While I know playoff performance isn't factored into the MVP
A player's actual performance in the playoffs in a given season is not factored into his MVP voting.
However, probable playoff performance should be factored in. Part of what makes a player valuable is how much he helps his team in the playoffs.
Thus, every MVP voter voting on a player P in a season S should include an estimate Playoff(P) of P's likely performance in the playoffs (or what that performance would be had his team made the playoffs).
More exactly, given an actual performance
regular in a regular season by a player P, each voter V has some estimate of P's playoff performance:
V_
playoff (
regular)
Clearly, for most of the last 17 years, and particularly '14 and '16, anyone would say that Playoff(Brady) was MVP level.
However, obviously Brady did not win the MVP in those years.
Therefore, each voter V who assessed Brady's likely playoff performance in 2016 (and likely previous years) as not being MVP caliber, should assess his own function
V_playoff(
regular)
in such a way that
V_
playoff(Brady)
is higher than it was.
In this way, eventually voters assessments of players' playoffs performance in a given year is corrected by their actual performance in previous years. This correction should eventually result in whatever factors Brady evinces in the regular season to be more likely to lead to playoff wins.
In other words: voters should admit they misunderstood Brady's abilities in previous years, and they should learn from their mistakes. So it is not strictly accurate that playoff performance is not considered in MVP voting, rather, only actual playoff performance in a given season is not considered for the MVP for that season, but previous playoff performance could have indirect effects.
[P.S. this is a kind of "reinforcement learning" for any computer science types out there: voters should learn eventually to predict playoff performance]