I thought it would be worth the try at 34-20, honestly, make them score 2 2 pt conversions to win.

But its definitely possible that there was less than a 50% chance of making any kick, so worth it to go for 2 each time. Really, at anything less than like 90% chance of making a kick, you should go for 2 every time, right?

In the early part of the game, your only concern should be maximizing your expected points. So, if (odds of 2pc) * (value of 2pc) > (odds of XP) * (value of XP), then you should go for 2 every time. Or, in other words, if you have a 45% chance of converting the 2pc, then you should kick iff you believe your chances of converting the XP are greater than 90%.

At some point later in the game, when the number of remaining possessions starts to decrease, you should shift away from maximizing points scored and start shifting towards minimizing your opponent's chances of putting together enough scores to catch you. As an example, imagine you score a TD with a minute left to go up by 8 with the PAT pending. Also imagine you estimate your chances of converting the PAT kick as 90%, and your chances of converting the 2pc at an amazing 80%. In this case, while the 2pc will yield more expected points, you should kick the XP every time. Why? Because whether you score 1 point or 2, you are taking the game from a one-possession game to a two-possession game. The only possible bad outcome here is failing to convert for any points at all, so you want to take the path that offers the smallest chances of failing to convert for any points, which in this case is the XP.

So, early on, go for 2 if your chances of converting the PAT are less than twice as much as your chances of converting the XP. Late in the game, if either an XP or a 2pc both result in a similar score differential (e.g. either will make it a two score game), go with whichever you think has a higher likelihood of converting.

The fact that Philly eschewed an XP when scoring to tie and again when scoring to take an 8-point lead tells us that they believed they were straight-up more likely to convert on the 2pc than they were on the XP.