International Saimoe League 2013

[iamadooddood]

Originally Posted by RegalStar

What is true is that the necklace matches are spices to get the regular season moving, and are not important to the overall picture. However, that's not what SDO is for. SDO is meant to be something that indicates the performance of characters over the period in question (and the period alone), and is always supposed to be the main basis for necklaces. The necklace match, to quote wontaek in an answer he gave to an interview (although the actual interview is translated to Chinese before publicized): "Still people may feel certain character’s SDO to be undeserved for whatever the reason. This is why we must have necklace group round. In a way, we are asking people which character deserves higher necklace score given the relative hardship they overcame."

Wait, remind me again why SDO needs to be taken into account in the final necklace score despite the bolded part?

I'm with Triple_R here; it makes no sense.

[exec]

Originally Posted by Triple_R

So why should some complex period-based calculation scheme matter more than how competitors actually do head-to-head against each other in an exciting 7-way match?

Valuing SDO more than the Necklace match itself really strikes me as the wrong approach to take. Using SDO to determine who gets into the Necklace match is a fine idea, very reflective of pro sports leagues using regular season performance to determine who gets into the playoffs. But then let pure vote count determine who wins the Necklace. In the unlikely even of a vote count tie, use SDO as a tie-breaker.

That's my position.

SDO topic on ISML forums, we had this discussion and my opinion was the same as Triple R, but apparently there are reasons... which don't make sense... they are forcing this SDO so hard it almost hurts...
They talk about making tournament unpredictable and fun, it almost sounds like they are making a show for media.. as long as it unpredictable anything goes... isn't it precisely because a chara is strong it's easy to predict her win. If it was other wise it wouldn't be so easy.

[iamadooddood]

Originally Posted by exec

SDO topic on ISML forums, we had this discussion and my opinion was the same as Triple R, but apparently there are reasons... which don't make sense... they are forcing this SDO so hard it almost hurts...
They talk about making tournament unpredictable and fun, it almost sounds like they are making a show for media.. as long as it unpredictable anything goes... isn't it precisely because a chara is strong it's easy to predict her win. If it was other wise it wouldn't be so easy.

Do we have any actual statisticians able and willing prove whether or not SDO makes things more or less unpredictable? Because it's clear that general consensus is that SDO makes things more predictable and not less, so unless the admins can actually prove otherwise, SDO has to go.

I know I said I decided my major to be mathematics, but I'm seriously considering switching to statistics instead...

[Triple_R]

Originally Posted by exec

SDO topic on ISML forums, we had this discussion and my opinion was the same as Triple R, but apparently there are reasons... which don't make sense... they are forcing this SDO so hard it almost hurts...

Reading through those first three pages on the ISML forums is honestly quite depressing. Why does there have to be a complex Necklace-awarding formula at all? Why can't the Necklace match results (i.e. the actual vote totals) simply speak for themselves?

But to be fair, that thread raises issues with SDO that I myself hadn't considered much. Yes, SDO considerations affect mid and lower-tiered girls as they become glorified tools for people's strategic SDO-based aims. Heck, I see that sort of discussion on ISML threads fairly often.

"It looks like Girl X defeated Girl Y in an attempt to lower the SDO of Girl A"

Direct sniping is one thing, but this sniping of mere proxies really hinders the integrity of ISML, imo.

Originally Posted by iamadooddood

Do we have any actual statisticians able and willing prove whether or not SDO makes things more or less unpredictable? Because it's clear that general consensus is that SDO makes things more predictable and not less, so unless the admins can actually prove otherwise, SDO has to go.

There may have been a time when the unpredictability argument in favor of SDO had some validity to it. But after 2013, I think that's no longer the case. In every Necklace match in 2013, the girl with the highest SDO going into the match came out of it as the Necklace winner. In all 5.

I think this shows how influential SDO has become over the years, and how many people vote with it in mind (I know that I usually do when I vote in Necklace matches).

[KholdStare]

The main reason is the necklace has to do with the period. Without the SDO of the period, we have a random match every 7 matches that has nothing to do with anything. So you have the question wrong.

It's not, "Why do we need SDO in a necklace match?" Because without SDO there wouldn't be a necklace match. In short, without SDO, the necklace match is deemed pointless and has nothing to do with an award based on performance during said period, so necklaces would be abolished.

