route off just enough for the incompletion. The ref looking at it is shielded by Avant's body, though--really hard for him to make that call. As it was, this is a great play by Youboty to skirt the edge of a call when he knows he's beaten and force the turnover on downs, but it's still holding. (CA minus the penalty)"

6th drive, 2nd and 9:

"Avant runs a deep corner route; Henne's throw is a little bit too far to the inside, allowing Salley to come up and make a play on the ball. A nice play from Salley but Henne gave him the opportunity to make it (as is usually the case on routes like that... can't stop a perfectly thrown ball). (CA... marginal, close to IN.)"

8th drive, 1st and 10:

"A ton of time. This one is thrown behind Manningham but the corner is too busy attempting to recover and does not locate the ball to make a play on it; Manningham makes the catch. (CA)"

These four plays would today have likely been classified as "MA" for "marginal." Barring a total redux of back UFRs for such things, I think we're stuck overrating Henne in posterity. On the other hand, ranking this among the top-ever games by a Michigan quarterback may not be that far off, considering some of the comments in that UFR:

"Drive Notes: Touchdown, 7-9, 5 min 2nd Q. A ten play, 36 yard drive… impressive or totally unsustainable going the length of the field against the Buckeye D? Er, the latter, as the rest of this game demonstrate. Still much competence displayed in portions here, especially by Henne, who looks like the guy we were all drooling over preseason."

Though execution errors on receivers and penalties (called and uncalled) did their fair share, the UFR of that game is clear: a 'Really Fucking On' day by Henne, and Ohio State's clear preference for loading the box to stop Hart, were not enough to deter the DeBordean School of Throwing Rock. When in the next 10-15 years Buckfards are witnessed heckling at Miami Dolphins games, Henne can blame DeBord.

Back to math, the Henne performance on here which receives the sternest look is that of Northern Illinois in the first-ever UFR, which you might expect would be buggy. Brian gave Henne a record 19 "Dead-On"s for that game. This persisted through the early 2005 season (if you divide DOs by DO+CA+MA for each game, the first four games of 2005 are 1 through 4, meaning Brian was WAY too liberal with DO for that game).

If you take the position that UFR during the MAC nummy-nums of 2005 was just a statistical fetus, the next down the list was Henne's performance in FOOTBALL ARMAGEDDON, and the 2005 Michigan State game. Sheridan's Minnesota game is dropped to 13.

Still, there's a definite problem here with the 2005 statistics. Either Henne couldn't miss early his sophomore year, or Brian was too kind with the ratings early on in the UFR process. I am going to assume the latter.

WORST PERFORMANCES

	Season	Quarterback	Game	Opponent	Result	DO	CA	MA	IN	BR	TA	BA	PR	ATT	DSR (w)
1	2008	SHERIDAN	11	Northwestern	L	0	9	2	9	7	2	0	0	29	29.70%
2	2008	THREET	9	Purdue	L	1	7	2	7	2	2	2	1	24	42.25%
3	2008	THREET	2	Miami (NTM)	W	0	6	0	4	1	0	2	0	13	42.86%
4	2007	MALLETT	11	Wisconsin	L	3	13	0	11	6	2	1	2	38	43.51%
5	2008	THREET	4	Wisconsin	W	1	15	0	9	3	7	2	1	38	46.79%
6	2009	FORCIER	6	Iowa	L	1	9	1	5	2	3	2	2	25	48.94%
7	2007	HENNE	12	Ohio State	L	1	13	0	12	1	1	0	4	32	51.14%
8	2007	MALLETT	4	Penn State	W	3	12	0	6	3	6	1	2	33	54.55%
9	2007	HENNE	2	Oregon	L	1	13	0	6	3	1	0	3	27	54.88%
10	2007	MALLETT	5	Northwestern	W	2	5	0	4	1	1	1	1	15	55.10%

On No. 7: Fie, FIE! upon the treacherous, lecherous Angry-Chad-Henne-Hating-God who hast tacken from our back of quarter his shoulder of Buckeye-smiting on this, his day of redeption. Of all the problems I might have with the universe and the way the space-time continuum has unfolded, this day still rankles me.

