2013-11-04, 01:00 | Link #1781 | ||||||
NOM.
Join Date: May 2013
Location: Singapore
Age: 31
|
Quote:
If you truly want to spice up the matches, the only way is by a fair showdown, and nothing more. The way SDO does it now can hardly be called fair. Quote:
Due to the complexity of SDO, manipulating SDO has had a lot more repercussions than manipulating VD alone. The effect of manipulating SDO by a certain successful fanbase ends up being greater than manipulating things like VD by the same fanbase. In the end, the manipulation of SDO has been proven to be at least as damaging, if not more so, than that of VD. Quote:
How about setting criteria as to who can vote in necklace matches? You can only vote for characters you have voted at least once for in the period. Or you must have cast a total of X votes in that period or something. Or both. Low vote count (where people don't vote for her) is not always a bad thing. It shows that the girl in question is weak, and that just by being in the running for ISML alone, she's punching above her weight. Nothing wrong about that. And if she does pull off an upset? Nothing wrong with people liking the underdog, right? "An underdog that needs protection = moe" is legit. Also, there are also those who do not ignore the regular season matches. It's just that, they manipulate the matches. If you encourage people to not ignore regular season matches, these guys will be sure to show up, sometimes regardless of the encouragement or lack thereof, but more likely to spite SDO. Which one will you prefer? (In short, any rule made to encourage people to vote regularly causes the tournament to get more susceptible to manipulation. Before I go on, I'd like to verify this with you.) Quote:
Also, seeing that you have a B.S. in Mathematics (though not in pure Statistics, it'll have to do; my university has separate degrees for Pure Mathematics, Applied Mathematics and Statistics lol), what is the median SDO handicap due to seeding? Quote:
Quote:
Speaking of which, I'd prefer having an even number instead of the current 7 contestants for a necklace match.
__________________
|
||||||
2013-11-04, 02:14 | Link #1783 | |
Senior Member
Author
|
Quote:
For Aquamarine, there would be an 1 on 1 Necklace match between the Nova girl with the highest SDO for the Aquamarine period, and the Stella girl with the highest SDO for the Aquamarine period. The match itself would be determined strictly by vote totals. The winner of the Aquamarine Necklace is then prevented from taking part in future Necklace matches for the same Calender year. For future periods, there would be 1 on 1 Necklace match between the Nova girl with the highest SDO for the period who hasn't won a Necklace yet, and the Stella girl with the highest SDO for the period who hasn't won a Necklace yet. Hhhmmm... Well, if you can't even get into the Necklace match without having the very best period performance of every girl in your division (who hasn't won a Necklace yet), then surely that alone is ensuring that the Necklace greatly honors/reflects period performance. It would also motivate people to go totally all-out for their favorite girls in each period, so long as she has a chance at getting the highest SDO for that period. If anything, it would make Regular Season matches even more important than what they are right now. The Necklace match itself would be between two competitors in two entirely separate divisions. So that likely gives the match at least a touch of unpredictability. It would also give a little hint, but not necessarily a big hint, to how Nova compares to Stella at regular checkpoints throughout the year. I would think that would appeal to stats-watchers. The match itself would be determined strictly by vote totals, which would make it appeal greatly to those who dislike handicaps in matches (including myself, of course). Well, though I might have missed something, it does look to me like this could be the ultimate compromise position in the ongoing SDO debate. I think this idea could really give us the best of all worlds. I would be supportive of it, under the specifics stated above.
__________________
|
|
2013-11-04, 06:03 | Link #1784 |
NOM.
Join Date: May 2013
Location: Singapore
Age: 31
|
OK, that's one down.
