Over the past few years, one of the more controversial aspects of NFL football games is the idea of going for it on fourth down when a team needs two or three (sometime more) yards for a first down. The reason? Something called “analytics,” which has upended decades of NFL truism and tradition. Anyone who watched the NFC Championship game between the San Francisco 49ers and Detroit Lions (or any number of other games this season) knows how contentious this has become.
Whether you agree or disagree with using analytics to help determine the best course of action during a sports contest, it is here to stay. Of course, “analytics” is just a fancy term for statistical research. Unfortunately, TV sportscasters and analysts, whose main job is to inform and entertain the casual fan, seldom if ever discuss how these analytics work, what’s behind them, and how they should be used in the context of any particular situation on the field. And most of the time, it’s only a topic of heated debate when it fails, not when it works.
Until a few years ago, unless it was desperation time toward the end of the game, you almost never saw a team go for it on fourth down from their own side of the 50-yard line if they needed to gain more than a yard. If a coach did so, it was considered bold (if it worked) and foolish (if it didn’t). Some coaches, such as Bill Parcells, Bill Belichick and Sean Payton, were known for successfully taking such chances. Fourth-and-three (or more), on the other hand, was virtually never attempted regardless of field position (again, unless it was toward the end of the game when a team desperately needed a score to avoid losing).
Today, however, because of analytics (and all teams have their own analytics people), this has become commonplace – with announcers informing viewers who they believe disagree with the decision, that “analytics says to go for it in this situation.”
This analysis takes a look at how the ESPN analytics model works during an NFL football game as it relates to recommending whether to go for it on fourth down, attempt a field goal, or punt. Different teams use their own analytics models, but they likely contain many of the same elements – as evidenced by the common refrain that “analytics says go for it.” I haven’t heard any broadcaster or analyst (or coach) say that analytics says don’t go for it on fourth-and-two.
Here is how ESPN explains various aspects of its analytics model, followed by my comments after each point.
- ESPN acknowledges its model is “more aggressive on fourth down than the average coach.” In fact, its model “recommends a go on fourth-and-1 in typical situations anywhere on the field.” There are a few reasons for this.
- “In the modern era when offenses are so explosive, possession significantly outweighs field position in terms of importance. A fourth-down conversion attempt might or might not fail, but a punt always turns the ball over to the other team.”
Comment: In my view, any model that says you should always do something regardless of context is inherently flawed. I also disagree that possession always outweighs field position in terms of importance. You are much more likely to win a game if your average possession starts at mid-field than if it starts inside your own 10-yard-line. Field position is one of the most significant factors in winning or losing. If you go for it on your own 20-yard-line and fail to convert, you are almost guaranteeing that your opponent score. If you go for it on your opponent’s 20-yard-line and fail, they still have to gain 50 or so yards to attempt a field goal and 80 to score a touchdown. Field position is at least as important as possession).
I’ve seen research done in the past that showed the increasing chances of scoring a touchdown or field goal as your starting position gets closer to your opponent’s goal line. I’d love to see an analysis for the 2023 season demonstrating the relationship between superior average starting field position and wins/losses. Of course, I would also note that superior starting field position is most often a result of a team’s superior offense and/or defense.
- “There are two factors for why the model often prefers going for it to a field goal in more situations than most fans would expect.”
- “First, it recognizes opting for a field goal attempt rather than a guaranteed three points. Field goals don’t always go in.”
- “Second, while touchdowns are worth seven points and field goals are worth three, both require kicking the ball to the opponent afterward. The possession and field position for the opponent is worth roughly one expected point in favor of the opponent. Once we consider the opposing kickoff, we can think of touchdowns being worth roughly six points and field goals being worth roughly two to the game’s margin. Thinking about scoring this way shows the relative value of a touchdown compared to a field goal is higher than you might think.”
