Splatfest results - A discussion and comparison

[LiX]

Inspired by a thread that got locked for it's rough approach, I decided to kick off a discussion here. Even when the gentleman in that other thread worded it drastically, he is partly right. Pearl and the topics she represents do win a lot in NA/EU, whereas Marina dominates in the same manner in Japan. For scientific and statistical reasons, I had a look at here and counted up the following:

Europe:

Number of Splatfests: 11
Victories (Pearl): 9 (81,8% of Splatfests)
Victories (Marina): 2 (18,2% of Splatfests)

NA:

Number of Splatfests: 11
Victories (Pearl): 7 (63,6% of Splatfests)
Victories (Marina): 4 (36,4% of Splatfests)

Japan:

Number of Splatfests: 11
Victories (Pearl): 3 (27,3% of Splatfests)
Victories (Marina): 8 (72,7% of Splatfests)

So you see, in Europa, Pearl absolutely holds the top spot for won splatfests with a whopping 81,8% of splatfests won. In NA, she is also very strong. In Japan, it's nearly the oppposite of Europe where Marina won nearly as much as Pearl did in Europe.

What do you think? Why is this so? Are European players just less paying attention to the topic and decide by their "favorite" host? Or does Pearl simply get the "popular" themes in Europe whereas Marina takes them in Japan? Why is Pearl's dominance slightly lower in NA and will it possibly get more balanced over there with Marina winning more?
Or is it all just a coincidence as the popularity vote is not always congruent with the winning scores, as today's splatfest shows, which Marina would have won, had the opposing team not barely won the solo category?

Will this become a vicious cycle in those areas? Will Pearl become even more popular in Europe as evervyone will think that "she wins anyway"?

 

[Gameboy224]

Character popularity is an irrelevant factor. Or at the very least a negligible one.

Pearl and Marina have little to no relevance other than being the representative of whatever trivial question is being asked. Pearl has been the popular option 3 times in NA, 5 times in EU, and 6 times in JP. Marinas been the popular option 8 times in NA, 6 times in EU, and 5 times in JP.

Popularity plays no factor in predicting the results. As it stands the popular team has won 14 overall Splatfests, the the less popular team has won 13 overall Splatfests. (Note that NA/EU do share some Splatfests, and Action v Comedy was a global Splatfest). 4 were clean sweeps.

In NA, we have a popular:unpopular win ratio of 5:6. EU has 5:5 (do not have data for EU's most recent). JP has 5:6.

Anybody who thinks they can actually predict the outcome is delusional. You can guess, and you may be right. But no more right than predicting a flip of a coin without foul play.

 

[MINKUKEL]

WARNING: LONG POST

I have a couple of things to say about this.
But, DISCLAIMER: I'm NOT a statistician, nor am I sure the data is fit for these kind of calculations.

1) You're counting some splatfests twice. Some took place in both NA and EU or even all three regions. So, those Pearl/Marina stats aren't right. I have a datafile with Splatoon stats made when the Nike Spaltfest was the latest fest. At that time it was 15 wins for Pearl and 10 wins for Marina. That's not too weird. You could flip a coin 25 times and get 15 heads and you wouldn't question the fairness of the coin.

2) At that time, the most popular team won 11 times, and the least popular team 14 times. Seems pretty normal. However, you would expect the most popular team to win more often since they already have one of three points in the pocket.

3) Now, if you look at how often a team won both the Solo and the Team categories, things get a little weirder. You get this:
-Most popular team won both Solo and Team: 4
-Most popular team won neither: 11 (in other words, least popular team won both)
-Most popular team won only one: 10

Now, here's the thing. Like I said, I am not a statistician (though I have to do a lot of statistics for my studies). So, I wasn't sure based on this rather small sample size if the data was distributed normally or not. So, to be safe, I looked at the relation between Popularity and Wins for both the scenario the data is and is not normal.

If you take into account that both of those scenarios could be true, you can conclude that popularity might have an influence on win percentage. BUT, not a very big one.

There is a very small (negative) correlation between Popularity and Wins (I combined Solo and Team Wins, controlling for the fact Solo matches outnumber Team matches 19 to 1) if the data is normal. So small, in fact (-0.174), that it's negligible, and it would mean Popularity contribute only <3% of all variation in Wins. BUT, the data is not significant, so you can't even say for sure that that's true. In other words, if that data is normally distributed, there's not really a lot going on.

