2014 Amethyst General Discussion Thread

[Hajime Tanegashima]

SDO might be one of the reasons Kurumi lost that match against Kotori. I guess many DaL fans did like kukimunstir: vote the best SDO candidate hoping to get a win for the series.

I've already said I think Kotori and Yoshino don't have what it takes to be a serious threat to the stronger Stellas because they share too many voters; they are just these voters' second or third choice and, at best, a weapon against other strong Novas. I think that happens a lot less in the case of Kurumi. Perhaps I'm wrong but that's what I think, and Aquamarine NM results seem to go for what I said. Now, we will see how Kurumi fares against two characters of the same series and with the highest SDO. This is a true test for her.

It would be ironic that Kurumi had fallen in Aquamarine period not against Kotori but against the randomness of loops and SDO building. This NM might tell us many things about what really happened back then.

[avery-kun]

SpoilerShow

Easy. DAL fans need to realize that this may be their only shot at DAL ever having the necklace, so all DAL fans need to vote for Kurumi without question. Otherwise, the DAL split will always be there in every necklace match while stella gets all the necklaces.

Color wise and SDO-wise the choice should be obvious for the fans indeed. However, was there actually even one time in ISML history since the necklace match creation when a contender won the necklace despite a 2nd (not even talking about three) character from the same series being in the necklace?

Edit :
Aquamarine 2011 : Kanade+Yuri -> no win.
Amethyst 2011 : Mio+Azusa ->no win
Emerald 2011 : Mio+Azusa ->no win
Ruby 2011 : Mio+Azusa ->no win
Diamond 2011 : Mio+Azusa+Yui ->no win
Aquamarine 2012 : Kanade + Yuri : Kanade win
Ruby 2012 : Kobato+Sena : no win
Emerald 2012 : Mio+Azusa ->no win
Topaz 2012 : Kobato + Sena ->no win
Aquamarine 2013 Ruri+Kirino ->no win
Aquamarine 2014 : Yoshino + Kotori -> no win
Amethyst 2014 : Yoshino+ Kurumi + Kotori -> ??

From this, seems pretty clear that as soon as the two contestants from the same series are even a little close in strength, they never won, although they were close several times. We can even say that K-ON suffered greatly from this, as in 2012 they only managed to win a necklace on topaz, when Azusa didn't qualify. Can DAL really manages a win against Kanade with a three way split between 3 characters of around the same strength? Doesn't sound really doable to me.
BTW, 2013 was almost free of split voting during necklace matches, with only aquamarine where it happened. But it looks like we're going to get a 2014 year with a LOT of split voting...
This is the second time in ISML history where 3 characters from the same series access to the same necklace. And the first time all of them got 7-0.

Great observation. This proves how screwed DAL are. Its funny that the biggest threat to the strongest girls in Nova are themselves.

As for HanaKana. I'm inclined to believe that it doesn't exist, or isn't very strong. Its just that Kuroneko, Kanade, and Eu have a very high overlap in voters, and most of them are not HanaKana.

[LOveLive!]

SDO might be one of the reasons Kurumi lost that match against Kotori. I guess many DaL fans did like kukimunstir: vote the best SDO candidate hoping to get a win for the series.

I've already said I think Kotori and Yoshino don't have what it takes to be a serious threat to the stronger Stellas because they share too many voters; they are just these voters' second or third choice and, at best, a weapon against other strong Novas. I think that happens a lot less in the case of Kurumi. Perhaps I'm wrong but that's what I think, and Aquamarine NM results seem to go for what I said. Now, we will see how Kurumi fares against two characters of the same series and with the highest SDO. This is a true test for her.

It would be ironic that Kurumi had fallen in Aquamarine period not against Kotori but against the randomness of loops and SDO building. This NM might tell us many things about what really happened back then.

Contestants who have the more loyal voters does not always win. They are just strong on free-for-all, not on 1-on-1. See Shana, she has the third most loyal fans(according to 2013 Bouns Finale) but she is still suffering now. Most winners were most voters' second or third favorites in ISML history.

