The trend in the NFL lately toward analytics--or sabremetrics, or Billy Ball, or whatever we should call this data-driven discipline in football--is out of control, misapplied, and misinterpreted. What makes it even more annoying to me is the dumbness in the name of being smart. Let's break it down a bit.
First, football is not baseball. Football has a much, much smaller sample set of data to draw from relative to baseball, so trying to apply pure analytics to football is already a bit sketchy. A pitcher thows maybe 30 games a year of 80 pitches, giving some 2400 data points, but a football team scores maybe 3 touchdowns a game for 17 games, giving 51 data points to draw from. And for any analytics, sample size matters. This puts football at an analytics disadvantage out of the gate.
However, even if we assume the metrics bear out in football, bad decisions are still being made with the data available.
For
example, in today's (12 January 2024) divisional playoff game between
the Tampa Bay Buccaneers and the Detroit Lions, the Bucs were down 31-17
in the 4th quarter (14 points). The Bucs scored a touchdown and decided
to go to the 2-point conversion. Going to 2 in this position is dumb
for a couple reasons, even without applying analytics.
- The obvious: 2 TDs and extra points is 14 points, which is enough to tie.
- Even if you want to play for a win, you can kick for the 7 points now and go for 2 if you score again when it really matters.
- If the opponent were to kick a field goal, you can't catch them with 16 points, so going for 2 twice has no benefit even if successful.
- And maybe the worst reason to go for 2 too early: If you don't succeed, your team, which just made a great play to score the TD, feels like they failed. This puts the team at a psychological disadvantage.
I think my philosophy can be boiled down to this: Don't go for 2 points until it really matters.
I've seen a college team go for a 2-point conversion when they were down by 24 points in the first half. To do so, I assume the thinking must be that the other team will not score again and that his team can get 3 TDs and 3 2-point conversions to catch up. That's ridiculous! There are so many combinations of scores that could happen in the game that you don't even know if the 2 points will matter later in the game, but the 1 point that you are giving up if you fail may be the difference later in the game. But you don't know that until later in the game. If you make a decision now about what the situation might be later in the game, you are out-thinking yourself.
Now to the actual analytics: Chris Collinsworth, after the failed Bucs 2-point conversion, attempted to break down why it was smart to go for 2 points. To the best of my recollection, here was his argument (paraphrased):
The
rate of successfully converting a 2-point conversion is 55%. The rate
of success in kicking extra points is about 95%. So if you add that all
up, it's a better chance to go for 2.
On the surface, this sounds right. If a 2-point conversion is successful 55% of the time, the average points per attempt is 2 times 55%, or 1.1 points per attempt. On the other hand, an extra point yields 1 point with a success rate of 95%, so the points per attempt is 1 times 95%, or 0.95 points per attempt. Since 1.1 points is more than 0.95 points per attempt, then going to 2 is on average better than attempting an extra point.
The problem is that the outcome of a game is not about averages; it is about situations. And each situation needs to be analyzed in terms of the desired outcome, or more specifically, the chances of all the possible outcomes that may be applied to a given scenario.
To show what I mean, let's break down 2 separate scenarios, assuming the team can score 2 touchdowns:
- They attempt a 2-point conversation on both.
- They attempt an extra point on both.
Scenario 1: Two 2-point conversion attempts
- Make 1st (55%) and make 2nd (55%)
55% times 55%
30.25%
Result: WIN - Make 1st (55%) and miss 2nd (45%)
55% times 45%
24.75%
Result: TIE - Miss 1st (45%) and make 2nd (55%)
45% times 55%
24.75%
Result: TIE - Miss 1st (45%) and miss 2nd (45%)
45% times 45%
20.25%
Result: LOSS
Scenario 2: Two extra point attempts
- Make 1st (95%) and make 2nd (95%)
95% times 95%
90.25%
Result: TIE - Make 1st (95%) and miss 2nd (5%)
95% times 5%
4.75%
Result: LOSS - Miss 1st (5%) and make 2nd (95%)
5% times 95%
4.75%
Result: LOSS - Miss 1st (5%) and miss 2nd (5%)
5% times 5%
0.25%
Result: LOSS
Note that the probabilities in both scenarios add up to 100%.
Now let's apply this to the desired outcomes.
Outcome 1: Don't lose the game
In order not to lose the game, the team must score at least 2 points.
For 2-point conversions, that means they must convert one or both of the 2-point conversions. The only scenario in which they lose is failing at both. Thus, there is a 20.25% chance of losing by attempting both 2-point conversions.
For extra points, they must successfully kick both extra points to avoid losing. If they make none or one of them, they lose. Thus, there is a 4.75% plus 4.75% plus 0.25% chance of losing. So attempting extra points gives the team a 9.75% chance of losing.
But wait! Going for 2 has a higher average points per attempt than kicking, but applying the probabilities to the desired outcomes shows that kicking the extra point results in less chance of losing the game!
However, a tie also gives the team a chance to win the game later.
Outcome 2: Win the game
Making both 2-point conversions is the only scenario that outright wins the game. However, we could combine scenarios in winning combinations (that is, kick an extra point and complete a 2-point conversion). Maybe I'll go through these scenarios another time.
However, a tie also gives the team a chance to win the game later.
Conclusion
To me, this is still a no-brainer. I would definitely take the extra point on the first TD to build on the success of the TD and take the free point. If my team scores another TD, in most situations I would take the extra point to tie rather than playing for the win. The tie gives my team a chance to play for the win on the next score (maybe in overtime).
What do you think? Do you agree that analytics is being misapplied to football?
