RESULTS: 2013 Prelims II-1

[Desufire]

If there was one in SAO, she would have gotten in the main season at the end of the first phase with little difficulty.

They pretty much all did do that anyways...

Side Note: http://archive.internationalsaimoe.com/ ... _p9-15.gif" onclick="window.open(this.href);return false;

I smell a bot in China

Why? I see darker color in Malaysia.

Im guessing theyre are more voters in China so it would be easy to hide a bot among those voters.

[maglor]

Now that's mentioned, can there be a possible explanation for that strange result of the regular season match last year where homuhomu went up against Kuroi Mato and didn't lose so badly? I'm probably thinking too much, but I'm still suspecting something to do China.

Anti-Homura voters have slight tendency to be anti- Hanazawa Kana. I guess for some, Anti-Hanakana sentiment was strong enough for them to vote for Homura instead.

[maglor]

If there was one in SAO, she would have gotten in the main season at the end of the first phase with little difficulty.

They pretty much all did do that anyways...

Side Note: http://archive.internationalsaimoe.com/ ... _p9-15.gif" onclick="window.open(this.href);return false;

I smell a bot in China

Why? I see darker color in Malaysia.

Im guessing theyre are more voters in China so it would be easy to hide a bot among those voters.

If so, how did you "smell" a bot from China despite the difficulty? How can that map tell you that there is a bot of significant power in China ?

[Desufire]

Suspicions. I'm not saying anything is a fact, I just find it slightly unlikely with all that orange, that there wasn't any cheating going on. Especially when brought into context what you mentioned about the Hidan no Aria voters...

[RegalStar]

China has always been blatantly pro-kugi and pro-any-show-that-has-kugi-tsundere-in-it.

[xcrossfacekillahx]

Yet somehow China is still a won territory for Mato which means there are still Anti-Homura voters rather than the Pro-Kugi voters.

[Akka]

Meh.
Lots of arena where I don't give a sh..., and the actual best moe character (those from Tari Tari & Kokoro, Watashi, Ibara...) are scoring low. while the boring stock characters from weirdly popular shows are curbstomping them.
Depressing.

[hinakatbklyn]

If there was one in SAO, she would have gotten in the main season at the end of the first phase with little difficulty.

They pretty much all did do that anyways...

Side Note: http://archive.internationalsaimoe.com/ ... _p9-15.gif" onclick="window.open(this.href);return false;

I smell a bot in China

Why? I see darker color in Malaysia.

Im guessing theyre are more voters in China so it would be easy to hide a bot among those voters.

If so, how did you "smell" a bot from China despite the difficulty? How can that map tell you that there is a bot of significant power in China ?

http://archive.internationalsaimoe.com/ ... _p9-24.gif" onclick="window.open(this.href);return false;

Closer matches like less than 100 I could see lead changes during the day. But a lead of 200 or more for Himeko within the last 5 hours turns into +100 for Neko (Net gain of 300)? Already reminded of 2011 KBM's bots/multi-voters. More impressive to get a similar vote rebound in ISML when more voters can counter a rebound than with KBM.

[Deathscyther]

I liked most results. I think Rin has what it takes to win her group. Kurisu and Yuki did great as well.

And wow, Kurugaya did much better than I expected her to.

[kuuderes_shadow]

http://archive.internationalsaimoe.com/ ... t_p9-2.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... t_p9-7.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... _p9-10.gif" onclick="window.open(this.href);return false; (this one is a bit less)
http://archive.internationalsaimoe.com/ ... _p9-12.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... _p9-13.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... _p9-15.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... _p9-16.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... _p9-19.gif" onclick="window.open(this.href);return false; (in this one all 3 characters got a boost so could be coincidental)
http://archive.internationalsaimoe.com/ ... _p9-24.gif" onclick="window.open(this.href);return false;

In every one of these cases, at 18 hours into the vote one character saw a sudden surge in their votes, while neither of their competitors did. Some are more extreme than others but for exactly the same bizarre surge to happen at exactly the same time, with an impact of at least a couple of hundred votes in no less than 9 arenas, potentially changing 3 or 4 results in the process, is certainly something that I would consider worth investigating.

Furthermore, there was no surge at all at the 18 hour mark during any of the prelims phase 1 days, so this cannot be attributable to there simply being a lot of people voting at this time.