I'm not sure who made the unpredictability argument but I wouldn't use that to talk about SDO...?

[Triple_R]

Originally Posted by KholdStare

The main reason is the necklace has to do with the period. Without the SDO of the period, we have a random match every 7 matches that has nothing to do with anything.

Going by my idea, SDO would still be used to determine who gets into the match. That alone would mean that the Necklace match is not "random". Of course we shouldn't just put 7 randomly selected girls into the match. Nobody is suggesting any such thing. Sure, use period-based SDO to determine who gets into the necklace match, and then the Necklace match itself reflects period performance.

What I'm basically suggesting is that periods should be thought of as mini-regular seasons, and Necklace matches as one big winner-take-all playoff match. I don't see anything wrong with this idea.

It would be a lot simpler than what you're going with right now, and I honestly think it would make Necklace matches more exciting and interesting on the whole.

As it is right now, SDO severely handicaps many Necklace matches. This is horrible because it unjustly punishes certain girls (Mikoto Misaka throughout 2013), and it also takes a lot of the natural fun away from what could otherwise be a wild 7-way match.

Increasingly, the girls with the three to four weakest SDO scores going into a Necklace match are getting passed over for girls with the three strongest SDOs going into a match. This is because people know the weaker SDO girls have no shot due to a severe SDO disadvantage going into the match itself. Without SDO, people would simply vote their heart in most Necklace matches. Right now, I admit I don't do that. I try to determine which two or three girls have the highest SDO scores going into the match, and I vote for my favorite girl out of those two or three girls. And I'm confident that many voters do the same thing.

It's good to have Necklace matches and awards to spice up the regular season. But it would be better if those Necklace match results reflected genuine strength, rather than encouraging loads and loads of strategic-based voting (both during the period and in the Necklace match itself) instead of pure heart-based voting.

[RegalStar]

Originally Posted by Triple_R

...it would be better if those Necklace match results reflected genuine strength...

For necklaces? It wouldn't be.

[KholdStare]

Correct, if necklaces reflected genuine strength (ie. each necklace match is basically a periodic performance rating), then you are once again off from how the staff envisioned necklaces to be.

You said SDO severely handicaps the necklace matches. I don't understand this statement. Whoever has the highest SDO should be awarded the necklace first and foremost. The necklace match only gives the ones with lower SDO a chance at stealing it.

[Triple_R]

Originally Posted by RegalStar

For necklaces? It wouldn't be.

Why not?

Originally Posted by KholdStare

Correct, if necklaces reflected genuine strength (ie. each necklace match is basically a periodic performance rating), then you are once again off from how the staff envisioned necklaces to be.

Maybe given the amount of negative feedback you're getting on SDO (both here and on that thread that exec linked to) the ISML staff should change the way it envisions necklaces to be?

You said SDO severely handicaps the necklace matches. I don't understand this statement. Whoever has the highest SDO should be awarded the necklace first and foremost.

Why? Should a pro sports team with the best regular season record also win the league's championship for that year? There's loads of cases where this is not the case, and it doesn't bother pro sports fans at all. No pro sports fan thinks that the team that happened to have the best regular season record should automatically win the league championship for the year, or receive a major handicap advantage on the competition. By the same token, I don't see why the girl who has the best period record should receive heavy favoritism in the Necklace match.

The necklace match only gives the ones with lower SDO a chance at stealing it.

I think that's a horrible way to look at Necklace matches. It makes it much less fun and interesting than a true 7-way contest where everybody is on the same playing field, without any handicaps.

The necklace match itself is much more fun than some complex period-based calculation is.

[exec]

Originally Posted by KholdStare

Correct, if necklaces reflected genuine strength (ie. each necklace match is basically a periodic performance rating), then you are once again off from how the staff envisioned necklaces to be.

You said SDO severely handicaps the necklace matches. I don't understand this statement. Whoever has the highest SDO should be awarded the necklace first and foremost. The necklace match only gives the ones with lower SDO a chance at stealing it.

