The Hogs may be off this week, but the stats are still flowing.
Arkansas enters the open date at 2-3 (0-2 SEC), but at least the first five games ended on a (relative) high note, with a strong performance against Texas A&M that at least suggested that the sky is not falling.
There will be a dedicated Kentucky preview next Thursday as usual, but it’s obviously worth noting that this game is huge if the Hogs are going to win an SEC game. Kentucky is the worst SEC opponent on Arkansas’ schedule, and the Wildcats are only slightly better than November non-conference foe Western Kentucky (73rd vs. 77th in our rankings).
Ranking the SEC
In the Texas A&M preview and box score breakdown, we debuted the concept of “adjusted points.” Basically, since our EV/EVA system already allows us to figure out where every point a team scores comes from, all we have to do is opponent-adjust as many numbers as possible and we can determine how many points a single team would score and allow against an average FBS team. Subtract Adjusted Points Allowed from Adjusted Points and you get Adjusted Scoring Margin, which is, by itself, a single measure for ranking all teams.
Here, for example, is how the Hogs shape up.
So we say that Arkansas is the 85th-best FBS team, based on this measure. That’s spectacularly close to where the major college football analytical systems rank the Hogs: Bill Connelly’s SP+ system has the Hogs at 82nd, ESPN’s FPI ranks the Hogs 80th, and Jeff Sagarin’s rankings (which include FCS teams) put the Hogs at 91st.
So I think this system that mostly exists on my laptop is pretty accurate, but I decided to take it a step further, building a score predictor that uses nothing but the adjusted points numbers generated by the system. Scores are calculated using a simple deviation from the mean formula. I picked 41 games against the Las Vegas spread last week and went 25-16 (61%). If that holds up this week when I pick 48 more games, I’m moving to Vegas and you’ll never hear from me again.
Now that I’ve (hopefully) established the credibility of these homemade rankings, let’s rank the SEC:
- Alabama (+25.0)
- Georgia (+24.8)
- Auburn (+21.8)
- LSU (+20.7)
- Missouri (+12.2)
- Florida (+11.9)
- Texas A&M (+11.4)
- Ole Miss (+8.0)
- Mississippi State (+7.1)
- South Carolina (+4.4)
- Kentucky (-3.0)
- Arkansas (-6.7)
- Vanderbilt (-7.6)
- Tennessee (-14.1)
Tennessee is trailing the field by nearly a full touchdown, and the Hogs jumped Vanderbilt, who looked unimpressive in a 24-18 win over hapless Northern Illinois on Saturday. Playing well and losing a good team is much better than playing poorly and barely beating a bad team.
Based on this math, Alabama is 32 points better than Arkansas on a neutral field. I add three points for homefield advantage, so I’d say Vegas is probably looking at Alabama -35 when the Hogs head to Tuscaloosa in a few weeks, unless the Hogs can get their numbers up before then.
Let’s back up for just a second...
I’ve gotten a ton of confused comments about the EV/EVA system, so consider this my best and final effort to explain how it all works. If you’re a long-time reader of Arkansas Fight, you’ll know that I’ve been working on this since 2015, so I get that it can be confusing, as it took me years to understand its full potential.
Q: What’s the difference between EV and EVA?
A: Expected Value (EV) is the number of points you can expect to score based on the current down, distance, and yardline. So if you have 1st and 10 at your own 20, you can expect to score 1.60 points, so 1.60 is your starting EV for that play. Every play has a starting EV and an ending EV; basically, your expected points before the play, and your expected points after the play. The change in EV — that is, the change in your expected points — is called Expected Value Added (EVA). So EVA is the actual value created by a single play. Touchdowns may be scored on the last play of a drive, but every play in that drive contributed in some way, and we can use EVA to calculate the “value” of each one.
So EV is mostly used as a baseline to calculate EVA, which is more important. When we do analysis, we use EV to calculate the value of starting field position, but then we’ll use EVA for every play in the drive. Runs and passes are calculated in terms of EVA. Penalties are in terms of EVA: the EV difference between 1st and 10 at the 35 and 1st and 15 at the 30 is Penalty EVA. Returns are in terms of EVA: if a return man fields a punt at the 12 and returns it to the 26, then the difference between the EV of 1st and 10 at the 12 and 1st and 10 at the 26 is given as Punt Return EVA, because his return “created” that value.
Q: What’s an example of how a drive is scored in terms of EV/EVA?
