×
  • remind me tomorrow
  • remind me next week
  • never remind me
Subscribe to the ANN Newsletter • Wake up every Sunday to a curated list of ANN's most interesting posts of the week. read more

Forum - View topic
The Best and Worst of Winter 2019, Jan 1-21




Note: this is the discussion thread for this article

Anime News Network Forum Index -> Site-related -> Talkback
View previous topic :: View next topic  
Author Message
Shay Guy



Joined: 03 Jul 2009
Posts: 2091
PostPosted: Mon Jan 28, 2019 2:29 pm Reply with quote
W'z taking the Chaos Dragon slot as is its rightful due, I see.
Back to top
View user's profile Send private message
MarshalBanana



Joined: 31 Aug 2014
Posts: 5294
PostPosted: Mon Jan 28, 2019 5:59 pm Reply with quote
I'm surprised at how high Fairy Tail is.
Back to top
View user's profile Send private message My Anime My Manga
zrnzle500



Joined: 04 Oct 2014
Posts: 3767
PostPosted: Mon Jan 28, 2019 9:41 pm Reply with quote
Shay Guy wrote:
W'z taking the Chaos Dragon slot as is its rightful due, I see.


Yeah, I think W'z has a pretty good grasp of that spot, unless Karakuri Circus really tests its remaining fans patience.

MarshalBanana wrote:
I'm surprised at how high Fairy Tail is.


While I can't say I don't understand why people would be surprised, it did pop into the top 10 of the weeklies 5 times last season. At what point will people not be surprised? I do think part of it is that this season overall is weaker than the last one, with Slime and Run with the Wind more comfortably in the top ten, though the other carryovers (aside from the aforementioned) are at about the same place as they were last time.
Back to top
View user's profile Send private message
Fred Lougee



Joined: 01 Oct 2018
Posts: 127
PostPosted: Tue Jan 29, 2019 2:05 am Reply with quote
zrnzle500 wrote:
Shay Guy wrote:
W'z taking the Chaos Dragon slot as is its rightful due, I see.


Yeah, I think W'z has a pretty good grasp of that spot, unless Karakuri Circus really tests its remaining fans patience.



Um...that should be "...remaining fan's (singular, possessive) patience."
Back to top
View user's profile Send private message
S0crates



Joined: 06 Jul 2018
Posts: 227
Location: Banned - Noticed our poor ethics
PostPosted: Tue Jan 29, 2019 6:56 am Reply with quote
I think you need to adjust the filler ranks here to a better degree. Shows shouldn't get punished for not having an episode a week. Look at the shows below Love is War, if you take the estimate of "score of average ranking between episode 1 and 2 as the filler" then all that happens is that Sword art And Fairy Tail gets pushed down (as there's no reason for them to be pushed up to begin with). The fact that Fairy tails runs past Shield Hero, even though Shield Hero Outscores them in both of their episodes, on the cumulative ranking would illustrate this inaccurate representation as well (same principle for My Roomate is a Cat).

It's just something to take into account. Personally I'd just place the "filler episode" as a default relative score (based of what's between them on the ranking from those two weeks before and after the week in question), if you need a filler slot. For example, "Love is War's "filler point" would obviously be "whatever would have been required for #7 in week 2". My Room Mate is a Cat would in contrast be "Somewhere between the previous ranks relatively".

This approach is especially clear in the weekly jump line. Almost all of the ones with episode in the middle Takes a severe jump above the shows that did not air. If you flateline those series and only make them compete for that week instead it's a more fair representation overall.

Illustration:


Surly this looks more representative?

This would also impact the cumulative rating(?), not to mention the cumulative rating's matrix has to be bugged. Look at Run with the Wind overtaking Dororo (and Neverland for that matter). Dororo outscored RwtW every week, so that's just not possible.
Back to top
View user's profile Send private message
zrnzle500



Joined: 04 Oct 2014
Posts: 3767
PostPosted: Tue Jan 29, 2019 5:08 pm Reply with quote
@S0crates The problem with putting the blank point in between where the previous week's episode and the next week's episode is that that only works when there is a week after, which is to say that it only works for the first entry of the season. Every other week where there isn't an episode, you could only compare it to the previous week's episode, which is exactly what they do now. I don't know that people put much weight on those points anyways, as they are just place holders which have no bearing on the cumulative ranking, as it states at the bottom of the page.

More generally, I think you misunderstand what the rankings are based on. They are not based strictly on the raw average ratings that you see on each episode page, but rather how people rated each show as compared to how they rated other shows, as explained in the Notes section at the bottom. Even if one show is on average rated 0.1 or 0.2 stars more than another show, that doesn't necessarily mean that the former is preferred over the latter, and for shows that close in rating (such as any of the ones you named) it is not impossible that the latter is preferred over the former.
Back to top
View user's profile Send private message
S0crates



Joined: 06 Jul 2018
Posts: 227
Location: Banned - Noticed our poor ethics
PostPosted: Tue Jan 29, 2019 5:56 pm Reply with quote
zrnzle500 wrote:
@S0crates The problem with putting the blank point in between where the previous week's episode and the next week's episode is that that only works when there is a week after, which is to say that it only works for the first entry of the season. Every other week where there isn't an episode, you could only compare it to the previous week's episode, which is exactly what they do now. I don't know that people put much weight on those points anyways, as they are just place holders which have no bearing on the cumulative ranking, as it states at the bottom of the page.

