#distributions

For the version where you vote the first digit of the number of votes on the current poll, it's deterministic (up to human/computer error) because:

Based on the total number of votes at the end, say N, you can predict (without Benford's Law) how many votes each digit should have, assuming everyone votes correctly, because you know how many numbers less than N have each digit as their first digit.

For example, when I voted and reblogged there were roughly 28000 votes, and the numbers for digits 3-9 were the same, which could have been predicted from the fact that the number of numbers <28000 that start with a 3 is the same as the number of them that start with a 4, etc: they are all 1,111:

For 3 it's 1000 (for numbers 3000-3999) + 100 (for numbers 300-399) + 10 (for numbers 30-39) + 1 (for the number 3), so 1111.

Same for those that start with 4: 1000 (for numbers 4000-4999) + 100 (for numbers 400-499) + 10 (for numbers 40-49) + 1 (for the number 3), so 1111.

The number for 1 would be the same 1,111 as above except with an extra 10,000 for numbers 10,000-20,000, so total 11,111.

And for 2 would be 81,111, for the same 1111 as before but plus 80,000 for 20,000-28,000 (where 28,000 is the total number of votes).

There would be some error due to people voting incorrectly, or timing issues (someone voting before seeing the page update from the previous person's vote), but the distribution would be mostly predictable.