Spoiler for Aquamarine:

Spoiler for Amethist:

Spoiler for Ruby:

Spoiler for Emerald:

Spoiler for Topaz:

You were saying? please do elaborate on this "chance" thing you are so forcing here? if you look at results you'll notice that not much changes on bottom of the SDO Top after voting. Who do you think would vote for a low SDO character? Would you do it? Or are you saying that none of those charas are strong enough to get this chance? What kind of conditions should be met to get this chance? 50% voting? are you even sane to suggest such things to be considered possible?
I'm sorry Kholdy, i do respect you as a person and statistician, but something tells me you don't include psychological factor into your statistics. No one would vote for low tier chara and if it's low on SDO it only makes it less possible to get votes. It's like people like to make labels, if chara is at the bottom, a regular voter will just ignore her and vote for possibilities, Mikoto was an exception, but did she actually manage to get anything? clearly there is no "chance" in Necklaces.

[Ithekro]

I don't vote possibilities. I don't even look at then until after the voting is done. I vote for who I want.

[Triple_R]

7-Way matches should always be fun.

Even in cases where the winner of the match is predictable, there can still be surprises in where everybody falls from 1st to 7th.

But SDO takes that fun element away. Unless one of them is named "Kanade", the girls ranked 5th through 7th in SDO scores might as well not show up at all. They will almost always fall near/at the bottom once results are tallied.

[wontaek]

It seems that many people are confusing what is cause and what is effect. Take a look at how things were from 2008 onward.

2008 : Necklace was given by cumulative VD. Very small group from Vietnam who controlled only about 10% of the voters successfully manipulated VD to give Topaz necklace to Sakurazaki Setsuna from Negima series. The real important thing here is mere 10% of the voters that sustained effort over all the necklace rounds were sufficient to give necklace to someone who eventually finished 40th out of 64.

Let me warn you all : If a group that controls mere 5% of the overall votes sustain effort for the entire regular season, while that group probably can't manipulate who wins the first 2 necklace, that group will be able to dictate which Tier 1 character wins the last 3 necklace and may be able to get a Tier 2 character of their choice win the Topaz necklace, if we enforce 1 necklace per year rule as well. Cumulative VD is much more easier to manipulate compared to SDO since your effort will count towards the final necklace score, regardless of whether your character won or not.

2009 : SDO is introduced. Rozen Maiden Faction appears in mid season and steals 3 necklace away with masterful manipulation of SDO. The highest Ranked Rozen Maiden character at the end of post season was 12th. They were able to accomplish this because at their highest point, they controlled up to 25% of the voters. They hastily beat a retreat and disappeared as ISML kept expanding in voter numbers, which meant their share of votes diminished. They gave clear demonstration that if you got good enough support for popular enough character, you can make things very unpredictable. Those arguing that SDO will make things less predictable should research what transpired in 2009 carefully.

2010 : Number of rounds got reduced. Necklace rules were identical to 2009. Kugimiya Rie Faction manage to capture 4 of 7 necklace. Aisaka Taiga and Nagato Yuki got a necklace even though they ended up finishing at 6th and 7th . Even though Shana captured 3 necklaces, she ended up finishing 3rd. Kugimiya Faction controlled close to 20% of the voters at their height. This did help them greatly in winning the 3 necklace, YET, the necklace they desired the most, Ruby, eluded them.

Do look at the magnitude of proportion a faction needed to control to manipulate the necklace here. In 2008, mere 10% was good enough to manipulate the necklace. In 2009 and 2010, Factions needed to bring in 20+% and even that wasn't enough for many of the necklaces. Rozen won only 3 out of 7 necklaces. If we went by VD, as many of you are supporting, Rozen and Kugimiya faction could have controlled something like 6 out of 7 necklaces.

2011 : Necklace Group round is added in. There are many reason for the necklace group round, but there is another reason that wasn't discussed. From 2008 to 2010, Groups controlling about a quarter of the voters were able to enforce their desire upon the 75% of the voters more than half of the times. By the end of 2010, however, the largest faction controlled only about 15% of the voters reliably, and this percentage was shrinking. By adding necklace group round, we have given the 85+% of the voters chance to undo whatever SDO manipulation the 15% succeeded in. We also instituted 1 necklace per year rule. This had made the later necklace period be very unpredictable, resulting in Eucliwood, someone who would finish at 12th to win the last necklace of the year. It should be noted that Eucliwood entered the necklace group round with record of 5 wins and 2 losses.

There is something very important to learn from Eucliwood case. The 2011 necklace system, which is almost the same as 2012 and 2013, did accomplish giving us unpredictable result by the end. The reason why 5 - 2 record girl can no longer win a necklace has more to do with current 2 division system instead of SDO.

