Five different teams from last year's DVOA top eight rank in the bottom half of the league through four weeks of 2014. What can we learn from other teams with similar starts in the past?
12 Dec 2008
by Bill Connelly
I explored EqPt values in three different ways: 1) looking at yard line, 2) looking at down and yard line, 3) looking at down, distance and yard line. I will talk about (2) and (3) in the future -- that sort of point derivation has a purpose and gives you a lot of credit for passing downs success and go-to ability on third downs. But when tying EqPts to real points, looking at yard line was (unexpectedly) the method most directly attributable to real point totals. So that's what I am following for now.
Well, today is the future. It's time to talk about what I'm calling second-level EqPts and third-level EqPts.
First, a refresher. The concept behind EqPts is a simple one. Each yard line has a point value assigned to it based on the number of points an offense can expect to score in a possession involving a play from that yard line. The 20-yard line is worth 1.179 EqPts. The opposing 20-yard line is worth 3.898. Each play is assigned a point value based on where the play started and where it finished. The slope is different throughout the field. A gain of five yards from your 20 to your 25 is worth 0.064 EqPts. From the opponent's 40 to the opponent's 35: 0.441 EqPts. From the opponent's 10 to the opponent's 5: 0.582 EqPts. The concept of EqPts is intended to give more weight to more important gains.
Second- and third-level EqPts are in essence derivatives of the EqPts concept. Second-level EqPts take two factors into account: yard line and down. First down from your 20 (1.488) has a different point value than third down from your 20 (0.808). Another example: First down from the opponent's 10 is worth 4.881 EqPts, second down is worth 4.477, and third down is worth 3.818. It's not a hard concept to explain; the points you can expect to score go down as the down markers go up.
With second-level EqPts, a play can have a negative value without losing yards. As mentioned in the quote above, second-level EqPts don't tie as closely to actual point values (because of the possibility of negative values, the sum of all second-level EqPt values is only about 82 percent of the sum of first-level EqPts), but looking at second-level Points Per Play (or 2ndPPP) adds a factor of timely play-making to the equation -- because in football, as in comedy and sniper-fire, it's all about timing -- while further subtracting empty gains on second- or third-and-long.
As you will see below, the differences in rankings between PPP and 2ndPPP are not tremendously significant, but the shifts are telling.
Important note: As with last week's data, all tables below are looking at only BCS-conference teams; not all non-BCS games have been entered yet.
|2ndPPP: Best BCS Offenses|
In effect, the 2ndPPP Top 10 is just a rearranged PPP Top 10. There are no new names at the top, but plenty of teams flipped quite far in the rankings.
Biggest Drops from PPP ranking to 2ndPPP ranking: California -17 (No. 14 to No. 31), Connecticut -10 (No. 42 to No. 52).
Biggest Gains from PPP ranking to 2ndPPP ranking: North Carolina State +10 (No. 43 to No. 33), Syracuse +9 (No. 55 to No. 46), Northwestern +9 (No. 49 to No. 40).
Let's look at defenses the same way.
|2ndPPP: Best BCS Defenses|
Defensive rankings were a bit more volatile; more teams moved up or down rather significantly.
Biggest Drops from PPP ranking to 2ndPPP ranking: Maryland -20 (No. 28 to No. 48), North Carolina -15 (No. 16 to No. 31), Northwestern -11 (No. 19 to No. 32), Pittsburgh -11 (No. 36 to No. 47), Rutgers -11 (No. 37 to No. 48), Penn State -10 (No. 3 to No. 13), Arizona State (No. 20 to No. 30).
Biggest Gains from PPP ranking to 2ndPPP ranking: Wisconsin +20 (No. 44 to No. 24), South Florida +13 (No. 22 to No. 9), Illinois +13 (No. 46 to No. 33), Auburn +11 (No. 25 to No. 14), Florida State +10 (No. 21 to No. 11), UCLA +10 (No. 49 to No. 39).
Naturally, when you begin to play with the concept of down-dependent EqPts, the thought to take it a step further persists.
With third-level EqPts, we look at not only down and yard line, but also down distance. Based on distances with similar success rates, I bunched down distances together as follows. And I went with the most evident category names possible.
For third-level EqPts, we assign EqPts to plays run on a specific down, on a specific yard line, and of a specific categorized distance. Put into action, using a series of 4-yard gains to illustrate, it looks like this:
First-and-10 (medium-long) from the 20: 1.834 EqPts.
