Ya dude, the script wasn't wrong the ratio data in the original json wasn't precise enough...it only went to 3 significant figures.

The above fix is decent but we will never know until we get the raw data that is the source for the aggregator itself...it really sucks that they didn't provide raw vote totals in the first place maybe as additional entries in the same or in a separate timeseries.

Interestingly enough this imprecision in the data cannot account for the live alteration of vote totals live on TV because those totals are likely based on two sources, which are the timeseries data in the json and another raw total source that should supplement this data in non-fractional form that is probably only available to companies that contract with Edison.

This means that in both sources, the ratio source and the raw total source, the numbers were switching unexpectedly in precisely the way displayed on TV in the live broadcast; these cannot be accounted for by virtue of simple rounding errors because both sources updated on screen simultaneously.

On the other hand, if CNN and ABC and so on are just as vulnerable to the same issue, perhaps this lack of precision was on purpose and perhaps it is the mechanism by which fractional vote totals are rounded and siphoned off in the tens or hundreds of votes from one candidate to the other. Your script might be capturing the real image after all, it is just a bit more nuanced than anyone previously considered.

Modified code or not, raw data has to be in integer form for these vote totals to make sense and for a pattern to emerge but as a consolation for the work you put into the script it did work as intended.

Ya dude, the script wasn't wrong the ratio data in the original json wasn't precise enough...it only went to 3 significant figures.

The above fix is decent but we will never know until we get the raw data that is the source for the aggregator itself...it really sucks that they didn't provide raw vote totals in the first place maybe as additional entries in the same or in a separate timeseries.

Interestingly enough this imprecision in the data cannot account for the live alteration of vote totals live on TV because those totals are likely based on two sources, which are the timeseries data in the json and another raw total source that should supplement this data in non-fractional form that is probably only available to companies that contract with Edison.

This means that in both sources, the ratio source and the raw total source, the numbers were switching unexpectedly in precisely the way displayed on TV in the live broadcast; these cannot be accounted for by virtue of simple rounding errors because both sources updated on screen simultaneously.

Modified code or not, raw data has to be in integer form for these vote totals to make sense and for a pattern to emerge but as a consolation for the work you put into the script it did work as intended.