Good points! But just to be pedantic:

If everyone is a self-interested, rational actor, at some point, so few people will vote that P will become significant. At that point, rational people (and by that, I assume you mean for those whom D=0) will start to vote. There's an equilibrium point there. And, oddly, if everyone were D=0, it would oscillate - in one election, no one votes, in another everyone votes. But we're not robots.

Secondly, P actually is significant in swing states. This is why so much money is being spent in those places. P has error bars. At least since the 2000 election was decided by the Supreme Court, these error bars have mattered.

Additionally, I would argue that, at least this year, there's a new variable in your equation (or a reinterpretation of PB): the amount of difference made in the outcome by number of votes by which the winner wins. (sorry, that was a mouthful)

This year, there's a high risk of a contested election. Given that so many people will be voting by mail, and that vote-by-mail has been considered by one side to present a high risk of fraud, and that vote-by-mail takes longer to tabulate, the margin of victory, and the timing of the counting of votes will play a significant role. Therefore, it is important to vote (and to vote early) this year for the first time.

Finally, let's not forget that there's more at stake than just the Presidential ticket. The Senate and the House are at stake at the federal level, and there are many local elections as well. Both P and B are higher than one would consider at first glance.