DiscoGary wrote:
Harambe wrote:
...
Without knowing how every precinct reports results and counties tabulate them and the state publishes them, there's nothing anomalous about patterned data.
The first mistake would be to assume that vote reports should NOT be ordered in some way.
Would it be surprising if a bank reported its coin counting audit in multiples of 40 quarters? lol
Maybe all counties have a pattern of reporting? Only when the few late-reporting ones are de-convoluted are they obvious?
There's tons of explanations that aren't OMG FRAUD but you seem to have a prior that vote reports should obey some random distribution.
Explain that.
There are others with more knowledge of this who have dismantled you repeatedly over the days, forcing you to scurry back to find new talking points. They should do a more thorough job than I.
That logic that says it might be possible that vote dumps happened in discrete predetermined numbers because of some procedural method is sound. The problem is that these are not predetermined vote "totals". They appear to be predetermined vote "margins" for Biden.
Are you saying someone gave out the order to count the ballots until Biden had a margin of ~4,800 or a multiple and then report it? Why would they do that if all the other updates in between had margins close to zero?
EVERYONE IN THIS THREAD, READ THIS POST.
Alright, I dug through the data and I actually figured it out.
The data curation is the issue. As you mentioned in point 2, the data are reported in vote shares rounded to the nearest 0.001 (thousandth). If you think about it for a moment, that accounts for all of the statistically unlikely numbers that you are finding. If the raw vote numbers were reported, you would not see this. Unfortunately, I do not have access to raw vote numbers.
Consider this. The data are reported in the form Biden 0.XXX Trump 0.XXX, reflecting the current cumulative vote totals. Changes in votes are INFERRED from changes in the fractional share.
So, what does this look like? On 11/20 (data point 596), Trump's fraction share changed from 0.492 to 0.493, reflecting a change of 0.001. At that point, there had been 4.9 million votes, so Trump's vote change appeared to be ~4800. However, this was not the true change in votes, but simply inferred by changing from 0.492 to 0.493. The actual change was smaller, but we do not know what it is since the fractional share is rounded to the thousandths.
At data point 576, Trump's fraction share changed from 0.493 to 0.492. Once again, this appeared as a gain of ~4800 for Biden.
Why did these changes in multiples of 4800 occur near the end of the vote collection? It's because changes in vote totals were relatively smaller than the cumulate vote number as time went on. Early on, changes were more likely to be greater 0.1% because the total vote number was smaller (recall that only changes in fraction share up to the thousandth show up in this dataset). Therefore, the changes look more random. Later on, the cumulative vote total is large, so changes are only reflected in changes in thousandths of the fraction share.
So there you have it. The issue is not fraud, but rounding errors.
Please let me know if you do not understand this.