Found one of Poindexter's old thread that may have foretold 2020's possible election fraud:
Is It Time For Algorithms That Flag Potential Election Fraud? -- Please Answer Both Questions.
Algorithms exist for almost everything else from credit card theft to consumer preferences. So why not for election fraud? Not to change or negate counts, but to flag highly unlikely results for further investigation.
An article I read a few days ago got me to thinking about this... it said something to the effect that new ballots were discovered in Florida that favored democrats 3 to 1 in an area that was previously 2 to 1.
There may very well be legitimate reasons for the shift. But if not, how likely is that due to chance?
Please provide your
intuitive responses to the following two questions before I post the answers (or you calculate them). Elections are not random processes like drawing colored balls from urns and this is an artificial oversimplification. But it does serve a useful purpose that I'll explain after replies are posted.
Again, I'm looking for intuitive responses only. Please do not calculate anything.
Situation: CNN County has an endless supply of voters and they favor democrats 2 to 1. Voting has ended. 500,000 votes were cast that favored democrats almost exactly 2 to 1.
Question 1: 1,000 more votes are found. Assume all 1,000 voters who cast them are a random sample of the CNN County population. What is the probability that those 1,000 votes favor democrats 3 to 1 or better?
Question 2: 10,000 more votes are found. Assume all 10,000 voters who cast them are a random sample of the CNN County population. What is the probability that those 10,000 votes favor democrats 3 to 1 or better?
https://www.elitetrader.com/et/thre...on-fraud-please-answer-both-questions.326973/
This thread is three pages long. The answer is here:
Nobody wants to take a shot at the answers? OK. Here they are. With a brief preamble.
Let's drop the strawmen. This is not about trying to "prove" fraud in Florida. But news reports from Florida got me thinking and I started this thread to show how counter-intuitive vote count probabilities can be.
Why? If votes are suddenly "found" that deviate significantly from overall ratios (e.g. from 2:1 to 3:1) there had better be good reasons. Because the odds of it happening by chance alone are astronomically small.
If liberal Unicorn Village's sealed ballots were accidentally misplaced for a few hours, no problem. But a suspicious scenario would be if a large number of ballots breaking 3:1 or more are suddenly "found" that had been randomly set aside throughout the day because of random machine breakdowns in otherwise 2:1 areas.
The problem,
as stated in the OP, is answered using the binomial CDF. For perspective I'll answer questions 1 & 2 in terms of the odds of winning the Powerball Jackpot (approx. 1 in 292,000,000 with one ticket).
Answer to Question 1: 333 expected republican votes versus 250 or less actually obtained is no big deal, right? Wrong! You're more likely to win the Powerball Jackpot with just two tickets.
Answer to Question 2: 3,333 expected republican votes versus 2,500 or less actually obtained is nothing to be concerned about, right? That's even less likely... a lot less. You have a better chance of winning the Powerball Jackpot 8 times in a row with just one ticket per drawing.
If you object to the problem as stated in the OP because in real life we'd be sampling without replacement from a finite population, fine. In that case the hypergeometric distribution would apply. The probabilities don't change much and the above answers would still be correct.
https://www.elitetrader.com/et/thre...wer-both-questions.326973/page-3#post-4760659