More generally, I think you misunderstand what the rankings are based on. They are not based strictly on the raw average ratings that you see on each episode page, but rather how people rated each show as compared to how they rated other shows, as explained in the Notes section at the bottom. Even if one show is on average rated 0.1 or 0.2 stars more than another show, that doesn't necessarily mean that the former is preferred over the latter, and for shows that close in rating (such as any of the ones you named) it is not impossible that the latter is preferred over the former.


Well, it says the Schulze method right on the cover, there' not much to misunderstand. The reason for why I mention it (with a question mark) is because the cumulative ranking is obviously miscalculated. Look at the numbers for Dororo and Run with the Wind. No matter what you do the math is just wrong. There is no way that Run with the wind can pass Dororo when Dororo always ranked higher.

Same could be said for several other shows, I'm just showing the obvious example everyone can pinpoint right out of the gate. Dororo has always been top #3, it can't possible drop to #4 in the cumulative ranking. At the same time Run with the Wind has never been #2 (or #1), yet it jumps up to it in the cumulative ranking.

Also your first statement is also irrelevant to the issue I'm bringing up. I'm just saying you can use the third week as a rule in the opening as a means to make it more "fluid", but even if you only use the first week the results should be closer to what I'm showing than what is being presented. The most clear way to show this is that you have several episodes that "don't have an episode" the same week that still manage to climb in the ranking comparatively. Look at My Roommate is a Cat/Rise of a Shield Hero and compare it to Magical Girl Spec-Ops Asuka.

Both of those series dropped below Magical Girl Spec-Ops Asuka in the weekly rating for week 2, when none of these 3 shows had an episode. If it truly is as you say, then that's impossible since both of them ranked better than Asuka the week before (and after). This is not me having any agenda of pushing those shows up (as a matter of fact I found the "My Cat is a Roommate" show to be rather dull), but the mistakes are just too visually clear as day.

I assume this presentation material got rushed out this time around, and that mistakes were made. I'd probably have someone look over the math if I were the crew at ANN, (and I urge picking up the tips regarding "empty week episodes", at least once the episodes after it is released so it can change the week after if needed), because as things stand now the math is just clearly wrong and is not following the methods give in the description. I don't have to know the raw data to see that, you just have to use process of elimination.
Back to top
View user's profile Send private message
zrnzle500



Joined: 04 Oct 2014
Posts: 3767
PostPosted: Tue Jan 29, 2019 7:07 pm Reply with quote
^If those were the only shows in the list, that may be a concern, but there are more shows in the mix, and we need to consider what happened with them, as the ranking is based on all of the shows. While the first episode(s) (within the time span) of One Piece, Boogiepop, Price of Smiles, and Morose Mononokean each ranked below the first episodes of My Roommate is a Cat and Asuka, the next episodes of the former shows were all rated at or above the first episodes of the former (The same could be said about Kaguya's first and Fairy Tail and SAO's second episodes) How exactly each of those chains of preferences end up is not clear with what we can gather just from the averages and the ranks.

Similarly, looking at Run with the Wind and Dororo, we need to also look at The Promised Neverland, which fell beneath both in the third week. If most people consistently rated Neverland above Dororo (and they did until the third episode of the former), and by the third episode rated Run with the Wind above Neverland, it is not impossible that that would lead to Run with the Wind>Neverland>Dororo. Again with the data we have access to as users, I can't say that for certain, but we cannot rule it out either. It's not that the math is wrong, it's just that the different points interact in a complex way that is not always apparent from just looking at isolated points or the data we have available as users.

Also, I don't think looking at the ranking of the first episodes is representative of what the ranking of the second episodes ought to be, especially when the second episode of shows that did have them could rank higher than both their respective first episodes and those of the ones that didn't. That would not have been representative of the increase in rating of the second episodes of One Piece, Boogiepop, Price of Smiles, Morose Mononokean, Fairy Tail, and SAO as compared to their own first episodes and the first episodes of My Roommate is a Cat, Asuka, and Kaguya, respectively.

I can't speak to Shield Hero's placement with the data we have, but that could be affected by a number of the people rating it most highly not watching/rating a number of the other shows, thereby leaving the ranking based on the preferences of those who had watched both it and any or all of the other shows in question (as it should be), who as a whole may not have rated it as highly compared to the other shows, at least during those weeks.
Back to top
View user's profile Send private message
S0crates



Joined: 06 Jul 2018
Posts: 227
Location: Banned - Noticed our poor ethics
PostPosted: Wed Jan 30, 2019 12:28 pm Reply with quote
zrnzle500 wrote:
zip

I think I see the issue. The cumulative rating is impacted by the "filler slot" because those who had an episode that week is rated against the previous episodes all while getting calculated into the cumulative rating (not the shows that didn't run, but the shows that did run).

