Pages

Sunday, May 19, 2013

TWTW!! Maybe Hawk Harrelson has it right



After listening to the recent debate between the Hawk and ESPN's Brian Kenney (YouTube below), I couldn't help but think that maybe the Hawk has a point, maybe two. Whether you call it TWTW or GRIT or whatever, intangibles are a quality that by definition do not lend themselves to being quantified by definition, but you know them when you see them.

from thefreedictionary.com:
http://www.thefreedictionary.com/intangible
3.intangible - hard to pin down or identify; "an intangible feeling of impending disaster"
unidentifiable - impossible to identify
Some guys have that quality of making the other guys around them better. They make the total of the team greater than the sum of its parts. They are leaders and chemistry builders. And some guys are the opposite.

Winners have the intangibles!!! And like many things, you don't appreciate what you've got until after it's gone. Then you realize the hard to quantify value-added that these guys provide after they are missing from the equation. Then the team is equal to ( mediocrity ) or maybe even less than the sum of its parts ( under-achievers ).  Of course, then it's too late.

That's what Hawk is trying to explain to the geek squad here, IMO.

from The Big Lead:



The first example that came to my mind was Tim Tebow in 2011 with Denver.

Artistically, not a QB work of art.
Statistically, not a work of art.
But his teams win consistently.

It took the Jets to break the streak and that says more about the Jets than it does Tebow.  The Jets have been
a cesspool of bad personnel decisions and Tebow was supposed to come in and be the team plunger how?
By holding a clipboard?

Anyway, I thought I would take a look at the 2011 year in quarterbacking in the NFL and began with the
premise that the better the QB statistically, the better the teams record would be.

So,  I wanted to see the correlation between the ESPN total QB rating and winning.

In addition, I wanted to see if there were any examples like Tebow of QB's who violated the statistical model,
so to speak. Who won sometimes in spite of their liabilities as a QB as defined by the statistical metric.
They were "bad" QB's who just won ball games.

Here is the metric and how they determine the QB rating.

There were some interesting results (shown below).

from ESPN:

Glossary

  • * Season Leaders: On pace for 250 action plays.
  • * All-time data reflects 2008 onwards.
  • PASS EPA: Clutch-weighted expected points added on plays with pass attempts.
  • RUN EPA: Clutch-weighted expected points added through scrambles, designed rushes and fumbles/fumble returns on running plays.
  • SACK EPA: Clutch-weighted expected points added (lost) on sacks.
  • PEN EPA: Clutch-weighted expected points added on penalties.
  • TOTAL EPA: Total clutch-weighted expected points added.
  • ACT PLAYS: Plays on which the QB has a non-zero expected points contribution. Includes most plays

  • that are not handoffs.
  • QB PAR: Number of points contributed by a quarterback over the season, accounting for QBR and

  • how much he plays, above the level of a quarterback who plays very rarely and is on the fringe of the NFL.
  • QB PAA: Number of points contributed by a quarterback over the season, accounting for QBR and how much he plays, above the level of an average quarterback.
  • TOTAL QBR: Total Quarterback Rating, which values quarterback on all play types on a 0-to-100 scale.


A quick primer on the fundamentals of Total Quarterback Rating:
Scoring: 0-100, from low to high. An average QB would be at 50.
Win Probability: All QB plays are scored based on how much they contribute to a win. By determining expected point totals for almost any situation, Total QBR is able to apply points to a quarterback based on every type of play he would be involved in.
Dividing Credit: Total QBR factors in such things as overthrows, underthrows, yards after the catch and more to accurately determine how much a QB contributes to each play.
Clutch Index: How critical a certain play is based on when it happens in a game is factored into the score.

For the W-L data for each QB, I used CBS Sports NFL data. The W-L data includes only games where the
QB started and includes playoff games if the team advanced that far. (TWTW)

from CBSSports.com:



The correlation between the Total QBR and the QB winning percentage (W_Pct) as a starter was 0.64.

No surprise, the NFL is a QB driven league nowadays.

Based on this small sample, the correlation number implies that about 40% of a teams winning percentage is
derived from the play of their quarterback.


The first thing that jumps out are the names at the top of the list are WINNERS. The creme de la creme of
the NFL. Aaron Rogers, Drew Brees, Tom Brady, Tony Romo and Matt Ryan are the top five and win at
about the rate that their stellar play implies

The bottom of the list gives us Blaine Gabbert, Curtis Painter, Sam Bradford, Tim Tebow and Mark Sanchez.
Painter and Gabbert won at about the level of play their QBR would indicate. Sam Bradford seemed to just
have one of those miserable season where everything went wrong for him and his team. But Tebow and Sanchez
both out kicked their coverage so to speak.

Sanchez had a QBR of 33.6 and an W_Pct of 50.0% for a plus 16.40 score.
Tebow came in with a 29.9 QBr and a W_Pct of 61.5% for a plus 31.64 score.