[Crisu]

Though it's no hope for Goko Ruri's third Amethyst Necklace, I still want to give my thanks to every staff.
With all your love, we will keep fighting till the end.

W-what -- someone did see! N-no! Forget what you saw!
The website just .. didn't get the actual lineup yet!! So it was simply using a placeholder containing values of last year's winner.

I'll have to ... have to go speak with someone about that. Yes, so I will take my leave now and fix this, hmph. You should thank me!

[Hajime Tanegashima]

You are right, but you need a fanbase too. Kotori's and Yoshino's results in a match in which the strong Stellas were, they showed that in a match against them they would lose too many votes. You need to be strong both as a first choice and as a second/third. Kotori and Yoshino don't fulfill the first condition. I don't know about Kurumi, but I think she should at least do better than the other two in that regard.

[Just]

W-what -- someone did see! N-no! Forget what you saw!
The website just .. didn't get the actual lineup yet!! So it was simply using a placeholder containing values of last year's winner.

I'll have to ... have to go speak with someone about that. Yes, so I will take my leave now and fix this, hmph. You should thank me!

This tsundere Crisu is so moe that I want the Seitokai to punish him more, for his mistakes concerning that website error.

[Momento10]

It is a shame that the only girl for the Amethyst Brooch who looks good in purple is Shouko, with her long purple hair and her long, dark hair onee-sama like personality. I guess Masami fits this color scheme as well. Everyone else seems better off in another color, particularly Topaz. Maybe Rin also fits Ruby for her fiery personality.

[Shmion84]

VF%-add keeps hanging around. Care to explain VF%-add model to everyone again? While I think Method-F is definitely superior to SWVO or VP only models, I can't say it has establish superiority over VF%-add, yet.

VF%-add is based on Votes For (VF) and Votes Against (VA) for a certain character.
We set the strength of character A as s(A) = VF / (VF + VA)
Prediction VF%-add for character A against character B: %(A) = s(A) - s(B) + 50 %

Also, it is interesting that SWVO-multiplicative has significantly improved in relative prediction power. It can be one match fluke, but it is something worth noting.

SWVO-mult - relative to Method F & absolute
Amethyst 1: 2.66 - 849.47
Amethyst 2: 1.69 - 564.65
Amethyst 3: 3.74 - 1466.11
Amethyst 4: 8.57 - 1308.50
Amethyst 5: 3.75 - 1201.62
Amethyst 6: 4.49 - 784.50
Amethyst 7: 1.57 - 436.16

So I don't think SWVO-mult has significantly improved.

[maglor]

VF%-add keeps hanging around. Care to explain VF%-add model to everyone again? While I think Method-F is definitely superior to SWVO or VP only models, I can't say it has establish superiority over VF%-add, yet.

VF%-add is based on Votes For (VF) and Votes Against (VA) for a certain character.
We set the strength of character A as s(A) = VF / (VF + VA)
Prediction VF%-add for character A against character B: %(A) = s(A) - s(B) + 50 %

Also, it is interesting that SWVO-multiplicative has significantly improved in relative prediction power. It can be one match fluke, but it is something worth noting.

SWVO-mult - relative to Method F & absolute
Amethyst 1: 2.66 - 849.47
Amethyst 2: 1.69 - 564.65
Amethyst 3: 3.74 - 1466.11
Amethyst 4: 8.57 - 1308.50
Amethyst 5: 3.75 - 1201.62
Amethyst 6: 4.49 - 784.50
Amethyst 7: 1.57 - 436.16

So I don't think SWVO-mult has significantly improved.

Thanks, and you are right about SWVO-multi not having improved. That 8.57 says we should stay away from it for a while. Perhaps we should revisit it in Topaz period. The reason for mentioning Topaz is because some method might have something in it which makes it better or worse as time goes by, so it is worth testing everything there. Right now, your method F and VF%-additive model are the two clear winners, to a degree that I don't think we need to test and validate in in Ruby, but if you want to, you are always welcome to do so.

