Brandon Mebane's Monster 2008, Modest 2009 and Following Up on WPA+ and EPA+

by John Morgan on May 10, 2010 1:14 PM PDT

Darryl Tapp was Seattle's best defensive lineman in 2009 as measured by EPA+, and by a wide margin. Plays he was involved in were worth 39.8 points for Seattle. That's more than Mario Williams, and just a hair below Dwight Freeney and Robert Mathis. This doesn't tell us the whole story, but it does add a little substance to Tapp's otherwise weak counting numbers.

That is one of a host of interesting finds found in Brian Burke's individual defender database. Another not too surprising fact: Brandon Mebane was better in 2008 than he was in 2009. In 2008, Mebane was among the cream of the crop. He ranked between Jay Ratliff and Jamal Williams for ninth in the NFL in EPA+. (I prefer EPA+ to WPA+ because it removes context) Mebane added 34.4 points. Colin Cole was Seattle's most valuable defensive tackle by EPA+ in 2009. He finished 15th with 25.7 EPA+. The big difference between Cole and Mebane, both 1-techs but in different seasons, is that Mebane contributed to both the pass and run defense. Cole contributed exclusively as a run defender. He had no sacks or quarterback hits. Mebane had six and 18.

The key to properly interpreting data like this is to combine it with scouting to create a total portrait. The aggregate of all run plays in the NFL is worth negative WPA, and so defenders that contribute to the run defense will tally high WPA+ and EPA+ values, but could be hurting their teams in ways that WPA+ and EPA+ do not properly measure. That's Colin Cole. That also accounts a little for Tapp's total EPA. Tapp is good, but not nearly as good as Freeney. Freeney contributes hundreds of pressures that are never recorded in the play-by-play. He forces outlet passes and double teams. So tread carefully. WPA+ and EPA+ are not definitive, but they're useful and yet another tool for interpreting the NFL.

Tweet Comment 14 comments | 0 recs |

Do you like this story?

Comments

Man, both of these articles were too Grim Truth for my fragile sense of fandom.

Talk more about bench press weight.

by DJ C-Raig on May 10, 2010 2:41 PM PDT reply actions

I wish there was somewhere on that site that explained their valuation

I mean a breakdown of how points are totalled for a defensive player in a series. How many points are being awarded for each type of play? Is it strictly field position? If a half sack is worth a whole sack why do they give Robert Mathis 10 sacks when he should have had 11? Why do they give Mario Williams credit for playing only 15 games when he played 16? Way too many questions about these stats.

by Kevaru on May 10, 2010 3:30 PM PDT reply actions

I'm sure they're

90% correct, 60% of the time.

by m_b on May 10, 2010 4:40 PM PDT up reply actions

Williams had a game where he accumulated no stats, that may be why he shows up as only playing 15 games.

Mathis had 9.5 sacks according to ESPN and NFL.com, I’d guess he gets full credit for the sack in Burke’s system.

As far as how points are awarded it’s done strictly by what impact on the opposing teams point expectancy that the player has. For example, If Mathis gets a sack that lowers the other team’s expected points value by 0.5 points he’s awarded that many points in his +EPA. +EPA only measures positive plays, if a player makes a tackle that increases the opposing teams EP then it’s not counted against him (which makes sense because, in an analogy Burke uses, you wouldn’t necessarily want to penalize a safety for making a touch down saving tackle on a 50 yard pass completion). +WPA is same, I think, except it uses win percentage instead of expected points.

You can read about Expected Points here, Win Percentage here and WPA/EPA here.

by Nate Dogg on May 10, 2010 7:27 PM PDT up reply actions

If Williams accumulated no stats but still played in the game then that should be reflected in his +EPA/G.

By not counting a game because he didn’t register a tackle, that raises this stat as the +EPA/G reflects the number of games played. As to the Mathis sacks, he had 9.5 sacks but in 3 of those games he registered a half sack. That would raise his total to 11 sacks (8 sacks + 3 half sacks). It’s these inconsistencies that really make me question that site’s legitimacy. Stats guys are usually meticulous and obvious deviations from accepted statistical data is often explained thoroughly.

As to his EPA explanation, I understand it if referred to the defense as a whole but what about the individual player? If a fumble and return happens resulting in a EPA of +2.1 how much of that is awarded to the DE who flushes the QB out of the pocket, how much is awarded to the DT who grabs at the QB’s arm, how much to the LB who strips the ball and how much to the DB who picks up the loose ball and returns it 30 yards? Are they all awarded the same amount? Does the DE who flushes the QB out and DT who grabs the QB get nothing? Obviously each person made a good play and contributed but are they all awarded equally? I just don’t see an explanation anywhere.

by Kevaru on May 10, 2010 8:08 PM PDT up reply actions

Only the players awarded with a stat would be given any points. The DT might get a half sack but the DE is out of luck.

As far as Williams and Mathis go thats something you should probably drop him about. I would guess his system rounds up half sacks and doesn’t count games where players don’t accumulate stats and that should be corrected.

by Nate Dogg on May 10, 2010 8:42 PM PDT up reply actions 1 recs

That photograph needs to be shopped

to include Ice Cube and McCoy. Meet your new seattle seachronics.

by hazbro24 on May 10, 2010 7:26 PM PDT reply actions

Or

http://www.theonion.com/articles/nfl-meteorologists-warn-steaming-blackguy-heads-oc,2340/

by jacobstevens on May 10, 2010 10:30 PM PDT up reply actions

So, to summarize

WPA+/EPA+ equates Daryl Tapp with Dwight Freeney, and says Colin Cole was Seattle’s best D-lineman in 2009.

How does this add anything to what we can already observe (or, in my case, read about other people observing) on game tape? It just seems far too noisy and context-dependent to provide anything other than post-hoc confirmation of previously held beliefs.

by cyberwulf on May 11, 2010 11:17 AM PDT reply actions