Second-and-6 (medium) from the 24: 1.758 EqPts (the 4-yard gain on first down was worth -0.076 EqPts.)
Third-and-2 (short) from the 28: 1.399 EqPts (the 4-yard gain on second down was worth -0.359.)
First-and-10 (medium-long) from the 32: 2.257 EqPts (the 4-yard gain on third down was worth 0.858.)
As you can see, third-level EqPts put an extreme premium on first downs. The slope changes throughout the field, but the main premise is the same: Until you move the chains, you are probably losing points. This is reflected in the averages as well. With the predominance of negative values, the sum of all third-level EqPts is about 30 percent that of first-level EqPts.
So with this focus in mind, let's take a look at the Best BCS offenses and defenses according to third-level PPP (3rdPPP).
|3rdPPP: Best Offenses|
Biggest Drops from PPP ranking to 3rdPPP ranking: Rutgers -52 (No. 12 to No. 64), Texas A&M -29 (No. 36 to No. 65), California (No. 14 to No. 33), North Carolina (No. 27 to No. 40).
Biggest Gains from PPP ranking to 3rdPPP ranking: Syracuse +18 (No. 55 to No. 37), Northwestern +17 (No. 49 to No. 32), North Carolina State +15 (No. 43 to No. 28).
Other than the baffling upgrade of Syracuse from a terrible BCS offense to simply an average one, these rankings do show you where some teams' downfalls lied. Three of four teams on the "biggest drops" list had offenses with all sorts of explosive potential, but their inconsistencies in moving the chains held them back. Meanwhile, Northwestern went 9-3 in 2008 despite ranking 62nd in total offense and 53rd in total defense. As their 3rdPPP suggests, though, they made the most of their opportunities and converted on third downs (46 percent for the season), and that's largely why they went 5-1 in games decided by a possession or less.
|3rdPPP: Best Defenses|
Biggest Drops from PPP ranking to 3rdPPP ranking: Northwestern -24 (No. 19 to No. 43), North Carolina -23 (No. 16 to No. 39), Maryland -21 (No. 28 to No. 49), Rutgers -17 (No. 36 to No. 53), Arizona State -15 (No. 20 to No. 35), Clemson -14 (No. 9 to No. 23), Pittsburgh -12 (No. 36 to No. 48).
Biggest Gains from PPP ranking to 3rdPPP ranking: Oklahoma State +47 (No. 50 to No. 3), Illinois +19 (No. 46 to No. 27), Florida State +16 (No. 21 to No. 5), Cincinnati +13 (No. 14 to No. 1), Oklahoma +12 (No. 34 to No. 22), Oregon State +12 (No. 41 to No. 29).
Once again, the defensive numbers are more volatile than the offensive numbers. Northwestern's appearance on the biggest drops list is a bit confusing (their 34 percent third-down conversion rate allowed was quite solid) and Oklahoma State's drastic leap from No. 50 in PPP to No. 3 in 3rdPPP is strange (they game-planned wonderfully against teams like Missouri, Texas and Baylor -- and their performance was impressive enough to land Defensive Coordinator Tim Beckman the Toledo head coaching job -- but that is still quite the leap). In all, though, you do see teams' defensive identities begin to stand out. Florida State was the best in the country on passing downs for most of the year, and that accounts for their leap into the 3rdPPP top 5. Meanwhile, teams like Cincinnati and Oklahoma rode timely defense to their respective conference titles, while Oregon State almost did the same.
The measures we are discussing today can go further than just evaluating teams, though. They can also be used to evaluated individual players, particularly at offensive skill positions. Let's take a quick look at running backs.
Below is a table of the top 10 rushers (by yards) from BCS conferences. Included with their yardage totals are their ranks in PPP, 2ndPPP and 3rdPPP among the 86 BCS ball carriers with at least 100 carries for the season.
|Top BCS Running Backs|
|Rank||Player||Team||Yds Per Game||PPP Rank||2ndPPP Rank||3rdPPP Rank|
|3||Javon Ringer||Michigan State||132.5||56||9||60|
|5||Kendall Hunter||Oklahoma State||126.5||9||6||17|
|6||Beanie Wells||Ohio State||121.2||43||46||63|
|8||Jacquizz Rodgers||Oregon State||113.9||58||15||48|
|10||Jonathan Dwyer||Georgia Tech||110.7||14||37||30|
You will probably also notice that nobody from that list makes the Top 10 of the 3rdPPP rankings. Why is that? Because that Top 10 is littered with quarterbacks (Texas' Colt McCoy No. 2, Oklahoma State's Zac Robinson No. 5, Ohio State's Terrelle Pryor No. 6, Baylor's Robert Griffin No. 7, and Florida's Tim Tebow No. 8). This makes sense; good rushing quarterbacks move the chains in key situations, and if you were to create a list of the best rushing quarterbacks, all of the above names would be on the list.
