Musings from Arledge: Sports Analytics

This week’s Musings is the result of a major failure. And, no, I know what you’re thinking, but you’re wrong. Most Musings are the result of simple, run-of-the-mill failures. This one was bigger than usual.

I interviewed Dr. Jeremy Abramson a Research Computer Scientist in the Networking and Cybersecurity Division at USC’s Information Sciences Institute, and a lecturer in USC’s Data Science Program. Dr. Abramson studies, teaches, and writes about sports analytics, a field that has exploded in usefulness and importance over the last few decades. That interview was supposed to go on YouTube as a Musings from Arledge Solo Edition video. 

Because I’m not a graduate of the Information Sciences Institute – how my day job is done hasn’t changed much since Abe Lincoln’s day – I had the audio settings wrong, and in the recording, Dr. Abramson sounded great, but I could barely be heard. And while that might sound like a net benefit to many of you, I’m far too vain to consider such a final product. So I took the information from that interview and turned it into a written Musings instead.

For most casual fans, our first experience with sports analytics was Moneyball, where A’s general manager Billy Beane (as portrayed by Brad Pitt) shocked and horrified the old-school scouts and executives when he turned conventional wisdom on its head in search for an edge that would allow the small-market A’s to compete with the Yankees and Red Sox. The answer – as it so often is in the modern world – was a math geek. Sports analytics is, simply put, quantitative decision-making, gathering data and making decisions based on what the data tells you rather than your gut. (In Moneyball that meant, among other things, not making draft decisions based on the attractiveness of the player’s girlfriend, a critical component of the analysis according to one A’s scout.)

Dr. Abramson believes there are at least three major applications of analytics in college football: recruiting (which is essentially roster-building through numbers a la Billy Beane), in-game decision-making (such as when should we go for it on fourth down?), and injury-prevention, which Dr. Abramson believes is the great white whale of sports analytics.

We started the conversation with in-game decision-making. Even most casual fans have now heard that football coaches should go for it on fourth down more frequently than they do. But there are coaches and media figures that push back aggressively against the use of analytics in decision-making, largely because they associate analytics with aggressiveness. If you make a particularly aggressive or outside-the-box decision, that’s seen as analytics. And when it backfires – as all decisions sometimes do – the old guard rails against the foolishness of listening to the math geeks. They’ve never even played football! (Well, some probably have.)

But Dr. Abramson believes this is a mistake. The reason why analytics and aggression are synonymous in the minds of many is because human beings are risk-averse – we fear the possibility of loss more than we value the possibility of gain – and the irrationality of risk aversion can lead to overly conservative decision-making. As Dr. Abramson put it: “People often make bad decisions based on their risk aversion. Analytics can help us curb some of the biases people have.”

Bad decision-making may be especially common when a particular decision is so ingrained in the culture of a sport. No coach gets fired for punting on 4^th and 3 from the 50. But if you go for it in that situation and turn the ball over on downs, the howls of outrage from fans and the media can be deafening. But as the football community comes to understand analytics better, the culture can change. Once people know that the numbers are strongly in favor of going for it on fourth and short from midfield, the pressure may soon come from the other direction: coaches may soon have to justify punting rather than going for it in those situations.

I asked Dr. Abramson about some of the criticisms of analytics. Is it true, I asked, that analytics fails to take into account the critical human component – the psychology of players, for example? When a team goes for it on fourth down and gets it, I think that can be particularly demoralizing for a defense. We’re so used to having to stop teams for only three plays, that giving up a first down on fourth down can hurt more deeply, maybe. 

But as Dr. Abramson pointed out, if certain bad outcomes are particularly demoralizing, this should show up in the numbers. So while the analytical models cannot test human psychology directly, if giving up a first down on fourth down is particularly demoralizing for a defense, and if that demoralization truly has an impact on the game, it should show up in the numbers that follow such an event. A defense should be more likely to give up yards, first downs, and points after a successful fourth-down conversion than a defense that is not demoralized in that same fashion. 

What this means is that a good analytical model can capture intangibles, albeit indirectly. Even if you can’t test “heart” with the numbers – and Dr. Abramson thinks you can – you can certainly gather data on all the things that “heart” is supposed to lead to: turnovers, tackles, first downs, etc. 

Second, Dr. Abramson argued persuasively that models can be fine-tuned for your particular situation. I asked about whether going for it on fourth down depends not on the general figures for all teams but on the figures for your particular team. Does it change the calculus whether 2022 USC is on the other side of the field – a team with Caleb Williams and top-notch backs and receivers but a questionable defense – rather than 2022 Iowa, a pretty good defensive team that most weeks could not find the end zone with three guides and a map? The answer, of course, is yes; you can choose data that fits closer to your individual situation. In college football, the data sets are almost always big enough to do so.