2012 : Partial scores for the losers is given out to lessen the SDO gap between characters with same number of wins. Number of Necklace got reduced to 5, which may be more responsible for many of the points raised before compared to most of the things that has been discussed. Before difference between win and loss in terms of points was 3 , 3 or 0. Now the difference is more like 2.2, as losers often got at least 0.8 points even in defeat. The lowest finisher to win a necklace was Misaka Mikoto, who ended up at 6th.

There is 2 big things people need to be aware of from 2012. If we criticize 2012 necklace winners for being predictable, then is there any system that can reliably hand out the 5 necklace to someone far below top 5? If you want a system where a character will earn a necklace through good performance, I doubt you can make one that will spread the necklace out to someone at 12th or below.

Even bigger thing is that the class of 2011, the characters that first appeared in ISML on 2011, are big anomalies . This group has been dominating ISML ever since 2011. These characters are so popular, and I think you can find plenty of commercial sales values to back up this claim of popularity, that it is hard to deny them the necklace if they want it. I tried many simulations, and basically came to conclusion that we have to either get 10,000+ new voters that are not a fan of Class of 2011 characters ( note not a fan does not mean they are haters. In simulation term, these are voters that will support class of 2011 characters less than (50 + 1 ~ 3) % of the times ) or simply ban them in order to prevent them from winning 2 of the first 3 necklaces.

2013 : Everyone said much about 2013 so I won't elaborate on those things. Still, there is one thing people are NOT paying enough attention. Misaka Mikoto never looked like a Tier 1 material during prelim and seeding matches. She finished 3rd in her group in the last seeding match. There wasn't a way to justify her to be in Tier 1 with those numbers, thus she was assigned to Tier 2, which have systematic handicap when it comes to earning SDO, if tier placements were mostly correct. That Seeding group result has as much to do with any SDO manipulations as to the reason why Mikoto did not win a necklace. Morale of the story is you can never decide to skip a match in ISML if you want to do well. Hanazawa Kana fans have always been model of consistency when it comes to voting. I do not wish to ruin a system that rewards consistent voters in order to bring in a system that rewards a character that only gets votes in one or two matches days.


It now comes to the biggest reason why we can't let Necklace Group match result to be the only thing that matters. Look at the history. ISML always has been a league that rewards diligent voters. If we increase the role of the necklace group match in deciding who wins the necklace, we would be rewarding those opportunistic voters who only drops in for 1 match days out of 8 days in a necklace period. So far most of the suggestion I read are very vulnerable to influence of these opportunistic lazy voters. One of key benefit of SDO is that you can't hope to accomplish all you want by concentrating on merely one of the 8 days, and perhaps not even 2 of the 8 days. Please consider this as you ponder upon various influence of SDO. Also remember that necklace group round eligibility is unaffected by SDO, unless in case of extremely unlikely tie.

[KholdStare]

So far, there have been "upsets" in the necklace matches where the highest SDO doesn't win. While this is rare, it happens. If people have a problem with necklace matches however, maybe we should go back to having SDO as the only decider.

[Triple_R]

Originally Posted by wontaek

There is something very important to learn from Eucliwood case. The 2011 necklace system, which is almost the same as 2012 and 2013, did accomplish giving us unpredictable result by the end. The reason why 5 - 2 record girl can no longer win a necklace has more to do with current 2 division system instead of SDO.

Yes, and is that 2 division system going away anytime soon?

So I don't see the point of discussing an unpredictable 2011 result that would not happen today due to changes made since that time.

In simulation term, these are voters that will support class of 2011 characters less than (50 + 1 ~ 3) % of the times ) or simply ban them in order to prevent them from winning 2 of the first 3 necklaces.

The first 2 or 3 Necklace matches are going to be predictable, no matter what, yes. At the very least, this will be true until "the Class of 2011" finally starts to weaken relative to everybody else.

But SDO makes even the final two Necklace matches predictable now. Without SDO being part of the results calculation for those matches, they would have almost certainly been less predictable.

SDO currently runs contrary to pure strength and makes Necklace matches more predictable. It's the worst of both worlds that Eater of All and I were debating back in 2011.