A: Some people understand better by seeing it in action. Let’s take Arkansas’ opening drive against Portland State, a nine-play drive that ended in a field goal. The play text is straight from the official NCAA play-by-play.
- Cody Williams kickoff for 51 yds , De’Vion Warren return for 16 yds to the Ark 30
On kickoffs, Field Position EV is assigned at the 25 (the value of a touchback). If the kick returner can get his team better position by returning the kick, he’ll get the difference. So 1.85 EV is given as Field Position EV, but the value of 1st and 10 at the 30 is 2.10, so Warren gets 0.25 Kick Return EVA for his efforts.
- (1st and 10 at Ark 30) Ben Hicks pass incomplete to Mike Woods, broken up by Jake Porter
The starting EV of this play is 2.10 (1st and 10 at the 30) and the ending EV is 1.58 (2nd and 10 at the 30), so both Hicks and Woods are charged with -0.52 Passing EVA for their failure to connect.
- (2nd and 10 at Ark 30) Rakeem Boyd run for 6 yds to the Ark 36
The starting EV is 1.58 and the ending EV is 1.66, so Boyd gets +0.08 Rushing EVA for helping the offense into a third-and-short.
- (3rd and 4 at the Ark 36) Ben Hicks pass complete to Treylon Burks for 12 yds to the Ark 48 for a 1ST down
Third-down plays are big, because punting drops your ending EV to zero, so the third-play is charged with the loss. Here the Hogs’ EV goes from 1.66 to 2.99 with the conversion, giving Hicks and Burks +1.33 Passing EVA.
Let’s fast-forward to the end of the drive....
- (3rd and 6 from the PSU 16) Ben Hicks pass incomplete, broken up by Anthony Adams
The starting EV was 3.66. If the Hogs punted, ending EV would be zero and this play would be charged -3.66 EVA. But the Hogs didn’t punt. They kicked a field goal. The value of field goals is based on their distance: this 33-yarder allows the offense to “retain” 2.5 EV, so this play has an ending EV of 2.5. That means this play was worth -1.16 EVA.
- (4th and 6 from the PSU 16) Connor Limpert 33-yard field goal Good
The starting EV was 2.5, and the play ended with a made field goal, worth three points (3.0 EV). So Limpert gets +0.50 Placekicking EVA. Had he missed it, ending EV would be zero, so he’d be charged -2.50 Placekicking EVA.
Here’s how the entire drive is scored:
- 1.85 Field Position EV
- +0.25 Kick Return EVA
- -0.52 Passing EVA
- +0.08 Rushing EVA
- +1.32 Passing EVA
- +0.59 Rushing EVA
- +0.99 Passing EVA
- -0.73 Passing EVA
- -0.18 Rushing EVA
- -1.16 Passing EVA
- +0.50 Placekicking EVA
Add all of that up, and you’ll get three points. If we aggregate the numbers, we get 1.85 points from field position, 0.65 points from offense, and 0.50 points from placekicking.
Q: What is PAN, and how do we opponent-adjust numbers?
A: Points Above Normal (PAN) is the value created above what the opponent normally allows. It’s an opponent adjustment for EVA. So if Kentucky is allowing +0.02 EVA per Rush, and Arkansas collects +0.05 EVA per Rush and 40 rushes, then Arkansas had an EVA advantage of +0.03 across 40 rushes, which comes to 1.2 PAN. That means that Arkansas’ run game generated 1.2 more points than what Kentucky’s defense normally allows.
Here’s a useful comparison:
- +4 Rush EVA means that the run game generated 4 points above the FBS average
- +4 Rush PAN means that the run game generated 4 points above your opponents’ average
If you have a difficult schedule, then PAN > EVA, because it is easier to get positive numbers against a tough defenses. If I told you Arkansas had +0.03 EVA per Rush in a game, would you be impressed? If it was against Alabama, then of course you would be! That would be a really high PAN. If it was against UTEP, then not so much. It’s hard to generate PAN against bad defenses because they allow so much anyway, so for team with weak schedules, EVA > PAN. Because of this, PAN is the best measure of a team’s offense and defense. The adjusted points formula (below) takes other things like special teams, field position, and defensive scores into account.
Other stats we use are similarly opponent-adjusted. We can opponent-adjust success rate (percentage of plays that go for positive EVA) by comparing it to the average to get marginal efficiency. If Arkansas gets 44% success rate and Alabama’s defense was allowing 40%, then Arkansas got +10% marginal efficiency in that game (44% is 10% better than 40%). Any stats you see labeled with the term “marginal” have been opponent-adjusted.
