You are overlooking huge factors about statistics. First, you're talking about taking the same action repeatedly because statistically it will work out more often than not in the end. Assuming that the action in question does favor the offense over time statistically, that's only true over the long term. Statistical results are based on analysis of large quantities of events to determine the most frequent outcome. The coaches are making decisions with short term results in mind where statistics are far from the most important factor in the decision.
Second, you are assuming that every time we are faced with a 4th & 1 the conditions and variables are exactly the same, which is the only way that statistical trials would be relevant. Every single 4th & 1 decision is different. The players on the field, the position on the field, the hash mark we're on, the time on the clock, the current score, etc. There are tons of variables that need to be considered and only in cases where the variables are identical would your theory of "do it the same way all the time, eventually it will work" hold any water, again assuming it has been shown that going for it in those situations results in a conversion more often than not.
Finally, and maybe most importantly, it's almost as if you are assuming the events are somewhat randomized, like playing roulette or blackjack (if you aren't counting cards), and that statistically the event favors Penn State in the long term. Both might be inaccurate. The results of a 4th & 1 conversion are not random. The players have direct ability to influence the outcome after the decision is made, while in random events that is not true.
One could also argue that the expected long term results do not favor Penn State based on the past 3 seasons. Our short yardage running game has not been good. Our OL has not been good. Our 3rd down conversions haven't been good. What if the long term results, given the variables we know about, lead to a negative expectation (not converting is more common) in the long term? Then your theory of always going for it would result in really bad results.