RogerDodger, can you explain whether & how the code adjusts one person's vote according to their average vote? I see from the hugbox that some people give out mostly very high scores, and some give mostly very low scores. So an adjustment to normalize their scores might make sense. But some people might be reviewing only stories they liked. They should know that their scores will be adjusted down and hurt those stories more than help them, if the scores are normalized.

3550381
Votes aren't normalised.

I considered doing so in the past, but there isn't any reasonable way to do it. Because the votes would need to stay within the range 0–10, you can't just stretch the numbers for anyone who has given at least one 0 and one 10 (which most people do). If on the other hand you enforced a certain kind of distribution, that's making assumptions about the data.

It's mostly not an issue, since most people vote on most entries, so any bias is consistent across their record.

Someone's hugbox score (arithmetic mean of votes) is probably not as interesting as the variance in their votes.

3550889

Someone's hugbox score (arithmetic mean of votes) is probably not as interesting as the variance in their votes.

The simplest standard normalization process is to set each score to

score_normalized = (score - score_mean) / standard deviation

where mean & stdev are calculated for each user.

Another normalization procedure is to sort all of the scores the user has given, then cut that set into N bins with an equal # of scores in each bin, and assign scores from 0 to N-1. This is useful for Netflix data, because users can assign scores 1-5, but almost never use a score of 2.

I agree there's no right answer.

What is a hugbox?

3555037
Hugbox score is your average score given to other stories in voting.