[Just]

So what about SDO Manipulation Analysis? I guess we'll still do that in Ruby right?

[maglor]

So what about SDO Manipulation Analysis? I guess we'll still do that in Ruby right?

You did magnificent job. I think it is needed, all the time, in order to detect anomalies. Knowing what SDO-manipulations might be present helps us better detect multivotes. I can tell you that your work helped us find some multivotes that tried to help with SDO ranking of certain characters. I would appreciate it if you continue this in Ruby as well.

As for SDO manipulation during Amethyst, I think some groups had some success in some of the rounds, but not in all the rounds. At worst, I suspect an SDO value to have shifted by about 6 for a character due to manipulation attempt, probably much smaller. In contrast, I think loops may be responsible for SDO swing of 9 or more for some character. This is why I think, for Amethyst, loops had more impact compared to faction SDO manipulation attempt. Still, presence of these loops does not diminish value of your work.

[Just]

SpoilerShow

So what about SDO Manipulation Analysis? I guess we'll still do that in Ruby right?

You did magnificent job. I think it is needed, all the time, in order to detect anomalies. Knowing what SDO-manipulations might be present helps us better detect multivotes. I can tell you that your work helped us find some multivotes that tried to help with SDO ranking of certain characters. I would appreciate it if you continue this in Ruby as well.

As for SDO manipulation during Amethyst, I think some groups had some success in some of the rounds, but not in all the rounds. At worst, I suspect an SDO value to have shifted by about 6 for a character due to manipulation attempt, probably much smaller. In contrast, I think loops may be responsible for SDO swing of 9 or more for some character. This is why I think, for Amethyst, loops had more impact compared to faction SDO manipulation attempt. Still, presence of these loops does not diminish value of your work.

Sure~ Thanks Mr moeglor for the compliment *blushes* I'll continue to do it for Ruby, though for Emerald I cannot guarantee being able to help. My work schedule for July is still pending. I'll let you know as soon as it is out.



For SDO shift as a result of manipulation attempt, I've done SDO predictions for each round starting Amethyst 5, and see how much the resultant SDO varies from the prediction. (I used figures from my model only, instead of average from all predictions obtained here). Only actual upsets for close matches would result in such change, and those matches were close to begin with.

I detected such shifts in Amethyst 5 and Amethyst 7, and both times the degree in SDO shift was about 2 (as the result of the upset of one underling) for each character. It's hard to do such analysis early in the Period, as the underlings are still not known; but I'd say Amethyst 2 also has some suspicious upsets. Summing all up, I'd say the worst case estimate would be SDO being affected by 4.



I've back-projected the average strength curve for all opponents of each 1st-tier character in the Amethyst period (the mindset is that early in the Period, the captain-underling relationship is still not known for each character, so the power estimate during the time shall be independent of this underling identity), and it shows that the strength of their schedules are roughly in line with their final SDO. For example, among Novas, Kurumi really did face the toughest schedule in Amethyst, so she having the highest SDO is reasonable.

This being said, there is some abnormality during Amethyst 5 & 7 (again), with Kanade's underlings suddenly gaining strength both times. This might be the result of deliberate manipulation efforts, but of course model error is also possible.

If you're interested, I can show you the curves I obtained; but after the Necklace Match. (I also plan to publish my findings on Baidu, after all the comedy of the Necklace Match and everyone has calmed down a bit. I won't want people to think I'm publishing things at this sensitive time with some evil intention )



(Oh, once again, I haven't really investigated the relationship of SDO with loops LOL Maybe you'd like to raise some ideas on how to analyze such relationship?)

[tehyc]

I'm actually wondering for a while now. How is the point of a loser is given?

[smartboyhw]

I'm actually wondering for a while now. How is the point of a loser is given?

