pouët.net

Go to bottom

How should entries be ranked?

category: parties [glöplog]
if nobody votes or only those whose choice does not represent the felt truth of the majority even the best system will fail.
@HellMood: If you want to be difficult, I can, too
added on the 2019-08-23 18:07:28 by Sesse Sesse
Quote:
With AVG scoring system, "Bold" would be the "best" demo on pouet (highest upvotes on all perfect 1.00 avg entries), disregarding all the nonvoters, and of course only if potential trolls resist to give it a pig right now ;)

And finally i know why i don't win compos more often ... the voting system! i knew it all along ^^

No: https://cardboard.pouet.net/wilson_score.php
added on the 2019-08-23 18:09:50 by Gargaj Gargaj
Note that the main reason this thread was started would not exist if voting would only be enabled after the compo as it used to be in earlier days:
Ditch live voting, problem solved.

Also, reviewing entries after the compo is much easier if you can fetch them from a party FTP, something lacking for a sadly increasingly high number of parties.
added on the 2019-08-23 22:05:09 by T$ T$
ditch voting completely and let truck decide.
added on the 2019-08-23 22:11:32 by nosfe nosfe
or Paul the Octopus
or Maalis Grandma
added on the 2019-08-23 22:55:55 by v3nom v3nom
FYI, I updated https://keyj.emphy.de/partyvoting/ with results from Evoke 2017 (which didn't have any noticeable voting result irregularities) and added the -1..3 system that has been proposed by a few people. I'm still waiting for data of other parties.

Wilson score looks very interesting too. I'll add it when I wrapped my head around the heavy math that's involved, or find a ready-to-use formula that works with 1..5 ratings.
added on the 2019-08-23 23:07:03 by KeyJ KeyJ
v3nom: i'd be happy bring her as jury, as long as you'll supply her with enough port wine! it's her vice... well, vice.. she's close to 90 now, so :P
offtopic:
Quote:
One more data point to consider: With Breakpoint's -1...5 voting system, there was probably only one entry that ever achieved a negative score, and that was because it somehow ended up in the wrong compo.
Yes! It is still one of my best works! If I remember correct no one got the fact that my release was the recording of a live event. Arranging the flowers was the wild entry.

I'll do a more indepth replay later. Sorry for this offtopic reaction, do carry on.
added on the 2019-08-23 23:55:18 by numtek numtek
mind you the reputation of the releases from any given party will mostly be completely off from the on-party rankings. at the party you have company, booze, sudden judgement call and a tendency to lean towards anything that is 'partylike'.
f.ex. out of a flawless, top-of-the-shelf downtempo tune and a slightly-better-than-average funky chip house bit, the latter will win at a music compo that's for sure.
added on the 2019-08-24 01:21:00 by nagz nagz
id take (and vote for) your downtempo tune any compo!
added on the 2019-08-24 02:12:34 by wysiwtf wysiwtf
the scene is not about making things better. it is about making better demos. at least that's what i have learned.
added on the 2019-08-24 03:20:38 by wertstahl wertstahl
Quote:
id take (and vote for) your downtempo tune any compo!


aw <3
added on the 2019-08-24 07:38:41 by nagz nagz
maali: if you bring her to deadline, she can be the jury. Port wine is the of course the least thing we can provide
added on the 2019-08-24 12:51:04 by v3nom v3nom
She would of course not need to judge all compos, that would be a bit too much to be asked for, but rather a compo of her choice. I'd say her judgement probably wouldn't be worse than the one of a bunch of drunk sceners ;)
added on the 2019-08-24 12:57:17 by v3nom v3nom
In order to make sure it's fair, I propose that Maali not be allowed enter anything in the compo his granny is judging. It would be too cruel to see him loose that one.
KeyJ: You cannot use the Wilson score interval with -1..5 ratings; it is purely for 0/1. But you can make a claim that it is sort-of normally distributed (being pretty close to continuous and *wave hands*). If so, what you want is a t-distribution interval:

C_low = avg - t(0.95, n-1) * S / sqrt(n)
C_high = avg + t(0.95, n-1) * S / sqrt(n)

where n is the number of votes, avg is the average of the votes, t(alpha, n) is the Student's t cdf (just call a math library), and S is the sample standard deviation, calculated as follows:

S = sqrt(sum((x_i - avg)²)) / (n-1)

Occasionally, you will get nonsensical results, like if you have only two votes and they are both 5, the upper end of the confidence interval will be above 5. Such is the nature of approximations.

(I still think Condorcet is the only sane voting choice, but it doesn't produce a single score that you can brag about.)
added on the 2019-08-24 18:16:15 by Sesse Sesse
Since I have a lot of data on voting from the last 15 years I can maybe weigh in on some things:

Generally vote participation has increased in the last 10 years due to the following changes:

1. Giving out the votekey at the beginning of the party instead of receiving it after the last compo in exchange for your wristband.

2. Switching from ranked voting to a point system, probably because it removed harder choices on who might be deserving of places 1-3. Btw this Also helped smoothing over the name voting bias. Or the fact that we're all older and vote more responsibly.

3. Introduction of live voting

Generally we went up from 50% participation at Revision 2011 to over 80% this year.

So for me the picture is clear that we have dramatically improved participation with the changes we started implementing at the ultimate meeting 2003 with the point system.

I think the problem is social, as gargaj suggested. After a long day of compos, talking and drinking, many people just don't care about / have the energy to check or update their votes.

We can think about better informing people what voting means and how it is applied. For example that you can change all votes up until the voting deadline and how things are calculated.

Btw 0 points are unbiased. It's just 0 and shouldn't obviously be counted into a average :) I thought a lot about reintriducibg the -1 because you can actually punish really really tasteless or fake entries. Though the results will mit change much unless you switch to an average voting based result.
added on the 2019-08-24 20:12:17 by D.Fox D.Fox
Small addition: party voting is and has always been a popularity vote and not a quality vote. So if the people chose not to vote and drink instead and not checking their votes then that was their choice. All solutions I have read have either been tried before or make things more complicated for the voter and will therefore hurt participation.

What we as organizers can do is educate and try to minimize technical issues.
added on the 2019-08-24 20:18:34 by D.Fox D.Fox
D.Fox: What do you think about making it stand out more than not voting on an entry = worse than giving 1 star with the current system? (just the general idea, without considering how this should be conveyed)
Quote:
Giving out the votekey at the beginning of the party instead of receiving it after the last compo in exchange for your wristband.


Ah, that awkward moment when someone is leaving the party and then shakes hands with you saying "I really enjoyed your prod, great work" while you notice that the wristband is still in place. :]
added on the 2019-08-25 11:18:41 by ham ham
They should be ranked by their content and not their character.
added on the 2019-08-25 12:42:19 by Zplex Zplex
Quote:
D.Fox: What do you think about making it stand out more than not voting on an entry = worse than giving 1 star with the current system? (just the general idea, without considering how this should be conveyed)


Isn't it already worse because it's 0 points instead of 1?
added on the 2019-08-25 14:49:39 by D.Fox D.Fox
What I mean is making it more obvious to the voter that this exactly is the case. From the responses in this thread it seems clear to me that many people are not actually aware that not voting on a prod is worse than giving it 1 star.

login

Go to top