2013 : Everyone said much about 2013 so I won't elaborate on those things. Still, there is one thing people are NOT paying enough attention. Misaka Mikoto never looked like a Tier 1 material during prelim and seeding matches. She finished 3rd in her group in the last seeding match. There wasn't a way to justify her to be in Tier 1 with those numbers, thus she was assigned to Tier 2, which have systematic handicap when it comes to earning SDO, if tier placements were mostly correct. That Seeding group result has as much to do with any SDO manipulations as to the reason why Mikoto did not win a necklace. Morale of the story is you can never decide to skip a match in ISML if you want to do well.

Real Moral of the Story: SDO makes it harder for mid-to-late season changes to character strength to have significant impacts. In other words, it retards natural character boosting, such as what Mikoto likely enjoyed due to mid-to-late season Railgun S episodes. It does this by attaching season-encompassing weight to prelim results, putting further constraints on character strength fluctuations. Yet again, SDO actually makes things more predictable for people who are paying attention. SDO systematically limits one of the few things that can really spice up the ISML regular season - A character being boosted by there being a currently airing anime. Wow, SDO is even more insidious than I originally thought it was.

Hanazawa Kana fans have always been model of consistency when it comes to voting. I do not wish to ruin a system that rewards consistent voters in order to bring in a system that rewards a character that only gets votes in one or two matches days.

Consistent voters would be rewarded under my suggestion as well, as using SDO to determine who gets into the Necklace match would make it important to ensure that your favorite girl(s) have a good record during the period. This is particularly true given the two division setup.

Right now, you actually punish mid-season newcomers to ISML by having their impact constrained by prelims results earlier on in the season. A new anime show brings in new fans, and new fans of a specific girl. Those fans decide to start supporting her by voting in ISML, only to find that ISML's rules have already severely disadvantaged the girl. ISML rules are essentially unwelcoming to these new fans that join mid-season.

The ISML system currently favors long-standing fans over new fans who wish to join anytime after prelims. No wonder 2011 girls continue to dominate, as the impact of new fans is reduced by the ISML system itself.

It now comes to the biggest reason why we can't let Necklace Group match result to be the only thing that matters. Look at the history. ISML always has been a league that rewards diligent voters. If we increase the role of the necklace group match in deciding who wins the necklace, we would be rewarding those opportunistic voters who only drops in for 1 match days out of 8 days in a necklace period.

No, that's not accurate. Because simply getting into the Necklace match requires a good period record. Or at the very least, it would if we used SDO to determine who gets into the Necklace match.

So far most of the suggestion I read are very vulnerable to influence of these opportunistic lazy voters.

How do you know that they're opportunistic and lazy? What if they're simply new voters? Everybody is new to moe competitions at some point.

Originally Posted by KholdStare

So far, there have been "upsets" in the necklace matches where the highest SDO doesn't win.

Not in 2013 there hasn't been. I think that voters are getting better and better at using SDO to game the system. It's really sapping a lot of the fun and heart out of ISML. It's a shame to see.

[RegalStar]

On the flip side, your system also means that showing up for prelims and regular seasons are much less meaningful, as all it really takes are the few matches that count (necklace matches, post-season matches) to decide the search direction of the graphic staff. Also, 2012 Eucliwood.

[Triple_R]

Originally Posted by RegalStar

On the flip side, your system also means that showing up for prelims and regular seasons are much less meaningful,

No, my suggested system does not mean that.

Under my suggested system, prelims would still be very meaningful simply for determining who even gets into the regular season.

Regular season matches are still very meaningful for determining who even gets into the Necklace matches, as well as determining who gets into the Post-season to have a shot at Circlets and the Tiara.

However, there is such a thing as making certain things too important, and giving them too much weight. Should a girl be heavily handicapped against winning a Necklace for an entire calender year just because of freaking Prelims? Do you really, honestly think that's fair, or good for ISML? It's certainly not fair to people who become interested in ISML mid-season, and not being fair to potential new voters is certainly not good for ISML.

So what I've suggested, and what other people have suggested, is superior to what is currently in place, for several reasons that have now been explained in detail, and for which I've read no good counterarguments for.

Also, 2012 Eucliwood.

What about 2012 Eucliwood?

[RegalStar]

I said "much less meaningful", not "meaningless". Regular season matches are supposed to be the backbone of the tournament.

