Sounds like the same argument (at least wrt a and b) would apply at least as well to deciding elections by the election-night vote counts.
Guess we should go straight to coin flips, and save the trouble of having a campaign.
Election night vote counts are only an issue when the result is close. and there is no evidence that recounts or recounts of recounts are more accurate than the original count. Count a jar of marbles three times. You get three different results and you have no way of knowing which of the three is the closest to being accurate.
Most elections aren’t close. The ones that are, have a way for settling those that is consistent and fair. One way would be, “the first count is the final count”. Another could be a coin flip. Another could be “challenger wins” since incumbennts have so many advantages that a close election almost always indicates deep voter disatisfaction.
Cite, please.
There are no studies, thus no evidence.
Well, you might consider that recounts, at the margins, tend to turn ‘no vote’ ballots into votes for one candidate or another. To the best of my knowledge, they almost never go the other way. The recount will be more accurate than the original vote if the determinations made by the persons doing the recount with respect to those formerly uncounted ballots are more accurate than chance would dictate.
Take this Virginia election. To indicate your vote, you filled in a bubble on a form with a pen. Such ballots can easily be read with a scanner - if filled out correctly.
Some ballots that can’t be read with the scanner still show a clear intent for one candidate or another. In the recount process, both the Democratic and the Republican recount officials have to agree on a change in status for it to be made. If only one of the two supports a change, it is reviewed by the court.
In a race between a Dem and a Republican, if a ballot had been counted as a ‘no vote’ in the initial count, but both the Dem and the GOP recount officials agree that a ballot is a vote for one candidate or the other, it seems inarguable that the accuracy of the election has been improved by that change.
Given that in VA-94, there were a couple dozen such changes, and only one that was questioned by either party, how do you argue that the post-recount count is no more accurate than the original count? That’s just silly. There’s an argument to be had over that last ballot, but either way, either a tie or a one-vote Simonds win is closer to accurate than the original 10-vote Yancey win.
I’ve heard that in Nevada, the official procedure for a tied vote is a single hand of five-card stud. Yeah, mathematically it’s still just a coin flip, but there’s something to be said for style.
This doesn’t actually work. Suppose, for instance, that your threshold for “too close to call” is 1000 votes: If one candidate wins by less than 1000 votes, then it’s too close to call, and it’s a coinflip, or a win for the non-incumbent, or whatever your rule is. But now, suppose that the initial tally comes in, and the incumbent has won by 1002 votes. The other party challenges, and says that there were seven votes that were debatable, and that they think that the actual vote tally was really only 995 votes in favor of the incumbent. And they demand the recount, and pay for it out of their own pockets, and in the recount not all of the debatable votes go their way, but some do, and now the total is a 997 vote lead for the incumbent. That’s less than the legally-required threshold, and so now the challenger wins.
Your proposal doesn’t change the fact that there’s an arbitrary sharp line between one side winning and the other. It just changes precisely where that line is.
They should have the two candidates compete in a “build fire” contest. You guys did watch the Survivor finale last night, right? 
Going back to the ballot pic, here’s the official Commonwealth of Virginia guidance on interpreting ambiguous ballots.
It’s got literally dozens of examples.
I haven’t had a chance to look at it, so I’ve got no opinion at all about its applicability to the disputed ballot. But I’m sure others will. And so will I, when I have a bit more time. 
First count is always final. The idea is to set the threshold higher than the margin of error for the election. So if we assume that election counts are accurate to within 0.2%, then a candidate at 0.21% ahead is going to win whether they do a recount or not.
Re: the ballot in question, If I ran things, that would be a no vote reflecting “this person is too damned stupid for their vote to count.” Probably would say the same about the one with 2 circles, one slightly lighter than the other.
That’s really no better than just setting the threshold at 50%, then. You might as well just say that 50% exactly is the cutoff, and still say that the first vote count is final.
This appears to be the relevant part to me, from page 8:
(8) Any ballot that has any mark, as above, in the target area or candidate area for one candidate, and on which other marks in the target areas or candidate areas for any other candidates have been partially erased, scratched out, or otherwise obliterated, shall be counted as a vote for the candidate for which the mark was not erased, scratched out, or otherwise obliterated, provided no other candidate is similarly marked.
IIRC, my ballot here in Arkansas told me to simply turn in the spoiled ballot for new fresh one, and not to try and correct mistakes.
That thin little line counts as “erased, scratched out, or otherwise obliterated”? Not in my understanding of those terms or phrases.
In the interest of finality and determinism and to avoid shenanigans, preference should be given to the initial judgement. The disputed ballot could reasonably be treated as a vote for the GOP, perhaps, but it could also reasonably be treated as No Vote.
FWIW, that’s a reconstruction attempt by a guy who wasn’t allowed to take a photograph.
~ ~ ~ ~ ~
Do you see that your post for which this was the defense could be misinterpreted? Was that ambiguity intended as a feature, or was it a bug?
It now contains a picture of the actual ballot. I’d say that was a relatively clear R vote, with correction of mistaken D vote, but I’m sure others have different opinions.
Especially because the instructions printed right on the ballot tell you to take it back if you goof it up and get a new blank one and start over. Wouldn’t that mean invalidating goofed-up ballots? Logically?
In what grade do kids start filling out those little Scantron circles for tests in school? If they haven’t figured out how to do it by the time they’re 18 and legal to vote, well, that’s a poll test I could stand behind.
My opinion is that I’m really glad I don’t have to try to figure that sort of thing out. I don’t feel like I could arrive at a logical way to make decisions about such ballots.
At least for me, the link goes to a Twitter thread, and it jumps straight to the third post in that thread, which is the photo of the actual ballot. The first post is a crude handwritten reconstruction of the ballot and what the voter filled in. The second post is a reconstruction of what the voter filled in, but on a real ballot. The third is the actual ballot that the voter filled in.
Apparently, only the Republican candidate was allowed to submit ballotsfor review.