A deeper look at the Edmonton fake punt (July 24, 2014)

Some background: Expected Points and Expected Points Added

Expected points (EP), and more specifically, Expected Points Added (EPA), are metrics I use on http://www.cflstats.ca that has been used by NFL analyst Brian Burke for quite a few years now.

Using historical scoring data, we can assign a point value for every down/distance/line of scrimmage combination in a game.  By looking at every play and then the next score for either team, then grouping it by down/distance/LOS, we can come up with an average expectation for that play.  EP can be positive or negative, indicating whether we expect the offense (positive) or defense (negative) to score next.  In calculating the values, some game situations are filtered out in order to keep the values more accurate; plays which occur in the last 4 minutes of a half, or when the score margin is 14 points or more are not included in the calculations, in order to decrease the effect of garbage time or 2 minute drill type possessions.

Once we have a value for each game situation, we can then calculate EPA, which is simply the difference between EP on the next play and EP on the current play (EP After - EP Before). Positive EPA means the play moved the offence into a better position, negative means they are worse off than before.

Looking at EPA and comparing some possible outcomes can give us clues to whether in-game decisions were good or bad, or if risks were worth it.

The Play

On 3rd and 10 from their own 6 yard line with 26 seconds to go, Edmonton opted for a fake punt and gained just shy of the 10 yards necessary for the first down.  Calgary scored a touchdown on the next play, and Edmonton was left looking like they'd made a bad decision.

But was it really a bad decision?

Outcomes and potential EPA

3rd and 10 from your own 6 yard line is a bad place to be, and the EP value reflects that.  EP in this position is -1.3 points for the offense, meaning most of the time, the offense will give up the ball (or a safety) and the defense will be the next team to score.

Going into the play, they had three options: 

1) Punt - Edmonton averaged 38.7 net punt yards on the night, so punting from the 6 yard line would expect to give Calgary the ball back somewhere around the 44 yard line.  1st and 10 from their own 44 yard line carries an EP of -2.4 for the Edmonton defense.  Over the past 5 years, kickers have averaged 81% on field goal attempts from this range, which was the mostly likely scenario given the time remaining in the quarter.  EPA for this outcome would be -2.4 - -1.3 = -1.1

2) Go for it (and succeed) - Lets assume they got those extra needed inches, and kept the ball on their own 16 yard line.  That gives Edmonton 1st and 10, which carries an EP of 0.3.  In this situation though, Edmonton would certainly have opted to kneel out the quarter, so in actuality their EP for this case would be a flat 0 EP.  EPA for this outcome would be 0 - -1.3 = 1.3

3) Go for it (and fail) - Or exactly what happened, in other words. On average, teams on 3rd and 10 that go for it are successful 23% of the time, for an average gain of 6.3 yards.  Plugging in the average yardage gives Calgary back the ball on the Edmonton 12 yard line, for an EP of -4.0.  EPA for this outcome would be -4.0 - -1.3 = -2.7

The success rate for 2 and 3 are linked, meaning the true value of going for it must be calculated as a fraction of both, however.  Historically, teams have converted on 3rd and 10 just under 23% of the time, which includes fake punts.  That means the average EPA is a combination of the two:

EP_Success * SuccessRate + EP_Failure * FailureRate

For this situation, we get an average EPA of -1.78

To recap, that leaves us two outcomes - punt for an EPA of -1.1, or go for it, for an EPA of -1.78.  Those are surprisingly close.

Conclusion

In a vacuum, or as a standard third down gamble, going for it here is the wrong decision.  Both options are bad, as both indicate that Calgary is more likely to score next, but going for it is a little more than a half a point worse over all.

But, this didn't happen in a vacuum, time was a major factor here.  Ordinarily, gaining your team a first down on your own 16 yard line isn't worth all that much, especially when it comes at the risk of a -4.0 EPA swing.  But in this case, gaining the yardage would have allowed Edmonton to kneel out the quarter, giving up zero points, opposed to giving Calgary the ball back in position to kick a field goal from a spot on the field where kickers are fairly successful (81%).

And it wasn't a gamble, it was a fake punt.  Fake punts are a bit harder to quantify the success rate on, as they rely heavily on the element of surprise, and whether or not the team has found a weakness they hope to exploit.  I don't know the league average for success on fake punts, but I would wager that they are slightly more likely to succeed than a 3rd and long gamble, especially if the punting team has spotted something they think they can exploit.  Edmonton only needed to convert at a 40% rate in order to break even vs the punt.

It's a very close call on this one.  In a tie game, giving up the ball and a likely field goal could have been the difference in the game.  Conversely, giving up the ball inside your own 20 is a huge risk, but with less than 20 seconds left, there's a very good chance that most of the time, Calgary only walks away with a field goal here anyway.  

So should Edmonton have punted? I actually like the decision here: if the Edmonton coaches felt that the Calgary defense was unprepared or likely to cheat back to block, they may have felt their chances of succeeding were much higher than 40%, and after a good defensive half, they had to feel that they could hold Calgary to a short field goal in the case of a turnover.  Unfortunately, Edmonton's defense didn't hold, and the gamble resulted in the worst possible outcome, but I give credit to Chris Jones and his staff for making the aggressive choice at a reasonable time.