Data Analytics for Football: Why the Vikings Lost to the Seahawks
Football is a wonderfully complicated game. It is akin to two armies lining up and moving up and down the battlefield. Coaches routinely read Sun Tzu’s The Art of War to hone their skills. Data analytics can help coaches get an advantage on their opponents; football is not merely a game of talent, but more a game of strategy like chess.
Football is an 11-on-11 matchup with many combinations of player-on-player interactions which will influence the play of the game. An 11-on-11 matchup can be complicated to model or predict, however, many of the possible interactions rarely happen, e.g.: a corner on the defense rarely interacts with the center of the offensive team. It happens but probably is pretty rare.
NFL Data Analytics
The NFL recently started using analytics compared to the other sports leagues, and the changes can already be seen. For example, teams have been more aggressive on 4th down electing to “go for it” rather than kicking a field goal or punting.
Other statistics teams are looking into are passing vs rushing on a particular down, predicting the number of yards a certain play will yield, going for a 2-point conversion or kicking the extra point after touchdown. Some of these are key to understanding why the Vikings lost to the Seahawks.
Sunday Night Football: Vikings at Seahawks
This particular game was a roller coaster ride for both teams and is why it makes an interesting example to analyze from a probability and statistics perspective. I am going to start from the end of the game to analyze what happened.
There are many other points in the game where play calling could have changed the outcome, but this portion of the game is the most controversial.
There are many other factors to consider which could have impacted the game such as: home field (there was no crowd due to COVID-19), weather, playing on Sunday night vs afternoon, injuries, etc.
4th and 1, 2:00 Left
Most who watched this game would say this was the pivotal moment. The Vikings had the ball on Seattle’s 6-yard line and one yard to go. According to ESPN’s win probability calculator, the Vikings had a 98% probability of winning by going for it and a 97.8% probability by kicking a field goal.
Going for it seems like it would have been the right call as a successful conversion would have effectively sealed the game. However, the starting running back, Dalvin Cook, was out with an injury. Some would argue that had Alexander Mattison run slightly to the right, he would have converted for a first down. Regardless, the Vikings did not convert, and they paid for it.
The coach, Mike Zimmer, may have been thinking about probabilities, however, in football events are not entirely independent. The following play is not independent of the previous play. The previous play influences which play is called next, which players are in the game, how many yards are left to go, time left on the clock, and so on.
If the Vikings had kicked the field goal and converted, their probability would have gone from 91.4% to 97.8% according to ESPN’s win calculator. Not converting the 4th down dropped the probability to 86.2%.
This is also an interesting situation for expected values and outcomes, and would also require Bayesian statistics for an even more in depth analysis.
Field Goal vs Going For It
Per Pro Football Reference, Dan Bailey, the kicker for the Vikings, has made 240 of 276 career field goals. This is approximately 87% and accounts for all attempts regardless of distance and weather conditions. We can likely assume the probability of his making a field goal is greater from a closer distance which was the 6-yard line.
Per Yale University Undergraduate Sports Analytics Group, the conversion rate for 4th and 1 plays during the 1998–2015 seasons for all plays is 65.7%. These figures have likely changed due to use of analytics as more coaches now go for it on 4th down as I mentioned above.
Missing the field goal would also give the Seahawks a better starting position which would also alter the win probability. I could investigate further into more accurate numbers, but for the purposes for this post I think it is acceptable to use these figures. Below is a table showing these combinations:
I used 67.9% for converting on 4th and 1 because the Vikings elected to run.
Keeping in mind these are rudimentary figures and rough estimates, I think it still gives us a good enough picture to show us why it didn’t actually make sense to go on 4th down.
Field Goal Made: 87% — Worst Case: Overtime, Win Probability: 97.8%
Kicking a field goal successfully would have given the Vikings a 97.8% win probability. They had an 87% chance of this outcome had they elected to kick. This means that the Vikings would have been ahead 29–21, a lead of 8 with less than 2 minutes left.
For Seattle to have won this game, they would have had to drive down the field from where the kickoff play ended, score a touchdown, and convert a 2-point conversion, only to tie the game.
They still would have to play in overtime which would have its own analytics to examine.
Field Goal Missed: 13%— Worst Case: Loss, Win Probability: 86.2%
Missing the field goal had a 13% probability. The win probability would have dropped to 86.2% with this outcome. The Vikings would have been ahead 26–21, and Seattle would have been given a shorter field to drive down to score a touchdown within 2 minutes.
We can’t know how the resulting Seattle drive would have been, but let’s assume they score a touchdown because they did in the actual game. This still only has a 13% chance of occurring.
4th and 1 Converted : 67.9% — Worst Case: Run Out the Clock, Win Probability: 98.0%
This is the scenario the Vikings were going for. Analytics had told them that converting successfully would give the highest win probability. The Vikings would have been ahead 26–21 but with the clock on their side. They would have run the clock out, or possibly have scored again with significantly less time left for Seattle.
What the Vikings didn’t consider was the probability of converting for a first down was 67.9%. There was a 19% difference in success for a field goal versus going for it.
4th and 1 Missed: 32.1% — Worst Case: Loss, Win Probability: 86.2%
This is the actual outcome of the game, and there is plenty of highlight film and commentary available. Seattle engineered an incredible drive to ultimately win the game 27–26.
There was a probability of 32.1% that this would be the outcome. We can argue that the Vikings’ defense could have played better, but moments like these are pivotal in the outcome of games and must not be taken lightly. Coaches are paid their salaries for making the correct decisions in these situations.
Why the Vikings Lost
The Vikings ultimately lost because they took a 67.9% chance for a 98% win probability instead of taking an 87% for a 97.8% win probability. The Vikings were willing to risk a 19% lower chance of success for a 0.2% increase in their win probability, and that is the grave miscalculation that lost the Vikings the game.