Q: How exactly are adjusted points per game calculated?
A: Recall that we calculate the actual points scored in a given game by doing the following math: start with Field Position EV, add EVA generated by Rushing and Passing plays, add EVA generated by penalties, add EVA generated by punt and kick returns, add EVA generated by placekicking, and then add EVA generated by defensive returns. That will always add up to the total number of points scored.
So here’s how adjusted points per game is calculated:
- Start with: Average Field Position EV per Game (inclusive of defensive returns)
- Add: Offense PAN per Game (this is opponent-adjusted EVA on runs and passes)
- Add: Average Penalty EVA per Game
- Add: Average Placekicking EVA per Game
- Add: Average Punt & Kick Return EVA per Game
And voilà, now you have adjusted points per game. All I really did to create this number is start with actual points per game, break that up into play types, and then opponent-adjust the run and pass plays. This formula isn’t really perfected yet: I still need to find a way to opponent-adjust field position and punt/kick returns. I’m using the raw numbers for now.
Q: What about the other stats?
A: Other stats like line yards, leverage rate, and available yards are simpler to explain and help provide context and details to understand what the EV/EVA/PAN/AdjPoints numbers are telling us. I link the glossary in every post that includes advanced stats. If you see an unfamiliar term, check the glossary.
Have any more questions about the EV/EVA system? Drop a comment!
So we see that the Hogs are 85th in adjusted scoring margin. Let’s dive in a little deeper to see how things look, starting with drive-level data.
The Hogs’ offense can move the ball just fine, ranking 45th in available yards, gaining half of all possible yards per game.. But finishing drives is one of the most important abilities in football, and the Hogs simply haven’t mastered it yet. Many drives enter opposing territory, but the Hogs are barely getting a field goal on average for every drive that includes a 1st and 10 inside the 40. In FBS play, eight drives have ended with the offense on the field but no points (four turnovers on downs, two fumbles, two interceptions) and two others have ended with a missed field goal.
The fact that the Hogs have won the turnover battle in four of five games, had the same or more scoring chances in all five, and won available yards in three of five... and are only 2-3 suggests that maybe the Hogs are closer to winning that we thought after the San José State debacle. Or maybe, like we saw with Bret Bielema, the Hogs will go on being competitive but never actually break through. It’s probably still too early to tell.
Up next we evaluate special teams and field position.
Arkansas’ marginal offensive field position is so high because defensive touchdowns are counted as “field position touchdowns” and 6.95 Field Position EV is assigned every time the defense scores. That’s because the defense quite literally allows the offense to start in the endzone, with no need for the offense to do anything. That’s what we’re trying to evaluate here: what kind of position is the offense (or defense) starting out in?
Arkansas’ high available yards number on offense means that, unlike last year, the offense rarely leaves the defense out to dry. The Hogs have jumped from 123rd to 69th in marginal defensive field position because the offense rarely punts from deep in its own territory. Punter Sam Loy hasn’t been great (38.6 yards per punt, 92nd FBS) but several of his punts have gone deep into opposing territory because of where he was punting from.
Kick returns is something I expect the Hogs to do better on. De’Vion Warren has been great for two years but hasn’t gotten much going so far in 2019, and he’s been injured for the last couple of games. Treylon Burks, on the other hand, has been fantastic returning punts.
Special Teams/Field Position Strengths:
- Punt returns
Things to Work On:
- Kick returns
These numbers tell us a few things:
- Arkansas’ offense — just 100th in PAN — has to be more productive during the second half of the season. We’ll do a run/pass breakdown in a moment.
- The Hogs are one of the most pass-heavy teams in the country, and probably the most pass-heavy team in the SEC. On passing downs, the Hogs run just 18% of the time.
- Arkansas’ standard-downs offense is really bad. It isn’t efficient and it isn’t explosive. Arkansas has to get more out of it.
- On passing downs, the Hogs are much better. Big plays come on passing downs, as the Hogs lead the SEC in passing downs explosiveness. That passing downs success rate number could probably stand to be a little bit higher.
Here’s the problem with the standard downs offense:
The Razorback run game is going backwards (often literally). The Hogs lose about four points per game when running the football. You can see that the run game is dreadfully inefficient, coming in at -16% marginal efficiency.