Comment: If kicking the ball to an opponent after a score is worth roughly one expected point in favor of the opponent, it has the same impact following a touchdown as it does following a field goal. It’s a wash. It doesn’t mean you subtract one point from a touchdown and one point from a field goal – that arbitrarily changes the value advantage of a touchdown vs. a field goal from the actual difference of 2.33 to 1 (7/3) to 3 to 1 (6/2). That is a substantial difference, and biases the model to such a degree that it will recommend going for it on fourth down virtually all the time. This seems to me to be nothing more than a math trick. It’s almost as if they went to their analytics folks and said come up with a math solution that justifies going for it on fourth down.
- “Close to the goal line failures still result in putting opponents in bad starting field position.”
Comment: This is true, but seems to go against the assertion that “possession significantly outweighs field position in terms of importance.”
- “In comparing a punt vs. a fourth-down attempt, ESPN’s model considers the win probability expected given a fourth-down success and fourth down failure, and weighs those by the expected conversion rate of that fourth down. (The expected conversion rate is determined by league averages in similar circumstances and adjusted based on strength of the offense and defense.) That produces an overall expected win probability given a fourth-down attempt, which can be compared to an expected win probability given a punt. The higher win probability is the recommendation.”
Factors not in the model include, weather, injuries, matchup advantages or disadvantages, and what happened previously in the game. Nor does it take into account your team’s success rate on going for it on fourth down in any given situation.
Comment: I can’t really disagree with this because I don’t know how they adjust the average based on the strength of the offense and defense. But I don’t like basing decisions on average league-wide results. – even if you modify the data, it’s still based on a huge database that isn’t as meaningful as people might think. The average of other teams, whose offensive and defensive capabilities can be vastly different from your own, does not provide the most optimal insights. Looking at your own team’s success rate in going for it on fourth and whatever combined with your opponent’s success rate in stopping it is a more valid analysis.
There’s also something to be said about what has already happened in the game providing an indication of what you should do next. In other words, if you went for it on fourth-and-three two times and failed, the model would continue to advise going for it a third time – probably not such a good idea.
None of this is to say I’m against the concept of analytics. I’ve been in the research field long enough to know the value of statistical analysis. But I’ve also been around long enough to know the disadvantages of using big-data black boxes to make decisions that logically make no sense. Context and human experience matter.
Of course, that doesn’t mean coaches just blindly follow analytics recommendations, but it often seems as if many do. If you look at the very best coaches – Bill Belichick, Andy Reid, John Harbaugh, Mike Tomlin, Kyle Shanahan, and a few others – sometimes they go for it and sometimes not in seemingly similar situations. Their experience and well-honed football instincts take precedence. Sometimes it depends on the flow of the game and whether a field goal will put them up by three scores. Sometimes it depends on whether they think their defense can stop the other team. Several years ago, Bill Belichick famously (or infamously) went for it on fourth-and-two from his own 28 yard line against the Colts because every time Peyton Manning got the ball back he drove right down the field. If the Patriots made the first down, they might have been able to run out the clock. Even though it failed, it was a chance worth taking.
Finally, there’s one thing analytics never take into account – the psychological impact and shift in momentum failing to convert on fourth down has on both teams. Just an example from the recent 49ers-Lions NFC Championship Game. The Lions are up by 17 points and the 49ers start the second half by getting a field goal. The Lions are now up by 14 points. They decide to go for it on fourth-and-two from the San Francisco 28 yard line. Had they attampted and made the field goal, they would have been back up by three scores and San Francisco players would have been deflated. But they went for the first down and failed. Now San Francisco was revved up, knowing if they scored again they would be right back in the game (they eventually took the lead). Then, midway through the fourth quarter, the Lions passed up a chance to tie the game with a field goal, opting to go for it on fourth-and-three from the 49ers’ 30-yard-line (and again failed to convert). The 49ers scored a touchdown, putting the game virtually out of reach.
Now, I’m not saying that they relied on analytics to make these decisions (the Lion’s head coach has been known for taking similar risks all season), but analytics likely would have recommended going for it in both situations.
Let the debates, arguments, and gnashing of teeth continue…
Comments