Main Points
Please read through this entire post to prevent the spread of misinformation.
I am happy to announce it's finally here: Our new scoring system to discount votes made by rating troll (aka duplicate) accounts!
This system has been in planning for 6 months. Before Chihayafuru 3, before Ishuzoku Reviewers.
For 10 years prior to this, these accounts were dealt with by admins via the Illegitimate Accounts thread.
The scores for all entries will be recalculated over the next 7 days.
During this time, please note that scores and ranks will be fluctuating for anime/manga entries which do not meet the following:- Top 50 Anime/Manga (adjusted on release)
- Currently Airing Anime (adjusted 3~4 hours after release)
The only change visible to users will be the weighted score. All other statistics will include rating troll counts (to make it more difficult to break the system). Furthermore, entries with insufficient votes will display N/A to dissuade manipulation.
Top 50 Charts
Anime
It's long been suspected that up/down votes approximately equal themselves out for most Top Anime, and this is confirmed after recalculation.
All scores in the Top 50 change by less than 0.10- 38 anime increased in score, from 0.0016 to 0.0758.
- 12 anime decreased in score, from -0.0003 to -0.0241.
- Most affected Top 50 Anime are Ginga Eiyuu Densetsu (0.0758) and Gintama° (0.0723). Regular rating troll reporters won't be surprised by this.
There is a little movement in the Top 10, mainly due to the scores all being very close together.
Manga
The Top 50 Manga show more changes in scores than Top 50 Anime.- Series typically have fewer members than Top Anime, which makes them easier to influence.
- Rating trolls for Top Anime often also upvote the manga, while the downvoters only target the anime.
- Most affected Top 50 Manga include Nogi Wakaba wa Yuusha de Aru (-0.4731), Monogatari Series: First Season (0.1916), and Re:Zero (-0.1420). The first is well-known for having been upvoted.
A bit more movement in the Top 10 here, but again, mainly due to the scores all being very close together.
Series Most Affected
After we first decided how to define illegitimate (duplicate) accounts from legitimate, we then applied that method to the entire anime database to ensure there wouldn't be any big surprises. From this data, I pulled entries with score changes of >0.10 and sizable member counts (>10k). These scores were recalculated again on January 31 and are presented below.
Half of these entries are movies or OVA related to typically up/down voted main series, and show much larger changes due to their low member count. Aside from the very obvious Chihayafuru 3 score increase, we see the continued trend of the Ginga Eiyuu Densetsu downvoting being corrected as in the Top 50. Other series we already knew were hit by rating trolls (e.g. Symphogear XV) also appear here, giving confidence that the system is working well.
Vote Brigading
Last week, a YouTube account posted a video instructing 1M+ subscribers to go to MyAnimeList, register an account if necessary, and vote Ishuzoku Reviewers a 10. This was followed by a second YouTube account, also with 1M+ subscribers. When an individual leads an organized voting raid on a website, instructing his followers to vote yes/no/10/1/5/green—and 15,000+ votes of that specific value come in less than 12 hours, the entry's score has been compromised due to vote brigading.
If you post in this thread saying that this is not vote brigading, that Ishuzoku Reviewers was not vote brigaded, or initiate any other circular arguments relating to the voting brigade (and YouTube videos made in the last week about it), you risk receiving a trolling and/or spam ban.
You had 50+ pages in the series discussion board to argue about it. That topic is not to be discussed here. This thread is about the illegitimate (duplicate) account system and should focus on that.
To combat vote brigading, we must have two systems at work:- The regular system to discount duplicate (illegitimate) accounts.
- An extra algorithm to discount scores given from legitimate accounts due to the brigade/meme.
Brigaded series cannot be calculated the same way as normal entries. Extra measures must be taken to try and discount both up and down votes given due to the brigade/meme + aftershock.
When the voting rush first began, I looked for a cache of the entry, and found this page:
If you compare the data on this page with the day this post is being written (February 10), you get a distribution graph that looks like below. Applying our rating troll system with our correction for vote brigading, we then get a weighted score of 7.57—nearly identical to the score from January 31.
The score is not frozen or forced to ~7.6.- If the community collectively agrees, the score can still rise to 8+.
- Both up and down votes from the brigade/meme were handled equally.
- No accounts involved in the brigade were banned or deleted. If these "brigading accounts" continue to watch and vote normally, their scores will be counted like all normal accounts.
Pingu in the City will also have its score adjusted for the vote brigading it received in 2017, moving from its current 8.43 to 6.80.
Likewise, Pingu in the City (2018) moves from 7.80 to 6.68.
Summary
- Please read through this entire post to prevent the spread of misinformation.
- This system has been in planning for 6 months; it was not "added because of an ecchi anime".
- Scores will be recalculated over the next 7 days. During this time, please note that scores will be fluctuating for many anime/manga entries.
- Only the weighted score will change; all other statistics will include rating troll accounts.
- It's long been suspected that up/down votes approximately equal themselves out for most Top Anime, and this is confirmed after recalculation. Top Manga shows a bit more change due to fewer members and Top Anime influence.
- If you post in this thread saying Ishuzoku Reviewers was not vote brigaded, or initiate any other circular arguments relating to the voting brigade, you risk receiving a trolling and/or spam ban. That topic had 50+ pages to run already and will not monopolize this thread.
- Ishuzoku Reviewers' score is returned to ~7.6—nearly identical to the score from January 31. This is not "frozen" and can still rise to 8+ if voting and account behaviour proceeds as usual.
- No accounts involved in the brigade have been banned or deleted and can still be used like normal.
Finally, the rating troll scoring system is not fixed, done, never to be looked at again. We anticipate malicious individuals will continue to try to break our new system and create even more accounts to manipulate votes in the future. From here, we will be monitoring for usual changes in scoring patterns and will attempt to deal with illegitimate accounts before they grab a foothold. If they do manage to get past us, we will attempt to correct it as speedily as possible.
Entries will still be vulnerable to YouTube voting raids due to the participation of legitimate user accounts. If new entries receive organized voting brigades, we will apply the second system to these entries (to readjust their scores) and will inform the community that we have done so. Please let Ishuzoku Reviewers be the last of it.
Thank you to everyone who took the time to report illegitimate accounts over the years and to give suggestions on how we might fix it. And thank you to the community at large for waiting patiently until now.
Edit: I thought the "Series Most Affected" titles would update on release, but seems they can take up to ~7 days to adjust as well. Chihaya should update faster since its still airing.
Edit2: Fixed some typos. |