[maglor]

http://archive.internationalsaimoe.com/ ... t_p9-2.gif
http://archive.internationalsaimoe.com/ ... t_p9-7.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... _p9-10.gif" onclick="window.open(this.href);return false; (this one is a bit less)
http://archive.internationalsaimoe.com/ ... _p9-12.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... _p9-13.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... _p9-15.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... _p9-16.gif" onclick="window.open(this.href);return false;
http://archive.internationalsaimoe.com/ ... _p9-19.gif" onclick="window.open(this.href);return false; (in this one all 3 characters got a boost so could be coincidental)
http://archive.internationalsaimoe.com/ ... _p9-24.gif" onclick="window.open(this.href);return false;

In every one of these cases, at 18 hours into the vote one character saw a sudden surge in their votes, while neither of their competitors did. Some are more extreme than others but for exactly the same bizarre surge to happen at exactly the same time, with an impact of at least a couple of hundred votes in no less than 9 arenas, potentially changing 3 or 4 results in the process, is certainly something that I would consider worth investigating.

Furthermore, there was no surge at all at the 18 hour mark during any of the prelims phase 1 days, so this cannot be attributable to there simply being a lot of people voting at this time.

(1) Thank you for your input. We are keeping an eye on this.
(2) Someone mentioned KBM, thus I need to show you something that happens VERY OFTEN in votes from MOST of EAST ASIA



There are very large anime series fan clubs in East Asia. They sometimes take interest in voting in events like KBM or ISML. The key point is that while some groups consists mostly of teenage student, some others are made up mostly of people in late 20s and beyond. Thus they tend to start voting when they get home from Middle/High School or work. During Weekends, this jump in voting usually happens around 8am, the time for breakfast, thus the jump you see in the graph above. KBM staff knows this is legit votes because they monitored fan club activities and also asked many voters the time they usually votes. For weekdays in East Asia the point when they usually arrive back at home from school or work is around 6PM which is right at the 18th hour point. If you look at vote graphs for East Asian countries, that jump at 18th hour always was there. If you see some time graphs from 2008 to 2010, you can see similar jumps there as well. In recent days, large influx of votes from all over the world drowned out this jump. What happened in the last match was that some key region's percentage of total votes increased while many others decreased, thus we were able to see this again.

(3) Here is the latest example from KBM. Do note that most of these voters are registered voters, and that recently registered voters needed to go through 2 hour vote delay in order to vote. In addition, KBM has many vote checking scripts running.



(4) Therefore, to our best knowledge this jump is legit, as most of it came from regions that consistently showed such patterns. We always have kept an eye on this, and will take action when we uncover wrong doings.

[maglor]

Also, Let's see how "little" that affect was during Phase I. Here are the last 3 arenas of Phase I last day







Even here, there were the jumps around 18th hour. The reason why they are weaker is

(1) Many voters left between Phase I and II. Many new voters came in as well and large percentage of them were from East Asia that shows 18th hour jump

(2) Pick 2 option tends to reduce radical behaviors by any group. So, in Phase I, because faction A supporters also were voting for Character B, C, or D , the jump got spread out among all the characters. Pick 1 nature of phase II has revealed what was somewhat hidden in Phase I.

[Zeando]

thanks for the explanation

in some cases it really looked odd, like:


i'm more surprised, given the international range of this tournament, of the results being usually so much linear
are there any data about the votebase nationalily(/timezone) composition?

[Bastion]

To break the chain of claims of multivoting (:P)
Saki is still losing, and Mikan's battle is a complete flip of what I expected otherwise, everything as foretold, maybe Sex Doesn't Sell in this case

[DurianDude]

To break the chain of claims of multivoting (:P)
Saki is still losing, and Mikan's battle is a complete flip of what I expected otherwise, everything as foretold, maybe Sex Doesn't Sell in this case

Sex doesn't sell super well in Saimoe. Otherwise Yuiko would have easily won.

[maglor]

thanks for the explanation

in some cases it really looked odd, like:


i'm more surprised, given the international range of this tournament, of the results being usually so much linear
are there any data about the votebase nationalily(/timezone) composition?

