Comment on Reminder: Three Days Left to Vote for the 2024 Lemmy Anime Awards
wjs018@ani.social 2 months agoIt does. There was a bit of a discussion about this in the voting announcement thread here. Basically, I wanted to do ranked choice voting, which isn’t as punishing to shows that don’t get a vote, but that type of question didn’t work in some browsers, so we fell back to the STAR voting. It should still result in more accurate results than a single, non-transferable vote though. It’s something I want to fix for the next time we do this (if we do).
Endmaker@ani.social 2 months ago
If you implemented the voting algorithm (STAR in this case) in the code instead of just using a library, consider using this more generalized variant of STAR; not sure if there is a name for it:
Instead of scoring candidates from 0 to 5, allow voters to give scores between ‘m’ and ‘n’, where ‘m’ and ‘n’ are integers. If the voter choose not to give a score to a candidate, assign a default score of ‘k’ where ‘k’ is an integer and ‘m <= k <= n’
For example,
What I liked about STAR - that I’m unsure if ranked-choice voting methods can achieve - is how expressive a voter can be.
For example, if I like anime A way more than anime B and hence want to give A a much higher chance of winning, I can express it in my scores (e.g. by giving A a score of 3 and B a score of 0)
Another case is if I absolute detest an anime C, I would be able to give it a score of -2.
To conclude, I don’t think other commenters’ complaints are about STAR itself, just its parameters of m, n and k.
If these parameters can be properly tuned such that unvoted candidates get a middle score instead of the lowest score, that should solve this problem where they appear to be voted against.
P.S. I’m still of the opinion that these awards are just another form of popularity contest, so if a candidate doesn’t get voted for because it was barely watched by majority of the voters, that means most voters don’t really want it to win anyway.
wjs018@ani.social 2 months ago
Pretty much what they are guaranteed to be. If people are looking to things like this to uncover hidden gems, then they are looking in the wrong place. Something like a critic’s awards would be better (or something like the jury selections from the reddit awards).
Your idea of modifying the default behavior of STAR is interesting. I don’t think I will have time to whip that together before the end of this awards campaign, but it is something you could kind of simulate by replacing the scores of all the blanks as the average of the non-blanks. That moves any blank responses closer to the
kvalue in your examples.I plan on releasing all the votes and the code I used to analyze them in a github repo so people can use their own tools to visualize or analyze the votes as they see fit (at least for anime + manga; ln’s are a bit trickier due to the low number of responses). So, we could essentially re-run the votes using alternate systems to see how things may or may not have changed.
For the sake of this awards season, I am just going to count blanks as a 0 since that is the default STAR behavior and what is implemented in the library I am using.