The only way I can see to make candidates threat every single state as being terribly relevant would be to have the election decided based at least in part by gains made in a state. So if in the previous general election D got 47% and R 51%, D getting 49% this time would be worth 'something' in terms of final score, even if they 'lose the state'. That way you'd have to fight to prevent any gains by the other side. This is just a mathematical thought experiment, meant to be an example of how to refocus campaign goals: there has to be incentive.
I think you're basically saying keep the elector selection statewide. But apportion the delegates proportionately to the votes cast in the state? (Where obviously that would end up being a function of how many electors a state has)
I could see good and bad things from that. It would make California and New York matter more, as even shifting a couple percentage points could mean several more EC votes(and fewer for the other guy). While Wyoming still wouldn't matter much, with three electors, the cutoffs would likely be 34% and 67% respectively. Although it could warrant a little more Democratic attention.
Trump won Wyoming in 2016 with 67.4% of the vote. Hillary won 21.6% of the vote. Which actually brings in another question when you're dealing with a fractional apportionment as would have happened there.
Also we need to revisit the bigger states for a moment as well, in this case I'm going to look at Utah in 2016. They have 6 EC votes. Every 16.66~% of the vote is 1 EC ballot.
Hillary won 27.46% of the vote. So she's at 1 and 2/3rds of an EC vote.
Trump won 45.54% of the vote. So he's at 2 and 3/4ths of an EC vote.
McMullin won 21.54 of the vote. So he's at 1 and 1/3rd of an EC vote.
nobody else obtained more than 4% of the vote in Utah. In this case it would like Utah would have split 2 EC votes for Hillary, 3 for Trump, and the final vote would have gone to McMullin for a total of 6 EC votes cast.
So now we have a three way race for president in the EC vote in the House.
But now let us take a look at California, where wither 55 EC votes, 1.8181~% of the vote(simplified to 1.8%) equals 1 EC vote.
Hillary won 61.73% of the vote. That nets her 34.29 EC Votes
Trump won 31.62% of the vote. That nets him 17.56 EC Votes
That makes 51.85 EC votes between them out of 55.
Johnson(libertarian) won 3.37% of the vote. That nets him 1.87 EC Votes.
Stein(Green) won 1.96% of the vote. That would net her 1.09 votes.
Which gets up to 54.81 EC votes cast
As per Wiki, McMullin won 0.28% of the vote, and "others" won another 1.04% of the vote. So not enough to decisively apportion another EC ballot.
Trump and Johnson are the two candidates with a greater than .5 EC Vote apportioned to them so they'll be rounded up.
34 EC votes for Hillary, 18 for Trump, 2 for Johnson, 1 for Stein for a total of 55 so thankfully we're not needing to play other games to get the correct total.
Except now just between Utah and California, we now have a 5 way race for PotUS in the EC. The constitution limits congress to the top three, so it'd be a race between Trump, Clinton, and Johnson(with 2 EC votes).
I'm not going to "math out" New York, but with 29 EC votes, a candidate would need 3.44% to get a full EC vote apportioned. Johnson obtained 2.29% of the vote in New York, and Stein obtained 1.40% of it. McMullin and others combined amassed a further 0.79% of the vote. So it would be a reasonable bet that Johnson would have picked up a third EC vote from New York as well.
Texas would have given Johnson another EC Vote as well. With 38 EC votes, 2.63% of the vote is needed for a full EC vote apportionment.
Johnson won 3.16% of the vote, for 1.2 EC votes
Stein/McMullin/"others" combine for 1.37% of the vote for a combined share of 0.52 EC votes
Trump won 52.23% of the vote, for 19.85 EC votes
Hillary won 43.24% of the vote, 16.44 EC Votes
So that'd be 20 for Trump, 16(+1?) for Hillary, and 1 for Johnson.
From a quick run down the table, I don't see that apportionment method throwing any other EC votes to a third party in 2016, but it would have thrown 6 EC votes to third parties in 2016.
Not sure how the rest of the electoral landscape would have altered with the change to methodology across the various states. Trump won 304 to 227 in 2016(7 faithless electors).
In making Texas, and Utah proportional, Trump loses 21 EC votes while Hillary gains 19 EC votes.
But in exchange Hillary lost 21 EC votes in California, while Trump gains 18.
So 304 to 227 turns into 301 to 225 to 3 to 1 to 1