It's not a heavy handicap. With theoretically perfect seeding, it results an average of ~2.2 SDO lost, which translates to an edge of 50-100 votes in necklace matches, which is pretty light all things considered. And yes, not doing well in an earlier part should translate to a slightly higher hill to climb over later. It makes perfect sense to me, given the structure of this tournament.

(The main three reasons for Mikoto's SDO this year are bad luck with scheduling, higher deviation of strength between prelims and regular season in Nova than in Stella, and a certain group - or perhaps several groups - attacking her SDO. There will be some countermeasures about the last part which will be implemented next year. Mikoto being tier 2 does hurt her SDO a little, but not much.)

Eucliwood was perfectly able to take advantage of her new season boost in 2012.

[Triple_R]

Originally Posted by RegalStar

I said "much less meaningful", not "meaningless".

"Much less meaningful" is still making it sound worse than what it is.

Regular season matches are supposed to be the backbone of the tournament.

Under what I'm suggesting/supporting, they still would be. A girl would still have to have an excellent regular season in order to make the Circlet/Tiara post-season, and a girl would still have to have an excellent period in order to get into the Necklace match.

It's not a heavy handicap.

Yes, it is. There's a reason why all five Necklace matches in 2013 was won by the girl with the highest SDO going into the match. This in spite of truly valiant performances by Mikoto. You are greatly downplaying this.

With theoretically perfect seeding,

If you need something theoretically perfect to support your argument, then your argument probably isn't very good, because it doesn't account for all sorts of imperfect variables that will typically be in play.

And yes, not doing well in an earlier part should translate to a slightly higher hill to climb over later.

Why? In most real world competitions, prelims is almost always just about determining who advances to the next phase. Most fans of pro sports, and other competitions, would find it downright offensive to have prelims do any more than that (aside from something relatively minor like "home field advantage"). And that's because they realize that handicaps run counter to the spirit of fair and honest competition, where people naturally want to see the best competitor win; not the one who simply has the most handicaps favoring him or her.

What you favor is not fair to voters who happen to join ISML mid-season.

It also makes it harder for character boosting to be effective, which is one of the few things that can spice up the ISML regular season.

So I strongly and vehemently disagree with you on this.

It makes perfect sense to me, given the structure of this tournament.

It makes no sense at all to me. I think its horrible and disgusting. I really do.

The main three reasons for Mikoto's SDO this year are bad luck with scheduling, higher deviation of strength between prelims and regular season in Nova than in Stella, and a certain group - or perhaps several groups - attacking her SDO. There will be some countermeasures about the last part which will be implemented next year.

Yes, instead of making things nice, clean, and simple, let's complicate things even further. Let's make ISML so complicated that only a master statistician could ever hope to understand all of the thinking and formulas that go into it. I'm sure that'll make ISML very inviting to newcomers (along with being unfair to mid-season newcomers, of course).

People like fair, and open, and transparent competitions (the perceived lack thereof hindering J-SaiMoe as we speak). The more things are layered down with complex rules and formulas and "countermeasures", the more it's a turnoff to newcomers who want to take part in something that's not hard to understand and grasp.

And that's yet another reason why it would be best to just let the girl with the most votes in the Necklace match win the Necklace. It's clear, fair, simple, and straightforward.

Eucliwood was perfectly able to take advantage of her new season boost in 2012.

Did she face the same prelim disadvantage that Mikoto did this year? (Honest question, as I don't know)

In any event, Eucliwood only did this because she never had people trying to game the SDO system against her to the degree that Mikoto did. When she won the 2012 Emerald Necklace, it was with her going into the match with the highest SDO of all the competitors.

[RegalStar]

Real-life leagues also have it that if a team performs horribly in the beginning, then that'd drag them down even if they suddenly spiked up in the mid-season. Real-life competitions also have it that if they perform bad in the qualifiers, they're going to get a worse seeding and thus be overall at a disadvantage in the playoffs.

And also, once again necklaces aren't meant to be awarded to the strongest characters.

If you need something theoretically perfect to support your argument, then your argument probably isn't very good, because it doesn't account for all sorts of imperfect variables that will typically be in play.

This isn't related but I'm picking this line out in particular due to how just... stupid it is. In any system whatsoever that relies on inputs with any amount of unpredictability behind it, the only way to go is to assume that the expected input is perfect. It's common sense.