clock menu more-arrow no yes

Filed under:

Observations and Questions: Oregon’s Rushing Offense

New, 21 comments

Investigating how the bread gets buttered

NCAA Football: Redbox Bowl-Michigan State vs Oregon Stan Szeto-USA TODAY Sports

Last month I investigated Oregon’s 2018 offensive playbook and went through how the playcalling fit within the gameplan. That article also laid out how I’m collecting the data for this series and the restrictions I’m placing on the dataset, including why I’m ignoring Oregon’s four FCS-caliber opponents. The Ducks’ tendencies made clear that it’s a run-first offense predicated on efficiency rushing that sets up manageable downs and distances. While that aspect of it looks to be pretty successful, many Oregon fans have noted that the run game doesn’t seem to produce the huge yardage from previous seasons in the last decade, and the purpose of this article is to look into how play selection, formation, personnel, and other factors might play into that.

By a wide margin, the variable that best predicted Oregon’s rushing success rate is not the ballcarrier, formation, play type, blocking scheme, direction, hashmark, drive count, down and distance, or even the quality of the opposing defense, it’s the week the game took place.

The story of Oregon’s rushing offense in 2018 turns out to be fairly straightforward, although I had to break down dozens of bi- and trivariate correlations to isolate it with confidence. There are three distinct phases to the 2018 season (four if you count the opening non-conference games, and I don’t):

  • Phase 1 - weeks 4, 6, & 7: Rush success 62.00%, Rush selection 54.35%
  • Phase 2 - weeks 8 & 9: Rush success 33.33%, Rush selection 30.00%
  • Phase 3 - weeks 10, 11, 12 & bowl: Rush success 51.43%, Rush selection 44.49%
NCAA Football: Portland State at Oregon Troy Wayrynen-USA TODAY Sports

Readers will recall from my January investigation into the offensive line that these times match up with the injuries to #58 LT Sewell and #75 RG Warmack. It really is as simple as that: to start the year (against very good defenses - #43, #13, and #4 in S&P+), Oregon ran the ball both dominantly and predominantly. Then, two o-line starters went down with injuries, and for a couple of games the run game suffered immensely; to their credit, the coaches cut down on rushing plays to match. Finally, though this phase is kind of weird, the team settled on a balanced attack, figured out replacement linemen, and stabilized their run game.

No other hypothesis that I tested comes anywhere close to explaining Oregon’s rushing numbers than the offensive line issues. While there are a few interesting wrinkles in other variables, most fan theories I’ve come across pointing to other factors in the Ducks’ somewhat disappointing rushing output fail to account for how good the team was prior to those injuries.

Before we move on to examining those other variables, I should explain what makes Phase 3 weird. First, while the rushing success rate against UCLA in week 10 clearly differentiates it from the disastrous outings in the prior two weeks, the gameplan against the Bruins more closely resembles a Phase 1 game in both rush vs pass selection and some other variables we’re about to look at.

Second, while I’m excluding from the dataset the final game of the regular season -- week 13 against OSU, as that team had an FCS-caliber defense -- the formation selection and other factors in that game deviated significantly from the rest of the Phase 3 games and made clear what the coaches’ preferred gameplan is when they enjoy an overwhelming advantage in the trenches, and as such helps to illuminate other more competitive games.

Third, the bowl game features two outliers: Sewell rejoined the lineup (though I think they rotated suboptimally on the right side of the line, as I argued back in January), but they also faced the #1 rush defense in the nation both in S&P+ and raw stats, an order of magnitude better ranking than any other opponent in 2018. I believe those two things cancel each other out, but that’s more of my subjective opinion than an objective conclusion from the numbers.

Running Backs

NCAA Football: Arizona State at Oregon Troy Wayrynen-USA TODAY Sports

Both #34 RB Verdell and #26 RB Dye had remarkably similar success rates -- both a couple points over 50% -- that stay the same comparing against almost every other factor, including direction, blocking scheme (man vs zone), and play type. Both of them go as their offensive line goes, with neither doing better than the other during the different phases. Verdell got about five-eighths of the carries in this sample size, which looks good to me – there’s not a significant dropoff in performance to justify giving the backup Dye fewer carries. The only thing that’s notable at all is that Dye appears to do better operating out of an offset formation rather than the pistol, but the sample size is small enough on that question that it’s only suggestive … just a couple plays going the other way would make the effect disappear.

I was a little surprised to see that both #10 QB Herbert and #20 RB Brooks-James had so few designed carries during competitive games that I couldn’t draw any decent conclusions from the small sample. Herbert is a pretty good scrambler when the pocket breaks down but I classify these as passing plays, one of the reasons I prefer using my own charts rather than the NCAA or other professional datasets. It became clear early in the season that the new playbook eliminated QB runs up the middle to protect Herbert’s health and opposing defenses in conference play almost universally stayed on the QB during read-option plays (which do constitute a very large portion of Oregon’s runs), almost always forcing the handoff.

Brooks-James’ season has an unhappier explanation: injury and ball security issues relegated him to special teams for most of his senior year. However, his excellent success rate (north of 75%) when he did actually get the ball suggests that ballcarriers other than power backs do have a place in this offense going forward.