However, the fact that SDO makes results easily manipulated still hasn't changed. For this, I propose that we bring back VD and VF — but with a twist. To encourage regular voting, we could use the geometric mean instead of the arithmetic mean so that not only each regular vote counts more, the results would be less manipulatable by a single faction as compared to using the arithmetic mean. So someone with, say, 6250 votes in one match and 4000 votes in another, will result in a geometric mean of 5000 votes for each match since √(6250*4000) = 5000 (wins and losses are counted separately). If it's still too manipulatable we can switch to negative power means such as the harmonic mean (-1 power), which may "mitigate the influence of large outliers and increase the influence of small values", said large outliers being due to manipulation of SDO as well as a lack of regular voting. Note that this dampening effect will reduce manipulation of results for as well as results against. However, the arithmetic mean should be kept for the postseason since manipulation does not affect the year's overall score nearly as much so the arithmetic mean is fairer there (not to mention that if it does, it should count as regular voting anyway). (EDIT: Also, using the arithmetic mean here would more accurately reflect their overall strength over the entire season than the geometric mean or any negative power mean.) Wins still matter, but I suggest 2-1-0 is better than the current system we have now which favours narrow wins over crushing wins (or even draws). The greater the difference, the lower the SDO becomes (from what I saw in the ISML forum's SDO thread, at least), which I hardly find fair, and it dampens the league. This should be changed to crushing defeats resulting in a higher SDO instead, so as to balance out the unfairness due to seeding of matches. Just as those who win against tough opponents be rewarded, so should those who win extremely convincingly against weaker opponents. This is only fair.
__________________
Last edited by iamadooddood; 2013-11-05 at 12:45. |
2013-11-04, 08:57 | Link #1785 | |
Senior Member
Author
|
Quote:
__________________
|
|
2013-11-15, 14:09 | Link #1786 | ||
Senior Member
Join Date: Oct 2006
Location: Planet Earth
Age: 54
|
Quote:
Quote:
Right now, I think biggest problem ISML has is the dominance of class of 2011 = those who were in 2011 ISML post season. All but one of ISML tiara winner can be found in 2011 ISML post season list. When you have small number of characters dominating the league, you make it very susceptible for all kinds of manipulation. One of my aim when designing a system for ISML is how much it can promote diversity. If we have very diverse groups of voters, and if we have many groups for many different characters, it becomes hard for any single group to taint the record, and we would be less susceptible for a deflation, like the one we saw when Rozen faction left in 2009 postseason . Therefore, I see much more reason to encourage the weaker character's fan base to come out and vote compared to stronger character's fan base. The long term health of ISML depends little bit more on what the losing side's voters will do, compared to what the winning side may do. ( The actions of winning side's voters are important as well. It just is that losing side voters are more likely to react in negative way, which often is of greater concern for the league, overall. ) In these aspects, rewarding a girl for crushing victory more than normal victory is likely to be more harmful than good. ------------------------------------ Most important of all, let's see what will happen in male+Touhou tournament where we don't have to worry about SDO nor VD. The first day match has been posted. ( http://internationalsaimoe.com/voting/ ) I'm sure all of you concerned with SDO can think up good tests to prove your point on what the results will be. I encourage any models and analysis strategy, provided that these models and analysis scheme is made known before the match starts. True test of models and hypothesis is to make verifiable quantitative predictions. I will accept submission of models or hypothesis until Nov. 24th. If your model and hypothesis succeed in predicting some important numbers ( it is up to you to tell me why a certain number or formula is important ) , I will give your proposal for necklace scoring system revision much greater attention.
__________________
Last edited by wontaek; 2013-11-15 at 15:04. |
||
2013-11-15, 14:47 | Link #1787 |
Gamilas Falls
Join Date: Feb 2008
Location: Republic of California
Age: 46
|
One thing to note in ISML is that the voters of the underdogs don't give up. Unlike in Saimoe were when a faction starts to lose, the faction mostly leaves, ISML, they press on and sometime are rewarded when the more dominate factions let down their guard. Every once in a while a mid-tier girl will beat a tier one girl for seemingly no reason (Ayu beating Nagato Yuki in one of the first years). Or the steady pacing of Tomoyo going from the 7th new entry girl to fighting for first place against Fate (before the finals).