There is, but because certain countries have larger proportion of votes that are extremely likely to be multivotes, I'm afraid I can't divulge that. I would like to add that for Korea, the Phase II Round 1 voting day coincided with the first day of school for 2013( Remember that East Asia is 8 to 9 hours ahead of GMT ) , thus many of the voters decided to not vote early, but vote late after their school was over. It also was start of Spring Break for many USA colleges. All these, though small by itself, may have added up to create this bump.

[10ZHAbin]

And one should not simply ignore the regular season match of Honma Meiko vs shimada minami from last year, china have a lot of voters. It make a even more difference in the first day of male match last year, and the incident with hinata happened.

And just a question, am I the only ISML voter in New Zealand? Since it combines with Australia in the ISML color map.

[Zeando]

thanks for the explanation

in some cases it really looked odd, like:


i'm more surprised, given the international range of this tournament, of the results being usually so much linear
are there any data about the votebase nationalily(/timezone) composition?

There is, but because certain countries have larger proportion of votes that are extremely likely to be multivotes, I'm afraid I can't divulge that. I would like to add that for Korea, the Phase II Round 1 voting day coincided with the first day of school for 2013( Remember that East Asia is 8 to 9 hours ahead of GMT ) , thus many of the voters decided to not vote early, but vote late after their school was over. It also was start of Spring Break for many USA colleges. All these, though small by itself, may have added up to create this bump.

i wasn't really asking for them, nor i'm talking about the bump now , rather if there is a way to know, compared to the national population, if the saimoe is more west centric, or anyway not really internationally distribuited

referring to how, except these 18h cases, the results are very linear, leading me to think of some cases for it to happen:
-there isn't a relevant difference in popularity for voters around all the world, in literal sense, meaning in any nation, continent, more or less there are the same fandoms (not true as we can see from the regional graphs)
-there isn't a relevant difference in popularity for voters around all the world, in practical sense, meaning only a part of the world is mostly influencing the results, while other zones with different preferences aren't enouth numerous to generate a visible change in the results during the voting (here why i was asking about real internationally distribution of the voters)
-the votes coming from different preferences somehow sum up to make a mostly linear result, at any given time during the 24h of voting (possible, but to have a linear results any time it seems more than a coincidence)
-there are too few voters globally to be able to see some relevant change, still it seems strange, even with few voters some changes should be visible

on a different note there are some things helping a more linear result, but they're unrelated to the voter national origin
-keeping the current votes hidden during the voting avoid for organized reaction votes to happen

[maglor]

i wasn't really asking for them, nor i'm talking about the bump now , rather if there is a way to know, compared to the national population, if the saimoe is more west centric, or anyway not really internationally distribuited

referring to how, except these 18h cases, the results are very linear, leading me to think of some cases for it to happen:
-there isn't a relevant difference in popularity for voters around all the world, in literal sense, meaning in any nation, continent, more or less there are the same fandoms (not true as we can see from the regional graphs)
-there isn't a relevant difference in popularity for voters around all the world, in practical sense, meaning only a part of the world is mostly influencing the results, while other zones with different preferences aren't enouth numerous to generate a visible change in the results during the voting (here why i was asking about real internationally distribution of the voters)
-the votes coming from different preferences somehow sum up to make a mostly linear result, at any given time during the 24h of voting (possible, but to have a linear results any time it seems more than a coincidence)
-there are too few voters globally to be able to see some relevant change, still it seems strange, even with few voters some changes should be visible

on a different note there are some things helping a more linear result, but they're unrelated to the voter national origin
-keeping the current votes hidden during the voting avoid for organized reaction votes to happen

(1) 3 things I can divulge without serious implication. (a) Number of votes from Asia, excluding Russia, is bigger than votes outside of Asia + Russia . This isn't surprising because Asian Continent holds 60% of World population, and since Anime is from an Asian country, you might say 70+% of Anime viewers are in Asia. (b) Number of Legit votes from outside of Asia never fell below 1000 since 2010. (c) Outside Asia voter numbers fluctuate by more than 100 , sometimes by more than 1000, from any match to the next match since 2011. Since there is a difference in taste between Asian and Non-Asian voters, this fluctuation often becomes the key reason why certain character won or lost.

