Imagine a football team trailing in the fourth quarter by 15 points. They are driving down the field. They then score a touchdown. A strategy decision now needs to be made: go for the two-point conversion now or wait for the next (necessary) touchdown to then attempt the two-pointer.
Their head coach chooses to kick the extra point.
Cue Analytics-influenced NFL writer, as if he or she is Batman seeing the Bat-signal for another mistaken coaching decision from the Big Bible Book of Football Analytics, 1st Edition (because the analytics do not need to be updated once the rules have been established. Did they have to ever update the Old Testament?). It's time to tweet!
"Should have gone for two points now. Better to know now if you need to go attempt an onside kick." (Aren't I smart? Where is the drop-mic emoji on Twitter, anyways?).
Look, I understand that the earlier on-side kick is advantageous since there may not be enough time for an on-side kick if you miss the two-pointer at 21-19 -- but that possibility still exists in the world where the head coach defers the on-side kick decision, by the way -- and that does play a role in the floating (and hidden) probability chart.
However, the "it's better to get the info ASAP" argument is not nearly as a slam dunk as the Twitteratti and The Analytics Say crowd suggests. There are at least three reasons why deferring might increase the probability of success on a deferred two-point attempt.
(1) More information is important -- and that also includes acquiring more play-calling knowledge regarding what is working (and why) against the defense. Another offensive series produces more observation and data regarding what is working and what is not. And if the probability of forcing overtime is highly dependent on the success of the eventual twp-point attempt, then getting into that play with the highest possibility of success takes on disproportional impact.
(2) The defense is getting more tired as the game goes on. Attempting the same two-point conversion play later in the game could see an uptick in the probability of success simply if it occurs later. I have heard several NFL coaches espouse their belief that players hit a wall where fatigue sets in. I have heard former Cincinnati Bengals and Dallas Cowboys offensive coordinator Paul Alexander identify this threshold at roughly 50 snaps per player.
(3) The heightened momentum from a team attempting to tie the score with a deferred two-pointer may be higher than the earlier two-point attempt. After another offensive touchdown and a defensive unit now on their heels at the precipice of blowing a 15-point lead, the probability of success may be higher at the moment for the offense than it would have been at the first touchdown.
Those three arguments are all playing the probability game that The Analytics Say folks revel in. But the best argument to defer the two-point decision is this: failing to convert the earlier two-point conversion might be too deflating for the trailing team.
Simply kicking the extra point keeps the trailing team at a hypothetical one-score deficit. And to quote the great philosopher Lloyd Christmas in Dumb and Dumber: "So you're telling me there's a chance!"
Keeping the deficit at only one score maintains hope for the trailing team -- and that is a team that still needs to score a second touchdown, either way. A missed two-point conversion is a deflating buzz-kill, with the metaphorical mountain needing to be climbed still very high.
This is a calculation that is dependent on the belief in momentum. And by momentum, I mean, that human beings' performance can increase or decrease from their baseline effort based on positive or negative feelings.
At its core, many of the applications of analytics presume that momentum does not exist. The experience of the moment is dismissed as a product of hindsight bias. Admittedly, most of the Gotcha! folks on Twitter are not aware of this. The savvier ones do appreciate that the application of the quantitative they are citing is dependent on every statistical moment being roughly the same -- a necessary component for their sample size to be valid and the application to the new situation (where they are now geniuses) being appropriate. So, many of the "get the info ASAP" folks will deny any negative impact of deflating a team from a failed early two-pointer because ... it is impossible to deflate a team, or something.
In the end, I don't think it is a bad decision to take the two-pointer early. If an offensive coordinator thinks his team has the goods to convert the play at that moment, then go for it. But this notion that head coaches who do not attempt the two-pointer ASAP are making some huge mistake in the Big Book of Analytics is just a take from someone living in Clownsville.
Best of luck -- Frank.