The list of over-achievers, possessing dare I say TWTW, were:

John Skelton @ +39.90
Alex Smith @ +31.98
Tim Tebow @ +31.64
Mark Sanchez @ +16.40
Joe Flacco @ +12.52
Tom Brady @ +10.63
Jay Cutler @ +10.20

The interesting name on the list was Skelton who out-performed a more highly regarded QB in Kevin Kolb
( 34.4 QBR - 33.33 W_Pct) for the same Arizona Cardinal team. Apparently, Skelton may have had had
that je ne sais quoi, the TWTW that Hawk is alluding to and he was able to rally the team around him as
Tebow did the Broncos in relief of Kyle Orton.

Skelton and Tebow are interesting because they took over the same team, with the same defensive squad
and rallied them from depths to heights. That result defines what Hawk is speaking about. We're just not
certain how to identify it in advance, but we can generally spot it after the fact pretty well.

The rest of the overachiever names you can generally say had good to excellent defensive teams around
them. How much of the teams performance can be attributed to good defense and how much of the
defensive performance is aided by good QB play (keeping them off the field, playing from ahead more
often than from behind, etc.) is fodder for another post.

Hindsight and many forms of statistical analysis are always 20/20.  I think that is the Hawk's frustration
with the SABR crowd and it is well placed in some instances.

Looking at the other end of the scale, the under-achievers yields some interesting names. These guys
QBR was significantly higher than their team record for whatever reason. Maybe they were stat
gatherers and not team motivators. maybe they were on bad teams.

 Kyle Orton @ -29.80
Tony Romo @ -21.40
Josh Freeman @ -18.63
Sam Bradford @ -18.60
Carson Palmer @ -18.16
Matt Schaub @ -17.50
Cam Newton @ -17.50

Some of these guys were on good teams, with good defenses, soooooooooo.......IDK.

Lets just say I would score this one in favor of Hawk Harrelson over Brian Kenney
and Harold Reynolds by a TKO.
STOP THE FIGHT!!

And they can roll their eyes and snicker amongst themselves all they want.







TWTW
Total QBR W-L W_Pct W_Pct Diff Player
86.2    15-2 88.2%        88.24        2.04 Aaron Rodgers
84.0            14-4 77.8%        77.78       (6.22) Drew Brees
72.7       15-3 83.3%        83.33      10.63 Tom Brady
71.4      8-8 50.0%        50.00     (21.40) Tony Romo
69.1    10-6 62.5%        62.50       (6.60) Matt Ryan
67.5             5-5 50.0%        50.00     (17.50) Matt Schaub
65.5    10-6 62.5%        62.50       (3.00) Matt Stafford
64.4      7-6 53.8%        53.85     (10.55) Michael Vick
63.6    11-5 68.8%        68.75        5.15 Ben Roethlisberger
62.8      9-7 56.3%        56.25       (6.55) Matt Hasselbeck
62.7      8-8 50.0%        50.00     (12.70) Philip Rivers
62.6      4-5 44.4%        44.44     (18.16) Carson Palmer
59.8      7-3 70.0%        70.00      10.20 Jay Cutler
59.7    13-5 72.2%        72.22      12.52 Joe Flacco
59.4    12-7 54.5%        54.55       (4.85) Eli Manning
56.6      6-6 50.0%        50.00       (6.60) Matt Moore
55.0    6-10 37.5%        37.50     (17.50) Cam Newton
51.2      4-5 44.4%        44.44       (6.76) Matt Cassell
50.5    6-10 37.5%        37.50     (13.00) Ryan Fitzpatrick
49.8      1-4 20.0%        20.00     (29.80) Kyle Orton
45.8      9-7 56.3%        56.25      10.45 Andy Dalton
45.8    14-4 77.8%        77.78      31.98 Alex Smith
45.3    4-11 26.7%        26.67     (18.63) Josh Freeman
43.9      5-8 38.5%        38.46       (5.44) Rex Grossman
40.1      4-9 30.8%        30.77       (9.33) Colt McCoy
37.7      7-8 46.7%        46.67        8.97 Tarvaris Jackson
35.1      6-2 75.0%        75.00      39.90 John Skelton
34.4      3-6 33.3%        33.33       (1.07) Kevin Kolb
33.7      2-9 18.2%        18.18     (15.52) Christian Ponder
33.6      8-8 50.0%        50.00      16.40 Mark Sanchez
29.9      8-5 61.5%        61.54      31.64 Tim Tebow
28.6      1-9 10.0%        10.00     (18.60) Sam Bradford
22.5    2-12 14.3%        14.29       (8.21) Curtis Painter
20.6    4-11 26.7%        26.67        6.07 Blaine Gabbert
0.641328597   Correl OBR - W_Pct
 GUILFORD’S SUGGESTED INTERPRETATION FOR CORRELATION COEFFICIENT VALUES
Value  Interpretation
Less than .20    Less than .20 Slight, almost negligible relationship
.20 - .40  .20 - .40 Low correlation; definite but small relationship
.40 - .70  .40 - .70 Moderate correlation; substantial relationship
.70 - .90  .70 - .90 High correlation; marked relationship
.90 - 1.00  .90 – 1.00 Very high correlation; very dependable relationship

No comments:

Post a Comment