Author’s Note: This week, I’m gearing up for the season; I want to discuss the mechanics of Saturday afternoon - what I look for, how I watch, and ways I apply analytics to in-game scenarios. Normally, this Thursday afternoon slot will be an advanced stats preview, focusing on TCU’s opponent and the underlying trends. For now, let’s talk about how I watch.

As the summer wraps up and the site shifts towards real, live, in-season football content, I want to reflect on what we’ve learned in the stats series these past couple of months: nothing. We’ve learned nothing. What we have done is taken some fun looks at what conferences will be the most entertaining, how conference standing affects recruiting, and how the polls have shaped the playoff narrative. Along the way, we got a little more theoretical, questioning the true random component of fumbles, identifying crucial game moments with a leverage index, and even extending the concept of success rate into the wxSR. My personal favorite this summer was unveiling the passing chart, an exciting way of plotting an offensive attack. Without spoiling too much, all season I’m going to have some fun with graphical representation of offensive tendencies, both for and against TCU. That’ll have to wait until we aren’t playing an FCS opponent, though.

This week, in preparation of meaningful football across the country, I want to document some of my thinking on what to look for in watching football analytically, or:

How to build a model of college football in your spare time

Lucky for you, all models look pretty much the same. There are some economic agents. They make choices in order to advance their objectives. The choices have to satisfy various constraints so there’s something that adjusts to make all these choices consistent. This basic structure suggests a plan of attack: Who are the people making the choices? What are the constraints they face? How do they interact? What adjusts if the choices aren’t mutually consistent?
-Hal Varian, in “How to Build an Economic Model in Your Spare Time”

Watching college football can be a religious experience for some; wearing the appropriate garb, cheering in unison with other congregants, investing a couple hours a week to get that emotional catharsis of winning or losing. But beyond rooting for a touchdown, what matters in college football? Today, I will briefly recap some of the most important stats in the game, and then construct a model of how I observe those stats and the underlying mechanisms to determine team quality and team performance.

How to Win a Football Game:

From Bill C.’s “Five Factors”:

But over time, I’ve come to realize that the sport comes down to five basic things, four of which you can mostly control. You make more big plays than your opponent, you stay on schedule, you tilt the field, you finish drives, and you fall on the ball. Explosiveness, efficiency, field position, finishing drives, and turnovers are the five factors to winning football games.

These five factors as the basic indicators of a team’s success. There’s a discussion to be had about the causal responsibility in winning that each of these factors have, as correlation in itself doesn’t mean much. What these five factors do give is a baseline picture of what to look for when watching the game, which leads me to constructing a model.

I start my model by discussing the stats I look at after the game for diagnostic purposes. By starting with the end, I can then trace back a list of in-game indicators to focus on.

• Success Rate: How often a team is gaining meaningful yards, contingent on situation. I have seen estimates ranging from a 30-50% increase in probability of scoring on a drive when you gain at least 4 yards on a first down. I tend to treat second down as an intermediary, and focus on first and third down success rate. If a team is not starting drives well, or can’t seem to get those last few yards to keep a drive going, I know they have issues.
• IsoPPP/wxSR: When a team does gain yards, how are they gaining them? The higher both of these stats (Isolated Points per Play and Weighted Expected Success Rate) provide measures of how strong a team was in moving the ball - measured by change in expected points or in points per successful play.
• Scoring Opportunities per Drive/Points per Scoring Opportunity: We are concerned with field position, as it dictates the flow of drives. More important than where you start a drive is how you finish a drive. Looking at how many scoring opportunities you set yourself up for (ball inside opponent 40) married with how often you convert that scoring opportunity (into TDs, preferably) is a great way to measure the potency of an offense.
• Net Points off Turnovers: This aggregate stat tells us in one fell swoop who won the turnover battle, and more importantly, who swung the game with turnovers. A large margin either way in NPT is a way to isolate a team’s offensive and defensive quality aside from the actual scoring margin.

With those five, I can more or less diagnose exactly what happened in the narrative of the game. I usually get that information from the advanced stats box scores off of team stat profiles, which should be out in a couple of weeks. In addition to those broad strokes, I like to get a little bit more detailed, especially when looking at TCU offense and defense.

To diagnose a specific team over the course of a game and season, I keep my eyes on the following stats, with an eye on the season average versus the single game values. That way, large differences can highlight specific weak or strong points.

Stats for a Deeper Dive

• Third Down Conversions: When you need a bailout, can you give yourself one? Can a team sustain drives?
• Unforced Errors: Miss a block, run the wrong route, throw it into the ground, drop an interception? All these and more are unforced errors, which subtly add up but explicitly tilt a game. A lot of games can be traced back to a couple pivotal unforced errors, and acknowledging them on either side of the ball can distinguish a good performance by your team from a bad performance by your opponent.
• Yards Per Attempt: I only include this to make the point that sacks should not count as rushing attempts, and really distort rushing numbers. Keeping that in mind, we get a more accurate picture of the run game.
• Hurries/Sacks/Passes Defended: All measures of aggressiveness of a defense, and underlying points of analysis for team effort/execution.
• % of Total Yards in the Air: This one is a little more nebulous, but evaluating a passing attack or defense requires an understanding of how passing yards are obtained. Of course, one is not better than the other, but it helps to parse out intent and execution to examine air yards.
• Play Sequence: I’ve got some great visualization tools for play sequence, and it is perhaps the area I’m most interested for this season - how do teams set up plays, how do they string together drives? I’m looking forward to learning much more about this as we analyze the upcoming season.
• Spread: This relates directly to the passing and rushing charts I introduced a couple weeks back. I am interested in how a team varies the direction and ambition (long pass or short pass, etc) of their plays, and visualizing this will be a large part of my stats analysis this year.

Conclusion

Watching college football analytically requires a framework for understanding, and by laying out the factors directly contributing to success, we provide that framework for ourselves. This season, as you watch, move past the numbers on the scoreboard and “total” statistics into the world of advanced stats. Look for how a team moves the ball, how unforced errors can pivot the direction of a game, and for how well a team does what it sets out to do. Enjoy the start of the season.