Much of this appears to be the fault of the offensive line, which is once again having problems. Power success rate, stuff rate, line yards, and opportunity are pure offensive line stats, and the Hogs rank 116th, 123rd, 108th, and 109th at those. Bonus yards are the only rushing stat unrelated to the offensive line... and that’s the only thing Arkansas does well. Both Rakeem Boyd and Devwah Whaley do a nice job of turning good runs into great runs. The Hogs have to find a way to become more efficient on the ground.
Interceptions are the cause of the marginal ANY/A numbers being low, but other than that, the Hogs’ passing attack is much improved. The line, struggling in run blocking, has improved its pass protection, particularly last week against Texas A&M. Both Ben Hicks and Nick Starkel do a nice job of sensing and avoiding pressure. There’s still a lot to work on: too many passes are dropped, and I’m not sure the Hogs are fully exploiting some of the advantages available in the middle of the field with Cheyenne O’Grady. Of course, the Hogs’ full trio of starting wide receivers (Trey Knox, Treylon Burks, Mike Woods) hasn’t been together for two full games, so it will be interesting to see what the Hogs try against Kentucky.
Here’s a look at some individual breakdowns of Hog runners and pass-catchers:
- Explosive pass plays
- Turning big runs into bigger runs
Things to Work On:
- Run blocking, particularly in terms short yardage and avoiding losses
The defense does a decent job of limiting big plays: Texas A&M in particular had a hard time getting anything down the field, and Ole Miss only got big plays late in the game. The Hogs’ main problem has been knocking opponents off schedule: 68% of opponent snaps are in standard downs. Arkansas’ marginal third down defense has also cratered from 42nd last year to 119th so far this year. I think the problem is that Arkansas simply cannot force incomplete passes. I’ll show you why as we go over these numbers.
The defensive line isn’t getting enough credit and that ends now. The Hogs’ defensive front has done a good job of stuffing the run and stopping short-yardage runs. The big problem is that the back end keeps giving up big runs. Hopefully that’s under control now, but Ole Miss and Colorado State gashed the Hogs with big run plays.
You may be wondering why Arkansas’ run defense numbers overall look decent given the gashing they took from Ole Miss and Colorado State. Well, we now have to review some of what I wrote back in the Colorado State preview:
Colorado State’s run game is quite bad. They lost last year’s leading rusher and through two games have had serious trouble generating explosive runs. They’ll use a lot of misdirection, including jet sweeps, to try and find something.
On the other hand, their inability to run the ball means that the Hogs can make them one-dimensional and break them down with interceptions or pressure.
Okay, so that was definitely true for last year, but it’s not true for this year. Colorado State is actually 21st in the FBS in Rushing PAN, including 4th in Rushing Marginal Explosiveness, so giving up 220 rushing yards to them no longer looks bad. Marvin Kinsey Jr., who rushed for 180 yards against the Hogs, including a 75-yarder on the second play from scrimmage, is currently 4th in the FBS in rushing yards by an individual.
The Hogs harassed Texas A&M’s Kellen Mond last week to the tune of five sacks. The Hogs have now faced FBS offenses ranked 11th, 27th, 37th, and 44th in marginal sack rate and still managed to generate decent pressure. Once again, the defensive line isn’t getting enough credit and that stops today.
Unfortunately, if the Hogs don’t sack the quarterback, he’s probably going to complete a pass for a nice gain. You don’t often see a team that is the best at getting sacks in the SEC but still the worst at stopping the pass overall. Arkansas’ very young secondary has been victimized pretty badly this year.
- Pressure on the QB
- Defensive touchdowns
Things to Work On
- Limiting high-efficiency passes
- Stopping breakaway runs
We’re hopefully not done with open date stats fun. I’ve applied the EV system all the way back to 2008 (Petrino’s first year) and am working on a piece comparing Arkansas football from 2008 to 2019. Look for that early next week.
For now, let’s simulate the rest of the 2019 season based on current data:
- Kentucky 29, Arkansas 22
- Auburn 41, Arkansas 15
- Alabama 49, Arkansas 15
- LSU 53, Arkansas 22
- Arkansas 23, Western Kentucky 23
- Mississippi State 34, Arkansas 23
- Missouri 39, Arkansas 23
If the Hogs want to do better than 2-10 or 3-9, there’s still a lot of improvement left to be made. Even WKU isn’t a pushover: the Hilltoppers are probably better than either Colorado State or San José State, and Ty Storey has them rolling after throwing for 189 yards and two touchdowns in an upset win over UAB last week.