I know the YC community frowns on general political content ... but there's a mathematical twist here.
Observing the primaries, it makes little sense to me why all delegates count equally. I want to propose a better algorithm for determining the winner of a primary (better than simply, "who has the most delegates?"). I'm going to assume that a party wishes to maximize the probability it will win the general election in November - no other considerations.
Here's the problem - take the Democratic primary for example. It shouldn't really matter which candidate wins Oklahoma (the Dems are going to lose it in November anyway), or DC (the Dems are going to win it in November anyway). Performance in swing states like Pennsylvania & Ohio should count for a lot, because winning there is crucial to winning the general election.
Currently, the power of a state is proportional (approximately) to its population. I propose a system where a state's power (ie delegates) is proportional to its population AND inversely proportional to its average spread (absolute difference between Rep. & Dem. percentages) in the past 3 presidential elections.
So,
Delegates = k* (Population / Avg. Spread),
with k selected appropriately, and the rounding done appropriately.
What do you think?