And while there are things that models don’t take into account that can sometimes loom very large – the sun shines directly into the quarterback’s eyes at this section of the Coliseum at 4:00 p.m. on October 15 – these “one-off, unicorn situations” don’t invalidate the use of models. If a factor is truly unusual and therefore renders the models less useful, the fact that it is unusual means that it will not render the model less useful except in unusual circumstances.

When it comes to building a roster, analytics is also helping solve some of the oldest problems in scouting and recruiting; for example, what to do with 40 times. We all know that 40 times do provide some useful information. If a defensive back often clocks 4.9 in the 40 – Guilty! – there are probably some limits as to what that player can do on the field, regardless of his other skills. Covering Zach Branch man-to-man may be a stretch. 

At the same time, many of the greatest players in history ran mediocre 40 times. Jerry Rice’s 40 time wouldn’t wow anybody; yet he was only the best receiver (and maybe best football player) who ever lived. 

But here, again, analytics can help. Once tracking data becomes universal, you will be able to tell far more about what you really care about – fast-twitch ability, suddenness, etc. – than a 40 time or three-cone shuttle can tell you. This technology is used extensively in the NFL and is often shown on broadcasts. As it becomes universal in high school and college football, 40 times are likely to become a relic of a less-sophisticated time.

What else can we do with analytics? Can a computer be a better play caller than Lincoln Riley or Norm Chow? The answer is quite clearly yes. Using artificial intelligence, computers can now beat the best chess players in the world. A computer can simply think through all of the possibilities and sift through the data in a way that a human brain cannot. Skynet could and probably should become USC’s next offensive coordinator.

But there are many things that analytics can do for us before resorting to artificial intelligence in the booth. And much of this is impossible to predict right now. Analytics can help to identify an edge. The three-point revolution in basketball was launched when the numbers proved its effectiveness over more-traditional offensive approaches. But such an edge will eventually be copied; what gave you the edge on your competition will eventually disappear as others adopt it, and there will inevitably be a search for a new and different edge. What that is we don’t know. If we knew, teams would already be doing it.

But we know at least one major field of study. Probably the most-important edge has to do with health. Calling better plays and making better fourth-down decisions has value. But keeping players healthy may be more important than anything. If you lose your star QB to an injury, there’s a good chance that no other edge – based on analytics or old-school thinking – will allow you to overcome that loss. Analytics is increasingly being used to analyze things like load management and rest. With time, new technology and the numbers that new technology spits out will help us to keep players healthier. 

To some extent it already has. Kickoffs are in the process of dying a slow death according to Dr. Abramson. The rules have changed and will likely continue to change. This is largely because the NFL analytics people were able to show that kickoffs were the most-dangerous plays in football. If you want to keep players healthy, you need rule changes to protect them from those situations where they are most likely to be hurt.

And, ultimately, analytics may be asked to save the game of football itself. The more we know about CTE and brain trauma, the more football’s future is in jeopardy. Rule changes and equipment improvements are probably critical if we still want to have a recognizable form of football decades from now. Imagine that: the math nerds may end up being that last refuge for the old-school football minds that push back against them today. They may be the only ones who can save the game we all love. 

---

Just a couple of random thoughts to close this out. 

First, over the coming weeks, we will see whether USC will be a true national-title contender in 2023. The roster is much improved this year. The offense will be great. The defense has more talent. The current roster has a chance to make the playoff. But I don’t think the current roster can win it. With a handful of major additions, that might change. USC has gotten bigger and faster in the front seven. If the Trojans can add a couple of additional puzzle pieces there before fall camp, USC may be back in the ranks of the truly elite a lot sooner than anybody thought.

Second, as I suspect all of you now know, Trojan great and WeAreSC contributor Kevin Bruce passed away this past week. I encourage you to read Greg Katz’s moving tribute to Kevin here. I offered some thoughts by video here. Those of you who knew Kevin know what a loss this was. I ask you to keep his wife, Patty, if your prayers.

All of us deal with the loss of friends and family; the older we get, the more experience we have. There is obviously nothing we can do at WeAreSC about human mortality. But one thing we can do – one thing I believe in very strongly – is the importance of keeping alive the history of the game and program we love. There will come a day when the 1974 Comeback or Bush Push are as ancient to USC fans as Babe Ruth and Lou Gehrig are to people of my generation. Fortunately, the prevalence of video will help keep many of those memories alive so we don’t lose this cultural history. But this is also the reason why I think it’s important to record the memories of those who were involved in those events, so we can keep the figures, teams, and major events alive for the next generation of fans. 

Along those lines, I will be recording and publishing part two of my interview with Pat Ruel this week. I encourage you to watch it; Pat is a great guest, incredibly knowledgeable about football, was a key actor in many of the major events of one of the golden eras of USC football, and is completely unafraid to say exactly what he thinks. Join us this week.