I agree that (1) seems way too hard and overkill
Re: (2), gathering the data from other elections is a bit of a pain in itself, so also kind of labor intensive
(3) seems easiest, especially if you already understand his theory (I have only made a cursory overview).
Could MC sim a number of normally distributed (or really any distribution) of ballot counts around a mean lead. Then show that no matter how you change it the number weights hold. An extension of what he did for the 0.25 & 0.17 stuff. Seeing the same number after changing the vote counts thousands of times would be a big normie red pill.
You're talking about adding noise to the original data? This may be a good approach but I really would like a control ("non-fraud") synthetic or real dataset, to make sure that this algorithm doesn't appear to manifest in random data
I.e. looking at a random bit string of sufficient length you will eventually find a run of 100 zeros. It looks like it is clamped to zero, and by itself it is extremely improbable, but it will eventually happen by chance (almost surely or something like that)
I agree that (1) seems way too hard and overkill
Re: (2), gathering the data from other elections is a bit of a pain in itself, so also kind of labor intensive
(3) seems easiest, especially if you already understand his theory (I have only made a cursory overview).
Could MC sim a number of normally distributed (or really any distribution) of ballot counts around a mean lead. Then show that no matter how you change it the number weights hold. An extension of what he did for the 0.25 & 0.17 stuff. Seeing the same number after changing the vote counts thousands of times would be a big normie red pill.
You're talking about adding noise to the original data? This may be a good approach but I really would like a control ("non-fraud") synthetic or real dataset, to make sure that this algorithm doesn't appear to manifest in random data
I.e. looking at a random bit string of sufficient length you will eventually find a run of 100 zeros. It looks like it is clamped to zero, and by itself it is extremely improbable, but it will eventually (almost surely or something like that)