To explain this easily I need to move down to Rock, Paper, Scissors. If you are up against a bot that always throws each 33% of the time with no strategy involved, that is called an un-exploitable strategy, but it's also an un-exploiting strategy. Even if I were to always throw Paper 100% of the time against this bot, we would always on average tie. This is the Nash Equilibrium point.
What you want to do is perhaps start with this random strategy, determine the realtive strength of your opponent, and only move to an exploiting strategy if they are weaker than you. If they throw Rock 40% of the time, you should throw Paper 40% of the time. If you were to enter a RPS tournament playing randomly, it is unlikely that you would reach the end. You need to exploit.
In Poker Cash Games, people just play the Nash and when opponents make flat-out mistakes they yield a little bit of profit. Do this enough with high enough stakes and it works out, even if its extremely boring. But this won't work in Poker Tournaments, where you only get payouts for top 10%, weighted heavily towards winning. So what happens in Poker is what happens in RPS - you attempt to find exploits, and your opponents attempt to find exploits in your play. In turn each of the players are attempting to move the environment of the play towards a position where they know it well and can find those mistakes.
If deviating from the Nash yields better results than the penalty of deviating because in that environment you are stronger than your opponent, you should take it. As this article points out, the same is true in Chess. And as you point out, this is also true of the Eephus pitch.