The point of a loser in each match is given by votes for / votes against of the match. Please refer to International Saimoe League 2014 Charter Article XI (Regular Season) Section E 1.(c).

[10ZHAbin]

How things are looking in Baidu right now.Show

Tenshi: supporting Tenshi with confidence.
Kuroneko: Supporting Tenshi with very little complaints.
Tokisaki Kurumi: Supporting Kurumi, but has talks and discussion split problem.
Itsuka Kotori: the veteran of ISML are encouraging the fans to support Kurumi to minimized split problem but many are reluctant and choose to support Kotori. Explanations are done during questions but a number of fans continue to support their own idea of voting only for Kotori regardless of the situation.
Yoshino: mostly the fans are following the words of veterans and are willing to give up Yoshino in favor for Kurumi. Some minors have conflicts but are all persuaded peacefully.
Tsutsukakushi Tsukiko: surprisingly, there is no talk of Amethyst necklace despite being the only character without split voting problem.
Eucliwood: I can't find their residential area.

[maglor]

How things are looking in Baidu right now.Show

Eucliwood: I can't find their residential area.

They still remain as elusive as ever.

[Vella]

How things are looking in Baidu right now.Show

Tenshi: supporting Tenshi with confidence.
Kuroneko: Supporting Tenshi with very little complaints.
Tokisaki Kurumi: Supporting Kurumi, but has talks and discussion split problem.
Itsuka Kotori: the veteran of ISML are encouraging the fans to support Kurumi to minimized split problem but many are reluctant and choose to support Kotori. Explanations are done during questions but a number of fans continue to support their own idea of voting only for Kotori regardless of the situation.
Yoshino: mostly the fans are following the words of veterans and are willing to give up Yoshino in favor for Kurumi. Some minors have conflicts but are all persuaded peacefully.
Tsutsukakushi Tsukiko: surprisingly, there is no talk of Amethyst necklace despite being the only character without split voting problem.
Eucliwood: I can't find their residential area.

In baidu opinions, this match is DAL vs tmall, and Tsukiko is just the referee.

[Progeusz]

What is tmall? Or rather, how was this word created?

[smartboyhw]

How things are looking in Baidu right now.Show

Tenshi: supporting Tenshi with confidence.
Kuroneko: Supporting Tenshi with very little complaints.
Tokisaki Kurumi: Supporting Kurumi, but has talks and discussion split problem.
Itsuka Kotori: the veteran of ISML are encouraging the fans to support Kurumi to minimized split problem but many are reluctant and choose to support Kotori. Explanations are done during questions but a number of fans continue to support their own idea of voting only for Kotori regardless of the situation.
Yoshino: mostly the fans are following the words of veterans and are willing to give up Yoshino in favor for Kurumi. Some minors have conflicts but are all persuaded peacefully.
Tsutsukakushi Tsukiko: surprisingly, there is no talk of Amethyst necklace despite being the only character without split voting problem.
Eucliwood: I can't find their residential area.

In baidu opinions, this match is DAL vs tmall, and Tsukiko is just the referee.

I went into unstoppable laughter when I saw this. (Especially when tmall does not only mean Kanade + Ruri in the Chinese online world)

SpoilerShow

Tmall is an online shopping platform. So online dating sim vs. online shopping platform?

Surprised that Yoshino fans would turn their support towards Kurumi. Nevertheless, the Kotori fans refuse to budge. Vote splitting still occurs, but at least not that too destructive.

What is tmall? Or rather, how was this word created?

SpoilerShow

Tmall means 天貓，天 denotes 天使, which means Angel (Kanade), 貓 denotes 黑貓, which means Kuroneko (Ruri)

[Just]

What is tmall? Or rather, how was this word created?

Tmall is an online shopping platform, and the Chinese name is 'Ten-Neko'.

I'd rather this concept not being introduced here. HanaKana faction is a good enough term that everyone understands.