For anyone who watched the SMU vs. Memphis game, you saw an example of a coach actually using analytics and made a smart decision to "Go for two". Down 14, SMU scored a TD to narrow the deficit to 8. Conventional wisdom says, "kick the PAT, score another TD and PAT, tie the game". But that is not what Sonny Dykes did. Instead, he opted to go for two, and cut the deficit to 6. Bill Connelly discussed this decision in an article today. 

Basically, if you make the 2-point conversion, you're in position to take the lead if you score another touchdown (at least, unless you're snakebitten in the place-kicking department -- earlier this year against Virginia Tech, Miami went for two, got it, scored again, and missed the PAT). If you miss the 2-pointer, you can still go for two to tie if/when you score again.

As he mentions in the article, this trend is becoming more prevalent this season; however, it has received minimal coverage, as both Herbie and Fowler were baffled by the decision, until a coach text Herbie during the broadcast, to explain the reasoning. 

Analytics take everything into account. They judge decisions based on multiple things. Summed up, you are left with this:

1. How your odds to win increase, if 2 point attempt is successful.
2. How your odds to win increase by kicking the PAT. 
3. How your odds to win decrease, if 2 point attempt is unsuccessful. 

Essentially, if you odds to win by Option 1 increase, more than your odds decrease by Option 3, then go for 2.

So how can it be advantageous to go for 2, while down 15? Let us assume teams convert 45% of 2-point tries (national average for 2019), so barely less than half your tries. Down by 15, you would have 2 tries to convert. If you fail on your first try, numbers say you have a high chance to convert the 2nd try, and thus, tie the game. However, if you convert the first try, then you just need a PAT to win the game. In essence, your odds to score 14 points via two PAT's, is just slightly higher than with two, 2 point conversion attempts (due to 45% conversion rate). BUT, that doesn't factor in the key statistic. Overtime. Factoring in overtime (50% chance to win), the odds favor Sonny Dykes strategy. (Increasing your odds to win by 5.2% over kicking two PATs).

Interestingly, this is not the only time when analytics disagree with conventional wisdom. If you score a TD to decrease the deficit to 4, analytics say "go for 2", even though a PAT puts you down a field goal. Why? Remember analytics factor everything into account. For example. You "go for 2" and convert. You are down 2 points. You get the ball back and kick a field goal to take the lead. Now, the other team has to kick a field goal to win. Now let's say we kick the PAT instead of going for 2. We are down 3, get the ball back, and tie the game up. From here, we are either losing, or going to overtime, where we have a 50% chance of losing. These swing the odds in favor of attempting the 2-point conversion. 

My apologies for the lengthy post, but I just found this absolutely fascinating because it goes against everything I thought. And maybe if OSU is ever in a position like this, and Ryan Day goes for 2, you'll know there is some reasoning behind it. 

So for all the "coaches" in the room. What call are you making in these decisions and why don't we see more of these analytic type calls? My belief is, I wouldn't want to be the one explaining analytics to a 70 year old booster, if things went poorly.