__________________
|
2013-11-15, 15:07 | Link #1788 | |
Senior Member
Join Date: Oct 2006
Location: Planet Earth
Age: 54
|
Quote:
__________________
|
|
2013-11-15, 19:04 | Link #1789 | |
Senior Member
Author
|
Quote:
Because that is what the success of girls in moe competitions depends on: Which girl has the most people show up and vote for her. If you're a supporter of a top-tier girl, and she's facing somebody from a lower-tier, there's a natural temptation there to take the match off since you know your girl is going to win anyway, and she won't be rewarded for a large margin of victory. But if a girl is rewarded for large margin of victory, her supporters are more likely to consistently show up in every match, because then there is something at stake even in her more one-sided matches. If one of your main goals is making regular season matches as important as possible, and to encourage voters to show up as consistently as possible, then rewarding girls for large margins of victory makes perfect sense to me. What's actually counter-intuitive is rewarding close victories moreso than dominating victories, all else being equal (i.e. the strength/tier-level of the opponent remains the same). And I don't see why rewarding crushing victories would discourage lower-tier girls' supporters from showing up. Rewarding crushing victories is not the same as punishing those who suffer crushing defeats. And honestly, I see very little evidence that mid-tier and lower-tier girls do that well in ISML anyway. Their voters stick around due to a sense of dedication, or because eventually their girl will have a match where she has a chance of victory, due to how the ISML regular season functions.
__________________
|
|
2013-11-15, 20:03 | Link #1790 |
Part-time misanthrope
Join Date: Mar 2007
|
When you can trust maglor about something, then it's creating a logical conclusion from correct premises.
If you want something to criticize about ISML, then blame these math majors for coming up with the most complicated rating system they could think of for something as simple as a moe contest. |
2013-11-15, 20:32 | Link #1791 | |
I desire Tomorrow!
Join Date: Dec 2005
Location: As far away from reality as possible
Age: 41
|
Quote:
__________________
|
|
2013-11-15, 21:45 | Link #1792 |
ISML Technical Staff
Graphic Designer
|
The preview for the first voting day is up:
http://www.internationalsaimoe.com/voting/ Remember that in the first two days, only 2 from each group will advance, so 2/3 of the contestants will be eliminated. They might be the most important two days so don't miss them.
__________________
|
2013-11-15, 22:01 | Link #1793 | |
Senior Member
Author
|
Quote:
But if you're going to have highly elaborate rules designed to maximize voter involvement throughout the entire regular season, then it's very much open to debate what are the best rules to have. If you think that rewarding crushing victories is a bad idea, then please tell me why. Right now, I think it's a great idea if your goal is to maximize voter involvement throughout the entire regular season.
__________________
|
|
2013-11-16, 00:53 | Link #1794 | |
NOM.
Join Date: May 2013
Location: Singapore
Age: 31
|
Quote:
And before I can comment on anything else, I'll say this: I see what you've done. You've not gotten rid of the vote manipulation problem. You've pushed it elsewhere. End result: A lot of voters. Sure, there's a high chance that ISML will survive for a long time. But guess what? Due to the emphasis on close victories, the winner will end up being the faction that can manipulate the most votes. While pushing a single faction may be hard, multiple factions ganging up on a single one is easy. Chances are, when a lower tier girl loses to a higher tier one, some of those voting for lower tier girls are actually from the higher tier girl's faction. As for VD and VF being easily manipulatable, even if they were (which I doubt would be the case in later years, assuming it still is) at least they only affect the integrity of one character's placing. SDO affects multiple characters, a lot more so than VD or VF ever will, hence any manipulation of it will have greater effects, even if it wasn't as manipulatable. I can't comment on anything else yet because I'm still calculating whether this is a good thing or a bad thing. Is sacrificing voters' integrity for ISML's general longevity a good thing? Does ISML still even capture the meaning of moe anymore? And so on. As for the male + Touhou tournament, I don't think as many people care about them so manipulation is less likely. Of course, it still happens, but... (Still, as Triple_R puts it, having Anime Saimoe Tournament's knockout system is still the best, but the mentality of their admins remains to be desired.)
__________________
|
|
2013-11-16, 02:37 | Link #1796 | |
NOM.
Join Date: May 2013
Location: Singapore
Age: 31
|
Quote:
Also, I've only followed JAST for a few months (didn't even vote once) and I already knew what's going on because of, uh, the people talking about it in the thread? Seeing that ISML is international, I can tell what kind of voters it has just by reading this thread because people posting here are likelier to be ISML voters (the fact that a lot of ISML voters also follow JAST helps some). I've even gone to the ISML forum a bit (where they're even likelier to be ISML voters, for obvious reasons) and got a gist of how they're like. For example, most of them appear to be anti-SDO. That, or the anti-SDOers are making the most noise. That said, having to use SDO as weightage in the necklace period rounds just to filter out multivotes, is already a bad enough sign.