I would have to see what math they use for the cumulative rating to make sure though.

Also, if Run > Neverland > Dororo, then it's still the case that Dororo > Run consistently every single week. I don't know how they calculate the cumulative ranking here, but it seems Dororo is getting punished for rating bellow Neverland "the week Neverland didn't run".

All in all this method seem poorly suited for weeks where less than half of the shows are airing. I actually think it would be clever to do like what was done with the first episode of Boogiepop, and either push that half's episodes forward one week and have a "mixed" week of the two episodes, or push them back so you got several "double first episodes" (in this case Boongiepop would actually have 3 mixed episodes, but instead you could put them as 1&2 for "first week" and 3&4 the second).

There's many other things that can be done as well, but overall the "filler week" is messing everything up. Next week might correct a few things if current standings stay as consistent as they've been up until now (it's not easy to do statistics at 2(,5) samples though, so it's understandable).
Back to top
View user's profile Send private message
zrnzle500



Joined: 04 Oct 2014
Posts: 3767
PostPosted: Wed Jan 30, 2019 6:03 pm Reply with quote
^No, the rankings of the placeholders don’t affect the cumulative, as it says in the Notes section on the bottom of the page.

Again, you need to look at how all the pieces fit together, not just pointing to two pieces that don’t quite connect by themselves. To order Dororo, Neverland, and Run with the Wind, in addition to looking at how Dororo compares to Run with the Wind, we need to look at how Dororo compares to Neverland and Neverland compares to Run with the Wind. Neverland was above the other two until the third week in the weekly chart. Just going by the weekly rankings, we can’t say for sure whether people preferred Dororo or Neverland over all the episodes so far, but I think you would agree that there is at least a possibility that people preferred Neverland over Dororo, i.e. Neverland > Dororo. Likewise, there is also at least a possibility that people preferred Run with the Wind over Neverland, or Run with the Wind > Neverland. If Run with the Wind > Neverland and Neverland > Dororo are true (as indicated by the cumulative) then Run with the Wind > Neverland > Dororo is at least a possibility too.

Now without knowing the preferences between those two and Neverland and which of those preferences win out, I don’t think we can definitively say that the order isn’t correct. It’s not that the math is wrong, it’s just that we can’t see how all the pieces fit together, at least with the information we have as users.

I will agree that it is a bit early to put too much weight on the cumulative rankings as they stand now, given we only have 2 or 3 episodes (or 4 in Boogiepop’s case) to go by, but that would be true no matter how we calculate things.
Back to top
View user's profile Send private message
S0crates



Joined: 06 Jul 2018
Posts: 227
Location: Banned - Noticed our poor ethics
PostPosted: Thu Jan 31, 2019 10:25 am Reply with quote
zrnzle500 wrote:
^No, the rankings of the placeholders don’t affect the cumulative, as it says in the Notes section on the bottom of the page.

To order Dororo, Neverland, and Run with the Wind, in addition to looking at how Dororo compares to Run with the Wind, we need to look at how Dororo compares to Neverland and Neverland compares to Run with the Wind.


It says it doesn't impact the cumulative ranking of the shows that did not run, but it does impact the rating of the shows that did run. The cumulative rating of Dororo is calculated, but not for Neverland and Run with the Wind. This is probably what makes it swing.

Also if what you say is true then that would also impact the regular rating at as well? Both can't be true at the same time when you only got 2 episodes. Run with the Wind would then rate above Dororo in the regular rating at the same time as there is no reason for why this phenomenon you believe to take place to ONLY impact the cumulative rating.
Back to top
View user's profile Send private message
yuna49



Joined: 27 Aug 2008
Posts: 3804
PostPosted: Thu Jan 31, 2019 10:58 am Reply with quote
Until they start reporting the number of ratings each show has received each week I pay little attention to this whole enterprise. Are we talking samples in the hundreds for each show, or are some ratings based on just a few dozen viewers?
Back to top
View user's profile Send private message
JacobC
ANN Contributor


Joined: 15 Jan 2008
Posts: 3728
Location: SoCal
PostPosted: Thu Jan 31, 2019 2:09 pm Reply with quote
yuna49 wrote:
Until they start reporting the number of ratings each show has received each week I pay little attention to this whole enterprise. Are we talking samples in the hundreds for each show, or are some ratings based on just a few dozen viewers?


The absolute lowest "pretty much no one is watching this but we had enough review slots to cover it" shows get at least 50 votes or so. The more popular shows get hundreds of votes, and it's gone into the thousands sometimes for the really big stuff.
Back to top
View user's profile Send private message Visit poster's website My Anime
Display posts from previous:   
Reply to topic    Anime News Network Forum Index -> Site-related -> Talkback All times are GMT - 5 Hours
Page 1 of 1

 


Powered by phpBB © 2001, 2005 phpBB Group