[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/sci/ - Science & Math

View post   
File: 430 KB, 1572x1422, rate.png [View same] [iqdb] [saucenao] [google]
15026821 No.15026821 [Reply] [Original]

Assuming that in the past, people voted somewhat honestly. But lately, increasingly, people want their liked works (even if they're not 10/10s to them) to "beat" other works, and vote 10/10 nonstop. This means newer works will be artificially skewed much higher than classic works, thus burying old classics.

How would you fix this? Find a curve based on inflation by time and invert it? Would it be linear or a log or something?

>> No.15026822

>>15026821
>. But lately,
Anon I...

Guess what, just ignore the internet rankings. People in the internet is so retarded that MAL had to hide review scores.

>> No.15027251

>>15026821
>This means newer works will be artificially skewed much higher than classic works, thus burying old classics.
Bleach is technically a classic at this point.

>> No.15027280

>>15026821
Show a percentage of the amount of users that gave a given anime a score higher than the mean score of their entire anime list

>> No.15027298

>>15027280
Sounds random. What would that do?

>> No.15027309

>>15026821
It's simple. Rate Steins;Gate the best. All other candidates are irrelevant

>> No.15027447

>>15027298
Show that gintama is based

>> No.15027452

>>15027447
>gintama
>implying we should try to show gintama is based when https://pastebin.com/H0zFNc9h

>> No.15027492

>>15026822
You are not on /a/. The point of /sci/ isn't the validity of the online anime opinions, it's the puzzle. Pls contribute to the puzzle or don't at all. Science isn't about why, it's about why not.

I'd say a good balance between simplicity and effectiveness would be to switch to a 100-point rating scale, and then weight each user's rating for a show based on, say, the average of the differences between each pairing of shows they've rated. That way users with more uniform (and just more) rating tendencies, who aren't giving out three 100s every season and/or didn't start two years ago, count for more.

>> No.15027690
File: 146 KB, 400x400, a9d9b93564afcdb76f5a761eead945b4f0c19284.jpg [View same] [iqdb] [saucenao] [google]
15027690

>>15027309

>> No.15028731

Other anons have proposed good ways of weighting reviews to de-value the zoomie shit muncher opinions.

I suspect in terms of discovering people with similar tastes you could use the heuristic they've proposed (tendency to give lots of new shoes 5 stars) to remove them from the data entirely, and only focus on more based people