Including Sonnyball won't be favorable for defensive stats, but will be bizarre. Wilcox leaving, and Rolovich taking over, won't change that either.
Regardless of who's the coach, Cal is still Cal as far as facilities and campus and community challenges go, and a Cal depth chart is what it is (has been thru multiple coaches), a recipe for a collection of bizarre rockfights, with the outcome dependent on whose' pillowcase has the most and biggest rocks in it.
This analytical series has been one of the most informative series I've read anywhere about any team, and eagerly look forward to seeing what the numbers and a Berkelium powered interpretation say about whatever comes along after Wilcox. (I'd love to see this applied to several other teams of note for reference.)
I appreciate the praise. This series could be replicated for any other team and posted on a weekly basis: it takes under 5 minutes for me to pull the most recent game's data, run the classifier, and generate the plots. The main challenge is the up-front costs: 1) pulling all the historic PFF data from a team and the bigger challenge of 2) having the domain knowledge to know how many categories to include and what those categories represent. Most people reading this site follow Cal closely enough to know what a typical Wilcox Rockfight looks like, but I have less of a sense of what kinds of categories would commonly apply to games played by a team like Oregon, which I don't actively follow. I could certainly talk about the patterns in each cluster and how they differ across clusters, but that deep domain knowledge makes it much easier to produce this series regularly and allow it to evolve over time (such as including new categories of games, like when Cal started getting involved in Pillowfights in the '22-'23 seasons).
I think for the Dykes games, you might have to include some type of game timeline, as most of Sonny's games have 2H offensive stats that are inflated due to garbage time and playing against the opponent's 2nd and 3rd strings. It seemed the GBs were down by 28 points or more by the start of the 3rd quarter, and our opponents went to prevent defenses for the whole second half!!!
Although it would complicate the analysis, I wish PFF had data at finer time resolution. Quarter-by-quarter data would be fantastic and provide deeper insight into how the team's performance changes over the game (like this week, where the first half was a defensive slugfest before it all fell apart in the second half). That would have been very useful for characterizing Wilcox's turtling teams in previous seasons and would be equally useful for those Bear Raid games where a flurry of late-game points made the final score look a bit more respectable.
Fortunately, PFF claims their grades account for garbage time, so the evaluations largely cover the game when the outcome is not already decided. How exactly they weight player evaluations on a snap-by-snap basis to account for garbage time is not clear to me (and may stray too close to revealing their secret sauce in their evaluation process).
Good points. I'm not sure PFF can account for many of the Sonny Dykes games where the whole 2nd half was garbage time - and as you say the "Bear Raid" was scoring points just to achieve respectability. The Holmoecaust and the Dykes Disaster were dark days for the program, let's hope our next coach can achieve more than mediocrity...
Including Sonnyball won't be favorable for defensive stats, but will be bizarre. Wilcox leaving, and Rolovich taking over, won't change that either.
Regardless of who's the coach, Cal is still Cal as far as facilities and campus and community challenges go, and a Cal depth chart is what it is (has been thru multiple coaches), a recipe for a collection of bizarre rockfights, with the outcome dependent on whose' pillowcase has the most and biggest rocks in it.
This analytical series has been one of the most informative series I've read anywhere about any team, and eagerly look forward to seeing what the numbers and a Berkelium powered interpretation say about whatever comes along after Wilcox. (I'd love to see this applied to several other teams of note for reference.)
I appreciate the praise. This series could be replicated for any other team and posted on a weekly basis: it takes under 5 minutes for me to pull the most recent game's data, run the classifier, and generate the plots. The main challenge is the up-front costs: 1) pulling all the historic PFF data from a team and the bigger challenge of 2) having the domain knowledge to know how many categories to include and what those categories represent. Most people reading this site follow Cal closely enough to know what a typical Wilcox Rockfight looks like, but I have less of a sense of what kinds of categories would commonly apply to games played by a team like Oregon, which I don't actively follow. I could certainly talk about the patterns in each cluster and how they differ across clusters, but that deep domain knowledge makes it much easier to produce this series regularly and allow it to evolve over time (such as including new categories of games, like when Cal started getting involved in Pillowfights in the '22-'23 seasons).
It's still 5 minutes so many others would wave their hands at... QED-style...
Not sure if I asked this before, but where are turnovers and penalties captured?
Those are captured in the individual-level grades and then represented in the team-level aggregated grades.
"With Wilcox’s departure, so too I will let our the Good-Bad-Rockfights-Pillowfights-BIZARRO categories depart."
Shedding a tear for ^^ THIS ^^
I think for the Dykes games, you might have to include some type of game timeline, as most of Sonny's games have 2H offensive stats that are inflated due to garbage time and playing against the opponent's 2nd and 3rd strings. It seemed the GBs were down by 28 points or more by the start of the 3rd quarter, and our opponents went to prevent defenses for the whole second half!!!
Although it would complicate the analysis, I wish PFF had data at finer time resolution. Quarter-by-quarter data would be fantastic and provide deeper insight into how the team's performance changes over the game (like this week, where the first half was a defensive slugfest before it all fell apart in the second half). That would have been very useful for characterizing Wilcox's turtling teams in previous seasons and would be equally useful for those Bear Raid games where a flurry of late-game points made the final score look a bit more respectable.
Fortunately, PFF claims their grades account for garbage time, so the evaluations largely cover the game when the outcome is not already decided. How exactly they weight player evaluations on a snap-by-snap basis to account for garbage time is not clear to me (and may stray too close to revealing their secret sauce in their evaluation process).
Good points. I'm not sure PFF can account for many of the Sonny Dykes games where the whole 2nd half was garbage time - and as you say the "Bear Raid" was scoring points just to achieve respectability. The Holmoecaust and the Dykes Disaster were dark days for the program, let's hope our next coach can achieve more than mediocrity...