So who are some of the best running backs in college football according to EqPt Values, who aren't on the list above? Here are some of the names that show up at the top of the list:
Want to make a sabermetrician rant and rave? Tell him (or her, ahem) how "clutch" a given hitter/pitcher is. They will tell you about small sample sizes and the like, and they will tell you that "clutch" hitting or pitching does not necessarily exist in any meaningful way, and for the most part they will be right.
But is it the same with, say, quarterbacks?
What if we looked at the following: 3rd-degree EqPts in the fourth quarter of a close (i.e., within two possessions) game? When you speak of coming up big in the clutch, chances are it would be a big moment in the closing minutes of a tight game, right? So is this a way to measure "clutch" play in quarterbacks?
The following table looks at 3rdPPP for quarterbacks in plays satisfying the above criteria (fourth quarter, close game). It looks at the offense's performance as a whole in those situations, not just pass attempts; I would suggest that clutch quarterbacking involves managing the offense as a whole, not just throwing the ball.
|"Clutch" BCS Quarterbacks|
|3||Graham Harrell||Texas Tech||115||0.411|
|6||Cameron Sexton||North Carolina||53||0.333|
|8||Russell Wilson||North Carolina State||72||0.301|
|11||Josh Freeman||Kansas State||64||0.283|
|13||Sean Canfield||Oregon State||45||0.273|
|14||Chris Smelley||South Carolina||78||0.250|
The list contains most of the quarterbacks you would expect -- Tebow, Harrell, Bradford, McCoy -- but some interesting suspects are interspersed throughout. It's not a perfect list, but it's an intriguing one.
The EqPts concept is one of the key components of the play-by-play numbers on which this column is based. It is a concept with potential beyond just Points Per Play and "+" numbers. The idea of second- and third-level EqPts begins to explore that potential. With these figures, you can further explore the ideas of empty yards (long gains that bolster stats but don't really help a team score or win) and the importance of timing in moving the ball down the field.
Thoughts or suggestions? Are there other criteria that would further explore the "clutch" concept? Other factors we should take into account when assigning EqPts?
Now some responses to comments from last week's column.
Given that both OU and TTU, for example, have S&P+ in Close on Offense greater than S&P+ in both overall Rush Off and Pass Off, while Mizzou is the opposite, putting both overall Rush Off and Pass Off doesn't seem to be providing very much valuable information. Relatedly, I also wonder about the balance in Rush Off and Pass Off, and if that's skewing your statistics. USC and Ohio State are two prominent teams that S&P+ is telling us have better Pass Offenses than Rush Offenses. Watching Pryor and Sanchez throw would seem to be a refutation of this observation. Here's the information I think you should have presented: Close Off and Def S&P+, Close Rush Off and Def and Pass Off and Def S&P+, Non-Close Off and Def S&P+ (so we can see what teams are good at blowouts).
I debated whether to include close-game Rushing/Passing S&P+ or overall Rushing/Passing S&P+. In the end I went with the overall numbers, though looking back at it, close-game stats might have made more sense. And while the goal was to focus last week's column completely on per-play figures, it definitely provided you with some context-free data. Though Ohio State's passing game was relatively efficient, and it ranked high in Big Ten conference play, that does not mean they should start passing 40 times per game -- as is clear to anybody who watched them. Terrelle Pryor was not meant for a high-volume passing game, but the way Ohio State uses the pass works very well for them -- better than the way anybody else in the conference uses theirs. In the future, we will look at both per-play and per-game "+" numbers, and context will begin to emerge.
And on that theme...
What we need is a metric that combines effectiveness with volume to truly have a comparison of best passing, best rushing, etc. A baseball analogy would be VORP or WARP I guess.
I agree completely. As I mentioned, I do have per-game "+" rankings too, and they are equally as important as the per-play numbers that we looked at last week. The goal with last week's column, though, was to focus on per-play figures and what they can tell us about teams in conference play. One of these days, we will pull it all together.
2 comments, Last at 17 Dec 2008, 4:00am by John