I put together a spreadsheet to help you understand what I think that rude zip person was getting at, -- but also, using zip's methodology to additionally break down the 17k anomaly. https://gofile.io/d/WYBsAz

Did you actually intend to highlight the PA:2020-11-04T04:08:51Z 17k source transaction? If so, perhaps zip thought you were pointing out that one as an anomaly. All I think rude-zip-person was saying is that the lowest possible share values that would round to the Edison numbers would result in a lowest-possible added vote diff (highlighted in yellow) that would fit inside the cumulative added vote count (highlight in green), therefore THAT transaction is not anomalous.

I spotted what seems to be a problem with your best-fit method (highlight in purple). It settled on 0.5625 which would round up, not down to Edison's 0.562 .... are you assuming Edison values are chopped/truncated/floor/whatever... or rounded-from-half? Rude-zip-person thinks rounded and almost gets it right though the numbers quoted indicate a lack-of-decimal precision problem too. (Used 0.561499?) For 0.562 the inclusive range would be 0.5615 to 0.562499999999...

That said, the 17k shown as three transactions in the next section of the spreadsheet is indeed odd.

• (1) Why send an update with only 54 votes?
• (2) Is this the result of modifying the shares or the underlying numbers? With those ratios (T=0.566 B=0.42 --> T=0.56 B=0.426) I can imagine someone thinking, "Let's move that last 6 from Trump to Biden." But no, that would result in the next transaction 'stealing' those votes back again and it looks like a normal distribution. So it must have been the underlying numbers.
• (3) I wonder if the cumulative vote total (2984522) was incorrect. If it was typed in by a human and one or more digits were mistyped, meaning the added votes was actually greater than 54. what could they be? Out of time right now but I will experiment.

I always suggest text posts for collaborative work because you can edit the text to fix links and typing errors in text, add new stuff. Thanks for your due diligence.

I put together a spreadsheet to help you understand what I think that rude zip person was getting at, -- but also, using zip's methodology to additionally break down the 17k anomaly. https://gofile.io/d/WYBsAz

Did you actually intend to highlight the PA:2020-11-04T04:08:51Z 17k source transaction? If so, perhaps zip thought you were pointing out that one as an anomaly. All I think rude-zip-person was saying is that the lowest possible share values that would round to the Edison numbers would result in a lowest-possible added vote diff (highlighted in yellow) that would fit inside the cumulative added vote count (highlight in green), therefore THAT transaction is not anomalous.

I spotted what seems to be a problem with your best-fit method (highlight in purple). It settled on 0.5625 which would round up, not down to Edison's 0.562 .... are you assuming Edison values are chopped/truncated/floor/whatever and or rounded-from-half? Rude-zip-person thinks rounded and almost gets it right though the numbers quoted indicate a lack-of-decimal precision problem too. (Used 0.561499?) For 0.562 the inclusive range would be 0.5615 to 0.562499999999...

That said, the 17k shown as three transactions in the next section of the spreadsheet is indeed odd.

• (1) Why send an update with only 54 votes?
• (2) Is this the result of modifying the shares or the underlying numbers? With those ratios (T=0.566 B=0.42 --> T=0.56 B=0.426) I can imagine someone thinking, "Let's move that last 6 from Trump to Biden." But no, that would result in the next transaction 'stealing' those votes back again and it looks like a normal distribution. So it must have been the underlying numbers.
• (3) I wonder if the cumulative vote total (2984522) was incorrect. If it was typed in by a human and one or more digits were mistyped, meaning the added votes was actually greater than 54. what could they be? Out of time right now but I will experiment.

I always suggest text posts for collaborative work because you can edit the text to fix links and typing errors in text, add new stuff. Thanks for your due diligence.

I put together a spreadsheet to help you understand what I think that rude zip person was getting at, -- but also, using zip's methodology to additionally break down the 17k anomaly. https://gofile.io/d/WYBsAz

Did you actually intend to highlight the PA:2020-11-04T04:08:51Z 17k source transaction? If so, perhaps zip thought you were pointing out that one as an anomaly. All I think rude-zip-person was saying is that the lowest possible share values that would round to the Edison numbers would result in an lowest-possible added vote diff (highlighted in yellow) that would fit inside the cumulative added vote count (highlight in green), therefore THAT transaction is not anomalous.

I spotted what seems to be a problem with your best-fit method (highlight in purple). It settled on 0.5625 which would round up, not down to Edison's 0.562 .... are you assuming Edison values are chopped/truncated/floor/whatever and or rounded-from-half? Rude-zip-person thinks rounded and almost gets it right though the numbers quoted indicate a lack-of-decimal precision problem too. (Used 0.561499?) For 0.562 the inclusive range would be 0.5615 to 0.562499999999...

That said, the 17k shown as three transactions in the next section of the spreadsheet is indeed odd.

• (1) Why send an update with only 54 votes?
• (2) Is this the result of modifying the shares or the underlying numbers? With those ratios (T=0.566 B=0.42 --> T=0.56 B=0.426) I can imagine someone thinking, "Let's move that last 6 from Trump to Biden." But no, that would result in the next transaction 'stealing' those votes back again and it looks like a normal distribution. So it must have been the underlying numbers.
• (3) I wonder if the cumulative vote total (2984522) was incorrect. If it was typed in by a human and one or more digits were mistyped, meaning the added votes was actually greater than 54. what could they be? Out of time right now but I will experiment.

I always suggest text posts for collaborative work because you can edit the text to fix links and typing errors in text, add new stuff. Thanks for your due diligence.