Blocking Scheme

Oregon v California Photo by Ezra Shaw/Getty Images

Oregon ran behind power blocking about 60% of the time over the course of these nine games. Interestingly, there is no correlation at all between formation (pistol vs offset) and blocking – it’s 60% power blocking no mater how they line up. Instead, the only strong correlation is with time: during Phase 1, the Ducks are pretty evenly split between power and zone; by Phase 3, zone blocking has climbed to about two-thirds of rushes.

In my opinion, this is the one aspect of Oregon’s rushing game that I think the coaches deserve criticism for poor optimization: power blocking remains successful at more than a 65% rate during both Phase 1 and 3; however, while zone blocking is 60% successful in Phase 1, it falls to only 42% in Phase 3. That’s a mismatch between selection and effectiveness in that period, though I don’t have a good explanation for why that is. Neither the data nor my film study make this any clearer other than to guess that Oregon’s zone blocking plays require more “gelling” and are harder for a shifting, injury-plagued line to execute. But if that’s so then coaches should have stuck with straightforward power blocking at least half the time.


NCAA Football: Oregon Spring Game Scott Olmos-USA TODAY Sports

Oregon only lined up in a set other than a single back (empty, Pro set, and that’s it) about 4% of the time, too few to meaningfully evaluate. On single-back rushing plays, the Ducks snapped the ball in the pistol formation about 72% of the time over these nine competitive games (the four games I excluded had very different numbers, but as explained elsewhere the coaches clearly knew they had those games in the bag and gameplanned differently). The remaining 28% were split between plays where the back is positioned offset of the QB for at least three seconds prior to the snap (20%) and plays where the back motions from pistol to offset just before the snap (8%). I could find no variable that performed significantly differently between always-offset and motion-to-offset, so for ease of discussion I will simply refer to all such plays as offset.

The puzzling thing about the predominance of pistol rushing is that offset rushing consistently performs about 10 percentage points better, regardless of other variables that I compared formation to. However, before detractors claim this is an open-and-shut case of a bad formation and poor playcalling, the picture that the rest of the data paints is that this is still a by-product of the shifting gameplan as a result of o-line injuries. Through week 10 (so Phase 1, Phase 2, and the first game of Phase 3), Oregon rushed more than 85% of the time out of the pistol. For the rest of the year (excluding OSU), the coaches shifted the gameplan radically to rush out of the pistol only 43% of the time. Furthermore, there is no other significant correlation (including a trivariate comparisons using time) between formation and any other variable – that is to say, Oregon lining up in the pistol vs offset does not signal whether the rush will be inside or outside, read-option or straight handoff, power- or zone-blocking, complex or simple, or anything else.

Instead, the flip in frequency between the UCLA and Utah games, combined with Oregon’s dominant performance in Phase 1 and their return to the pistol over 80% of the time against OSU, indicates that:

  1. The pistol is Oregon’s preferred rushing formation,
  2. When they have o-line superiority they use it very well,
  3. It sets up the offset, under ideal o-line conditions, as a change-of-pace that works even better because of its rarity,
  4. When forced into suboptimal o-line rotations, the offset takes over a majority of plays, and
  5. Offset-majority games are probably the best available option in those circumstances, but still worse than healthy o-line, pistol-predominant games.

In other words, the coaches came up with a Plan B to deal with offensive line injuries, which in addition to the rise in creative screen passes incorporated a whole lot more offset and zone rushing. Plan B was mostly successful, and although I still have some complaints about the line rotations and zone blocking effectiveness, ultimately I think it worked as a stopgap that produced three wins in the final four competitive games. The relative success of that solution, however, doesn’t indict the choices that went into Plan A -- that is, dominant power rushing featuring the highly efficient halfback dive out of the pistol -- because Plan A was still more successful, and against better defenses.


  1. What do fans think of the increase in zone blocking after the offensive line injuries? My best guess is that this is trying to work out of the old playbook as “comfort food” to the backups in the lineup, who after all were recruited to play under Steve Greatwood’s system. Could this factor have an impact on the offseason battle for right guard?
  2. Since any explosion play has to first meet the yardage-gain threshold to be a successful play, I generally use efficiency as a simple proxy and just assume a certain percentage of successful plays will break big based on the stars aligning (usually the safety missing a tackle). Oregon did get several explosive runs against Stanford and Cal in Phase 1, then UCLA and Oregon St in Phase 3, but they were absent for much of the year. I think I know why that happened against the Huskies -- they had a phenomenal corps of safeties who flew out of the defensive backfield to trip up several nearly explosive plays just as they were getting going -- but I don’t have a great explanation for why we didn’t see many against teams like Arizona St. What do fans think of that issue - bad luck, freshman backs, defenses having adapted since the heyday of the option offense, or something else?
  3. I’ve got over a hundred contingency tables comparing every variable I collected in my game charts, but most of them don’t show any significant correlation, and I can’t write an article listing everything I looked at but came up with no relationship or insufficient data. But I’m happy to answer any specific question fans have about the data I’ve gathered - just ask in comments.