The Good, The Bad, and The Rock Fights: Cal Football Mid-Season Review

What does the 2023 team's overall performance profile look like, and how does it compare to recent Cal teams?

Photo Credit: Rob Hwang

Due to the bye week we will take a different approach this week. For those unfamiliar with this series (which may be a decent number of you, as this week’s edition is not behind the paywall), we use Pro Football Focus data to categorize the most recent Cal game. PFF grades twelve distinct categories across the team (e.g. passing, tackling, pass blocking, etc.) and we use those grades to determine what kind of game it was: Good (grades are better than usual across the board), Bad (grades are worse than usual across the board), a Rockfight (defense earns excellent grades while offense usually struggles except for running), or a Pillowfight (pass offense and/or running carry the team while all aspects of the defense earn sub-par grades).

Rather than looking at a single game, this week I have taken the average grades across all seven games so far this season to see how the whole season fits into our usual categories. For comparison, I also took the averages across each of the previous seasons under Wilcox (except the bizarro 2020 season) to see how those seasons fit into our four buckets. Before we feed the results into our machine learning classifier to see what it spits out, let’s take a look at those season averages.

PFF Season Grades

Below I have created a boxplot to show the variation in each of the game-level grades across the Wilcox Era. I have superimposed the season-level averages as colored dots on top of those boxplots. Each color represents a different year.

Each season’s color-coded average is superimposed over boxplots capturing the range of grades in each category. The blue box captures the middle 50% of the data and the horizontal line represents the median, or midpoint. Peripheral data falls into the blue vertical lines and single blue dots represent extreme outliers. This caption is officially too long now.

Looking at the overall score on the far left, the 2017 and 2018 teams are in a near-tie at the top, followed by the 2021 team. Interestingly, the current 2023 team’s average overall grade sits exactly at the midpoint of all these overall grades. 2019 is just below that; despite the best record of all the seasons, the midseason losing streak during Garbers’ injury featured a distinct decline in overall scores that brought down the average. 2022 fared worst, sitting near the 25th percentile.

Offensively, the Musgrave-era teams (2021-22) have earned better scores at the skill positions (passing, receiving, and running) while the Baldwin-era offenses (2017-19) had better O-line grades (pass protection and run blocking). The 2023 team aligns more with the Musgrave-era teams due to stronger performances from the skill positions and weaker performances from the O-line. With three different starting quarterbacks, the passing grades have been all over the place this season. If Nando keeps playing well as the starter, 2023’s average passing grade should climb over the remaining games.

The average defensive grades show a big difference between the Tim DeRuyter era and the Peter Sirmon era. TDR’s defenses had the highest average defense, run defense, and tackling grades. The 2021 and ‘22 teams saw notable steps back in those categories. Pass rush has always been a challenge for Wilcox’s defenses (it’s usually the lowest-graded defensive score) and there has been no notable trend from year to year. Interestingly, the 2018 team has the best coverage by far, while the rest of the teams all have average scores below the median. Only the 2018 team consistently achieved excellent pass coverage; while other seasons had some excellent individual performances, there were enough below-average performances that the aggregate score dropped when we averaged across the whole season. Other than tackling, the 2023 defense’s grades have been all over the place this season. They’ve had some strong performances (Auburn, 2nd halves of UNT and Idaho) and some awful performances (UW, OSU), but most categories have fared worse than usual this year.

There’s enough variation in these average 2023 scores that I did not have a strong guess as to where the classifier would put the 2023 team. It’s neither been a uniformly better-than-usual nor a worse-than-usual season. It should not be a rockfight, as the offense has been too good and the defense has been too bad. And yet the offense has not been consistently good enough nor the defense consistently bad enough to qualify for a pillowfight. So where will 2023 end up?

Cluster Analysis

I fed the season-average results into our classifier and it placed the 2023 team in…

Actually, before we get to the result, a quick description of the plot below. The figure visualizes all the data points that go into the cluster analysis, along with how the clustering algorithm sorts them together. Points closer together are more similar, while those farther apart are more different. The x- and y-axes don’t map on to any specific variable; rather, they reflect the dimensions along which the data are most different. We only have two dimensions because the substack servers would melt if I tried to upload a three- or four-dimensional figure. Reading the tea leaves, the x-axis seems to capture something about defense: defensive performance seems to get worse in the further right parts of the chart (consider the 49-point explosion against OSU in 2018 on the far left edge of the plot in comparison to the 7-point faceplant against Oregon in 2019). The y-axis is a little more convoluted, as games higher up on the plot tend to have worse defense than games near the bottom—of course, there are many exceptions to this trend (like that 2019 Utah game, where the defense was one of many, many, many problems that day). Anyway, back to classifying the 2023 season…

(Click the image for a much larger version) A visualization of our four clusters of game types. Data points closer together are more similar while those farther apart are more dissimilar. The most coherent groups of data points are clustered together, with those boundaries showing which teams belong to which clusters. Each season-level average is highlighted in a color-coded box.

The Good? It does not really feel like a Good season, but upon further investigation the result makes some sense. 3 games this season belong in The Good (North Texas, Idaho, Arizona State), 1 belongs in Rockfights (Auburn), and 3 belong in Pillowfights (UW, Oregon State, Utah). So The Good and The Pillowfights would be the most natural home. In fact, 2023’s position in the plot above is on the edge of The Good, just across the border from The Pillowfights. With three of the last four games falling in The Pillowfights, the team is trending in a pillowfight direction. If Nando keeps leading a productive offense and the defense continues to struggle, we may see further movement into the Pillowfights by the end of the year. We’ll have to do this exercise again after the season ends to see where the 2023 finishes the year.

Previous Wilcox teams have finished in a variety of different positions. 2017 finished in The Good, while 2018 and 2019 were Rockfight years. 2021 moved back into The Good, but continued regression on defense (and the pleasant success of the Plummer-led offense) moved the 2022 team into The Pillowfights. There’s a good chance 2023 follows that same path, beginning with a likely Pillowfight against the woeful USC defense this weekend.