__________________
Last edited by iamadooddood; 2013-11-17 at 00:43. Reason: Mistake |
|
2013-11-16, 07:43 | Link #1797 | |
Senior Member
Author
|
Quote:
ISML does display age, gender, and nationality/regional statistical breakdowns for its rounds and matches. But I'm not sure what that has to do with what we're talking about here, though I'm certainly open to any explanation on that.
__________________
|
|
2013-11-16, 16:23 | Link #1798 |
Senior Member
Join Date: Oct 2006
Location: Planet Earth
Age: 54
|
1. Despite our best efforts, it is inevitable that some people will try to multivote. There always is a chance that we may miss some for couple of rounds. One of the aim of the current design is to minimize the impact these multivotes may have.
2. I have not found any real life example where promoting crushing victories had positive effect on the league as whole. One of the hopes ISML organizer has is for people to interact with other people with different thoughts. You can't have that unless you find ways to cheer up the fans of losing girls in hope that they will stick around. 3. Simple VD system showed how mere 50 voters can skew the whole thing in 2008. Only argument I have seen that says such a thing can't happen, is one that says because we have more voters compared to 2008. I have countered that argument by pointing out that mere 5% of the voters can successfully skew the results to their favor in VD/VF scheme, if those 5% votes in all the rounds. We have seen from events like Best Moe Tournament where a faction may bring in 2000 voters to all the matches. 2009 Rozen faction consistently brought in over 1000 voters. Do you wish to be at mercy of these factions? I shudder to think how bad it could have been if 2009 system was VD/VF, not SDO. 4. A faction that can manipulate the most votes will have the greatest influence in the result no matter what kind of scoring system you have. Only real way to counter this is to encourage diversity. It is my fear that simple VD/VF system for necklace will discourage diversity. 5. Any system has to pursue 2 potentially conflicting goals : (1)Abiding by Majority's desire (2)Protecting the voice of the minority . Simply going by VD will give people no incentive to court the factions that supports minor or weaker characters. 6. Most problems in 2013 necklace had its roots not in SDO system; the root of the problem is dominance of class of 2011 and certain factions tendency to show up only for "key" matches. ALL the systems discussed will reward voters that consistently vote much more than those who show up only for the important matches. For some reason, most of the outcry was against factions that diligently voted in all the matches. SDO system in most cases do better job of rewarding the faction the sticks around and be courteous compared to VD, where it is hypothetically possible to win the whole thing with massive single burst of support. 6-1 . It is very important to remember that some girls got stuck at very disadvantageous starting point due to her fans ignoring prelim seeding matches. Misaka Mikoto is a very good example. She got stuck in 2nd Tier due to poor performance in the seeding matches. No matter what scoring system we used, Mikoto's chance for necklace will suffer because her schedule being less favorable for the necklace. 7. If you check all ISML records from 2008 to now, you will see that there is clear inverse relationship between schedule strength and VD. Obviously, it is because you can build up your VD number much easier if your opponents are weaker. VD/VF system will ensure that a character that beats several 1-6 and 2-5 opponents will win the necklace over another character that faced two 6-1 and two 5-2 characters. Are you sure you want to reward someone for having easy schedule? Unless VD/VF system is modified to reward characters that beat strong opponents, we cannot accept that. SDO was adopted in 2009 because of this, and at that time, it had very wide support. The question comes down to this : Which character do you want to have the necklace? (1) Someone who beat two 0-7, two 1 - 6, one 2 - 5 , one 4-3, and one 6-1 characters with average margin of victory near 10%. (2) Someone who beat two 2- 5, one 3- 4, one 4- 3, one 5 -2 , and two 6-1 characters with average margin of victory near 7%. I am interested in your answers, but also will point out that in most of real world cases, people have overwhelmingly favored case resembling (2). This is why we have RPI ratings being used in College Baskeball and BCS system in College Football in USA which relies on polls and computer models that rewards teams which had harder schedule. I also know BCS is flawed, but all the alternatives suggested by someone serious about college football placed heavy emphasis on schedule strength.
__________________
Last edited by wontaek; 2013-11-16 at 17:35. |
Tags |
isml, saimoe |
Thread Tools | |
|
|