pouët.net

Exactly. With the current voting system the intuitive thing is to vote all prods you saw from 1-5 stars and leave out the ones you missed / can't remember etc. At least that's how I did it long time as well.
But by doing that invited entries are "punished" twice, because it's more natural to give that one annoying prod 1 star instead of leaving out the vote (because it's nowhere stated what this means and the intuitive thing would be that it's neutral then, since you didn't do anything)
Intuitive: Action (=voting has a consequence), not doing anything doesn't

*invited=unvoted (damn autocorrect)

OK I get what you mean. Though my intuition works different apparently. how can "not voting" have no consequence if we're using a system where we're adding votes :)

I guess having a page that explains voting can't hurt. Maybe changing the stars to a different icon. Not sure if that'll change anything though.

Regarding the votetools, why don't try a system similar to what we have here in pouët?
A piggy must be the default option. So that (a piggy) would be three stars. Anything below that would mean either a weak thumb down (2 stars) or a full thumb down (1 star). Same for thumb up (4 and 5 stars as weak thumb up of big thumb up). Makes sense or not?

Let me propose a counter-suggestion: If you don't rank a given entry, that entry gets the average of all of your votes in that compo. (It gives you a way to give entries e.g. 3.5, but that doesn't sound like anything that would be easily abused.) So average across entries, not across voters.

Not a bad idea, but what if you really don't like it? I think this might be harder to communicate than the current system, where (as of my knowledge) not that many people have a problem with. Also, this would elevate quite some entries to a much higher position, which might not be "fair" in some ways.

I don't think it's clear to most people that the votes are simply added up, and there's several ways to apply somewhat "neutral" votes as already suggested several times in this thread. I think your intuition is a bit different than most people, given that you wrote the thing ;-)

I was surprised to learn that no vote was worse than the lowest vote now after Evoke. I always thought it was somewhat neutral, although I'm aware of the difficulties in making a fair voting system. I think making it clear should be the first step. I also think that this, combined with how live voting works, is quite unfortunate. I suspect that we can do better by combining live voting back into the normal voting system instead of bolting it on the side.

If you really don't like it, you give it one star? As it is now, "no vote" is simply just a hidden 6th star. I don't think that eliminating that 6th star would really limit the palette of expression your satisfaction with the prods, 5 options are still plenty :)

https://keyj.emphy.de/partyvoting/ has been updated with Sesse's ideas, both the mathy t-distribution stuff and the much easier to grasp "not voted = assign the user's average for the compo" method.

Regardless of which method is used, the main takeaways for Evoke 2019 are the same: "Eisenerz" would get at least second place, and in most cases, "The Spiral" would get third place.

By the way, there was definitely some votekey abuse going on at Evoke 2019: There's a set of about 30 votekeys that always voted 5 stars for the same one or two entries in a compo (12 entries in total) and nothing else. Some of these votekeys have been registered in short "bursts" of up to 12 keys in 2-minute intervals. (No, I don't have any personal data about these "people", and even if I had, I wouldn't disclose it.)
For comparison, similar stuff has been going on at Evoke 2017 already, but at a much smaller scale (around 10 keys), and with some 3-star votes for selected entries mixed in.

So, yes, it's definitely a social problem, at least in part. As said before, we can't stop people from abusing the vote system (any vote system, that is), but we can make it slightly harder for them and at the same time more intuitive and comfortable for those who do vote consciously.

Making the UI of the current PartyMeister voting system more transparent as to what happens with uncast votes (e.g. by simply labeling the "-" or "not voted" option with "0" instead) is a good first step, but I'd still prefer a different ranking method. Any of those included in my comparison would be fine.

30 key is not exactly a trivial amount given the number of people who voted. Fuck that shit.

i gained some weight, so, lemme address the elephant in the room. maybe it's those nostalgic oldskoolers who make up their 5 star judgment after just seeing the 80s cracker group name on the beam slide regardless what utter artistic diarrhea comes next :P

god forbid you call it out too -- they'll cry disrespect, as if anyone that cracked a few lame games before I was born automatically deserves it (:

i was in fairlight, they can suck it! ;)

I personally don't think the voting system ist that broken. But you can always improve stuff.

Lug00ber's argument about remembering prods in the combo could be solved by adding a "pin" option. This way you could mark the entries you like and compare them in the 5 minutes after the compo, or when you vote later on.

Ruling out the 0 votes by keeping a minimum of 1 would reduce the impact slightly. But I highly doubt the actual benefits, since it only battles the symptom, not the cause.

LJ's idea of coupling the voting with some sort of reward (doesn't have to be merch imho) is a real good option in my eyes. Maybe something like a "Whoever got the most winners in most compos right" system. If several people are right either compare how near they're voted to the average or let luck pick the winner.

You can't rule out any manipulation at all, no matter the used system. And people snatching 30 votekeys like those at evoke are just plain assholes who deserve to get their butt spanked with a sledgehammer.

If someone actually grabbed votekeys, well, shame on them. If people bought votekeys from others or something… well, there's always been cheating.

I don't think we should introduce a Keynesian beauty contest by adding prizes for voting “right”; it's not likely to pick better winners. At Solskogen, we gamify the voting process by putting up the number of people currently live voting at the compo slide; it's a good reminder, it adds societal pressure, and everybody wants that number to go up, so voting gives your brain that little reward. (There could be adjustments to exactly how we do it in the future.)

http://hype.retroscene.org/blog/events/839.html some extra statistics

Have fun telling people to at least vote one star. In this scenario, having the -1 back might be the better solution IMO.

Which would actually make "0" the default / "did not vote" option. if i really want to vote i can do something and if I really hate it, i can do so, too.

Not sure if it makes sense, but options where we have to TELL the visitor to take action will probably not work, because you can't really influence their behaviour like that.

so far this is the most sensible idea here

So, Truck has to visit _all_the_pardeyz_ :)

Yes, and that's exactly why many users here would prefer a truly neutral option as a default.

Ok. Question is, which party is going to want to test it?

Generally, when I first used the 5-point-system, I basically extrapolated from the old "rank the first three entries" way from back then and never thought about using something else than just adding up.

But yeah, it should be doable to to integrate a bayesian average voting option in PM3 and have the organizers switch it (or even show both). In the end it's up to the organizers what they want to use and how to apply it.

https://github.com/Gargaj/wuhu/blob/master/www_admin/votesystem.inc.php - FWIW Wuhu has all the vote systems separated and you can add your own easily.

I experimented a bit with a new "prediction" option over at https://keyj.emphy.de/partyvoting/. The basic idea is to keep sum-based ranking, but fill the "holes" of unvoted entries in a smart way that tries to predict what the user would have voted.
The general idea is to assign the average score of all voters for that entry, but that alone would be too simple: If a user has a general tendency to vote very high or very low, the average across all users wouldn't fit with their typical voting behavior. So a correction is applied, by comparing the user's average vote score with the average of all users, and applying the difference as an offset to the predicted vote.

Currently, I tend to prefer these "vote prediction" methods (which includes Sesse's earlier proposal), but unfortunately these are slightly harder to implement: You need a full entry/uservote matrix to work on, a simple "select sum(score) from votes" query is no longer sufficient. In that light, I'd be fine with simple average-based ranking too; as said before, anything is better than the current method.

it's not "prediction", it's average/regression imputation. maybe if you'd start altering the significance of a vote based on the amount of missing values, it will become "prediction". then again, you'd make it more a mathematical exercise than a "vote" at that point which would obscure voting even beyond the point of what we expect voting to be :)