(2) What I am trying to say from (1) is that even though non-Asian voters number less than Asian voters, the number is great enough that mere fluctuation of just the number of non-Asian voters often are good enough to determine who wins and loses.

(3) East Asian voters are more likely to be below age of 18 and also be female. Throughout the world, Above 18 voters usually vote within first 2 hours of start of vote, or in random time spread out through the remaining 22 hours. Due to this random nature of time of votes by above 18 year old voters, despite that each nation has their own preferences, the time-vote graph looks rather straight between 3rd to 15th hour due to all the random fluctuations cancelling each other out. Below 18 voters, in other hand, do tend to vote between 6 to 9pm mark in their local time, unless they vote within first 2 hours of start of voting. This time concentration and regional concentration of below 18 voter contributes to the bump we often see in time-vote graph.

(4) Thus

-the votes coming from different preferences somehow sum up to make a mostly linear result, at any given time during the 24h of voting (possible, but to have a linear results any time it seems more than a coincidence)

This is closest to what is happening for most of the times, except that linear result is not coincidence, but very natural outcome due to central limit theorem ( http://en.wikipedia.org/wiki/Central_limit_theorem" onclick="window.open(this.href);return false; ) . A easy way to visualize this may be to imagine number of ISML votes per minute to equal to a number you get from rolling a 6 sided dice . If you try that, you might get 6 votes in the first minute, 1 vote in the second minute, 5 votes in the 3rd minute, and so forth, but you will see that you will seldom be far away from 210 votes per hour rate in any hour, thus the whole time-vs-total graph would look pretty linear.

(5) If you wish to carry the thought experiment even further, imagine that you have divided the world into about 5 regions and assigned a dice to each of the regions. This would mean you would be rolling 5 dices every minute. Even though you get random results from all the dices, in end, you will see the 60 minute = hourly trend to be very near 1050 votes per hour in any hour. If you calculate Standard deviation of each hourly rate and divide it by average of all hourly rate, you will find this stdev/average ratio be much less than what you will get from (4) . This is why we usually get linear results most of the time.

[Zeando]

(1) 3 things I can divulge without serious implication. (a) Number of votes from Asia, excluding Russia, is bigger than votes outside of Asia + Russia . This isn't surprising because Asian Continent holds 60% of World population, and since Anime is from an Asian country, you might say 70+% of Anime viewers are in Asia. (b) Number of Legit votes from outside of Asia never fell below 1000 since 2010. (c) Outside Asia voter numbers fluctuate by more than 100 , sometimes by more than 1000, from any match to the next match since 2011. Since there is a difference in taste between Asian and Non-Asian voters, this fluctuation often becomes the key reason why certain character won or lost.

(2) What I am trying to say from (1) is that even though non-Asian voters number less than Asian voters, the number is great enough that mere fluctuation of just the number of non-Asian voters often are good enough to determine who wins and loses.

ah ok, that drops the possibility of a very uneven international distribution of the voters
leaving basically only 2 groups with different preferences which mostly influence the results and so only 1 visible bump in there (i've noticed there is also a very little bump around 7-8h near the end of phase1)

(5) If you wish to carry the thought experiment even further, imagine that you have divided the world into about 5 regions and assigned a dice to each of the regions. This would mean you would be rolling 5 dices every minute. Even though you get random results from all the dices, in end, you will see the 60 minute = hourly trend to be very near 1050 votes per hour in any hour. If you calculate Standard deviation of each hourly rate and divide it by average of all hourly rate, you will find this stdev/average ratio be much less than what you will get from (4) . This is why we usually get linear results most of the time.

that explains well the total voting being linear,

even if before i've misused the linear term while i was meaning more vote difference linear growth, but guess that too gets all scattered and distributed over the 24h
so the only way to see a non-linear growth in the vote difference would be by reaction voting? since it seems unbiased votes normaly result with linear vote differences

anyway thanks for the little infos, they were more than enouth for my curiosity/thought experiment :3

one assumption i was wrong on was assuming the vote timing for over 18 to be similar to the one of under 18, while from what you said over 18 tends to be far more random with the vote timing, excluding the first 2 hours, leading to timezones differences not influencing the regional timing in voting
said regional vote timing was what had lead me to find the linear results to be unusual
but since there is almost no influence from the regional/timezone vote timing, then a linear result for the most part makes more sense