By weighted DSR, Sheridan's Northwestern game is even deeper into my middle school love life. 29.7 percent is off-the-charts hideous.

The rest of the Top 10 juggled but minorly. Almost every extended Mallett appearance for Michigan makes the list.

An important note here is that even among the worst performances of the last five season, 4 of 10 are still wins. Threet put in some awful games in 2008, but yet of those three appearances, we won two. And though in his limited sample for that game Threet was God-awful, our offense was more than good enough on the one loss (Purdue) to win it. The reason: the offense.

If we're talking about coaching wins despite bad QB play, forget the 2007 Northwestern performance by Mallett -- Henne had almost twice the attempts in that game, and the offensive strategy was so bad it made Brian erupt in disgust in its aftermath. As for Penn State '07, which was a defensive win and offensive nightmare. No points for DeBord.

But RR? The Miami (NTM) win last year is telling. In Brian's post-UFR discussion, he postulated that this performance was historically bad:

"Dude, Tacopants is going to catch 400 balls this year.

No, because even he’s watching these sail over his head, and he can be whatever height he wants to be because he is made of dreams and snails and puppy dog tails.

But seriously, folks, this may have been the worst quarterbacked game in the modern era(defined as Moeller forward) of Michigan football."

And he was probably right, since Northwestern hadn't happened yet. Yet we WON :) And this is because:

"Under Rodriguez, Michigan either has a quarterback just take off—easy to do if you’re a fast guy and even functional if you’re Threet, apparently—or throws a bubble screen to a guy like Odoms and hopes he gets a block."

And against Wisconsin we won. And against Purdue, the offense was still enough to win but the defense was D.C.-firingly awful. And against Northwestern, with the worst quarterbacking job by a guy not trying to lose in history, we had the ball down a touchdown on our own 42 with a minute to go.

Point: offensive idiocy by DeBord cost Michigan an OHIO STATE win on a day that we got Peyton Manning-on-his-best-day-level quarterbacking, while offensive genius by Magee/Rodriguez gave Michigan two wins and kept us in position for three more on days of historic QB incompetence.

Sorry, this is supposed to be about the quarterbacks.

When we calculate the careers of our QBs, it should again be no surprise that:

BEST CAREER

	Quarterback	DO	CA	MA	IN	BR	TA	BA	ATT	DSR (w)
1	HENNE	163	417	0	139	55	49	35	899	71.21%
2	FORCIER	17	97	9	19	15	19	2	194	70.27%
3	SHERIDAN	5	57	6	19	11	9	2	110	60.34%
4	THREET	16	97	6	45	21	24	11	231	55.46%
5	MALLETT	16	58	0	38	17	13	10	159	51.04%
6	ROBINSON	1	5	5	4	3	3	0	21	46.34%

In other words, just a few games into his freshman year, Forcier is already just under the sophomore-to-senior performance of Henne. And that's still including Henne's inflated Dead-On gimmes from the first half of 2005.

It's also interesting to note the Forcier/Threet comparison. Tate is now 1 or 2 games away from Threet's number of attempts, but already thrown one more Dead-On and the same amount of catchables. In bad reads, Tate has 15, while Threet had 21 -- a slight but not huge improvement. Threet had way more batted balls (11) than Tate (2), which I think is a function of Forcier rolling out a lot more.

The real difference: Incompletes. Of those balls that weren't bad reads, throwaways or batted, just 12 percent of Forcier's throws ended up uncatchable, while for Threet that number was 25.6 percent.

Part III: Components?

So maybe what we really have here is an accuracy statistic, not a quarterback rating. So let's do straight-up accuracy.

Of balls thrown in the direction of open-receiver-type objects (those charted as DO, CA, MA or IN), here's your QBs in percent that made it there:

ACCURACY RATING:

	Quarterback	Throws	Accuracy	%DO
1	FORCIER	142	86.62%	13.82%
2	HENNE	719	80.67%	28.10%
3	SHERIDAN	87	78.16%	7.35%
4	ROBINSON	15	73.33%	9.09%
5	THREET	164	72.56%	13.45%
6	MALLETT	112	66.07%	21.62%

I guess it's true: Forcier = Accuracy. Sheridan is third and within striking range of Henne (though with far fewer dead-ons), perhaps justifying RR's strange confidence in him. But before you try to screw your memory toward making Sheridan as a less-on-target Henne, I warn you now this is just a sampling error; Sheridan still sucked.

