The discussion of "playing to win" is an interesting one. It should really mean making the decision that best increases your win probability; in some cases that might be the aggressive decision, and in some cases that might be the conservative decision. The problem is that win probability models in football are based on so little data (compared to say, baseball), and there are many more moving pieces in general. Hence, coaches need to rely heavily on experience and intuition. There are many cases where coaches choose what seems very likely to be the less optimal decision based upon "conventional wisdom" and a strong desire to not be blamed for making the game-costing decision (punt vs. go for it provides many examples, as does goal-line FG vs. 4th down attempt). But there are other cases when there is just too much gray area to know which decision is really optimal, even in hindsight. And the evaluation cannot be made just on outcome, because sometimes probability is on your side, but you end up rolling snake eyes (e.g., your kicker misses a 25 yd FG).