The other side of the coin: decision-making. Your QBs in percent of Attempts that result in Bad Reads:

	Quarterback	BR/Att
1	HENNE	6.12%
2	FORCIER	7.73%
3	THREET	9.09%
4	SHERIDAN	10.00%
5	MALLETT	10.69%
6	ROBINSON	14.29%

Couple things of note here. For one, we see that any ground Henne may have lost to Forcier on accuracy/finding Tacopants is made up for with decision-making. On the other hand, Forcier is a freshman, so that mark is almost certain to improve in years to come. Already it is firmly ahead of Threet, and way beyond Mallett at Tate's age.

So now we have two good statistics, one measuring accuracy and the other decision-making, which can be used to qualify a Quarterback. From here we can combine these to create a QB Rating (Accuracy divided by % of bad reads). I think a rating of 10 is an average QB.

Quarterback	Accuracy	BR/Att	Rating
HENNE	80.67%	6.12%	13.19
FORCIER	86.62%	7.73%	11.20
THREET	72.56%	9.09%	7.98
SHERIDAN	78.16%	10.00%	7.82
MALLETT	66.07%	10.69%	6.18
ROBINSON	73.33%	14.29%	5.13

As expected: Henne is awesome, and Forcier is not quite sophomore-to-senior Henne but is a major upgrade over Threet.

So, does this metric work?

Adding in the charted appearances for Sheridan bumped him up several of these lists, and reveals a major problem with it: a 20-yard pass to a guy open for a half-second against Ohio State (CA) gets charted the same as a 4-yard screen pass that takes so long to get there the receiver gets blown up immediately after contact.

The other problem with including Sheridan is that two of his most telling performances (Penn State's 2nd half and the entire Ohio State game) were not charted because they were so bad. Sheridan thus gets an "INCOMPLETE" for all career metrics and should not be considered among them. If someone would like to go back and UFR the unwatchable chapters of last year, I would love to incorporate those stats.

Of course, Sheridan (as you would expect from a great future coach) also points out, using his Toledo performance, that per-game metrics also suffer from the This thing needs a yards-per-attempt metric.

So let's add one, and create a complete revised QB rating from the UFRs. Really simple: multiply the rating above with yards-per-attempt from the official stats.

We get:

Quarterback	ATT	DSR	DSR (w)	%DO	Rating	YPA	QB Rating
HENNE	899	67.6%	71.21%	28.10%	13.19	6.98	92
FORCIER	194	67.5%	70.27%	13.82%	11.20	7.16	80
THREET	231	52.8%	55.46%	13.45%	7.98	5.53	44
MALLETT	159	48.7%	51.04%	21.62%	6.18	6.41	40
SHERIDAN	110	60.2%	60.34%	7.35%	7.82	5.01	39
ROBINSON	21	37.5%	46.34%	9.09%	5.13	5.80	30

As I said before, Sheridan's inclusion is meaningless if you can't count OSU and PSU against him. Still, the Yards-per-Attempt metric busts him and Threet way down for all the screens and dinkies thrown last year, and gives us a final metric I think we can believe in.

Note Forcier's yards-per-attempt this year have been higher than Henne's career average. That will come down as more Big Ten teams and fewer MAC nummy-nums work their way into his statistics.

Our UFR-based QB rating isn't like the NFL's -- it doesn't count touchdowns, for one. But if we're looking for new ways to visualize the Hennechart, and use it to gauge how good our quarterbacks really are, I think this might be it.

Part IV: Oh, But What if We Multiply Weighted Downfield Success Rating With Yards-Per-Attempt?

We get this:

	Quarterback	DSR*YPA
1	FORCIER	5.03
2	HENNE	4.97
3	MALLETT	3.27
4	THREET	3.06
5	SHERIDAN	3.02
6	ROBINSON	2.69

Which is eerily close to star ratings, isn't it?

So muh-guh-bluh-gers, which one do you prefer, DSR*YPA, or the Rating thing?