However, if we assume the data isn't normal, and we judge the data by how it ranks compared to other splatfests instead of the absolute values, we get a slightly different result.
Here, the correlation IS significant, but only when you combine the Solo and Team wins. Considering there is quite a bit of difference between Solo and Team if you do it this way, I'm not sure if this would be correct, so to be on the safe side, I looked at both of them seperately instead.

Popularity would contribute about 6.5% of the variation in Solo, and about 10%* in Team.
Combining them you get about 17%. But only by combining them do you get a statistically significant result.

I don't really feel confident enough to give a definite answer on the non-normal calculations, but you could say that if there is a statistically significant effect, it would be moderate. Not big, not all that small either.

4) Of course I could've made mistakes or assumed something that I shouldn't have.
But what does this crap even mean?
Well, ultimately it doesn't really matter all that much anyway. There's a fair chance of winning if you're in either team, which is the most important part IMO. Also, while there seems to be something going on here (possibly), this could be any number of things. Your guess is as good as mine. Unless your guess is that it's a conspiracy, in which case your guess is much worse.

And please keep in mind that, again, the data isn't a binary win/lose thing.

If you're good at stats, please do comment.

 

[supgamer76]

ive found the more popular looses. well from my experience

 

[The Salamander King]

I believe the more popular team usually loses because of player size. The more people on one team, the harder it is to find players on the opposing side. The popular team has a higher chance of fighting their own players, which does not alter the score whatsoever. Thus, the "hipster" team gets or loses points more often than the other.

 

[Gameboy224]

I believe the more popular team usually loses because of player size. The more people on one team, the harder it is to find players on the opposing side. The popular team has a higher chance of fighting their own players, which does not alter the score whatsoever. Thus, the "hipster" team gets or loses points more often than the other.

Irrelevant. There will always be the same number of matches that count played by each team. Only difference is the people playing those matches.

The Splat Power system exists purely to mitigate any difference in skill population. Population size for each team actually doesn't play too much of a factor. The game actives tries to only pair you with players of similar general skill.

Hence why if you look at the Splatfest results ever since the implementation of Splat Power in Splatoon 1 and all the Splatfests in Splatoon 2 have been significantly varies.

 

[AllToonedUp]

Irrelevant. There will always be the same number of matches that count played by each team. Only difference is the people playing those matches.

The Splat Power system exists purely to mitigate any difference in skill population. Population size for each team actually doesn't play too much of a factor. The game actives tries to only pair you with players of similar general skill.

Hence why if you look at the Splatfest results ever since the implementation of Splat Power in Splatoon 1 and all the Splatfests in Splatoon 2 have been significantly varies.

But the game tries to match you up with players of the same team if it takes too long to find the opposing side and the score only counts if you fight and win against the enemy team.

The last 3 splatfests in 1 + Vampire VS. Werewolves and Sweater VS. Socks had both popularity and wins for their victories as well, so I don't think Splatfest Power is a reliable factor. I think @The Salamander King's theory is more plausible.

 

[sakono]

Out of all the games I played for the Raph vs Leo splat-fest I played against another team Leo about 90% of the time. same with my roommate. Yet Raph still won. I think this skews it so the lest popular one wins.

I saw lots of posts online of others on team Leo complaining about not fight more team Raphs too.

 

[calamaro]

I'm still trying to figure out what the reason would be to fix a splatfest is in the first place...

Maybe the more popular has more bad players and that tilts the scales to the other side?

 

[Gameboy224]

But the game tries to match you up with players of the same team if it takes too long to find the opposing side and the score only counts if you fight and win against the enemy team.

The last 3 splatfests in 1 + Vampire VS. Werewolves and Sweater VS. Socks had both popularity and wins for their victories as well, so I don't think Splatfest Power is a reliable factor. I think @The Salamander King's theory is more plausible.

Ever since the implementation of the Splat Power system, most solo win differences typically remain close to 51~52%. Highest any Splatfest has differed since was 55%.

Compare this to 5 Splatfests in NA immediately prior to Splat Power, the lowest win difference was 55%, and it went as high as 59%. Heck, JP's last splatfest before Splat Power had a win difference at 64%.

The fact that up to this point no solo score has broken 55%, with the vast majority not even breaking past 51~52%, shows that the system is indeed working. Teams tend to show a larger on average difference, but have still never broken 56%.

My point is Splat Power is supposed to make the win score much less predictable since the game will actively try to make matches where it can go either way.

 

[MINKUKEL]

Thus, the "hipster" team gets or loses points more often than the other.

The fact that you can both win and lose points is precisely why it doesn't matter.
Only team A vs team B matches do anything.

 

[Cephalobro]

The problem with the current system is that it disbands the team automatically after a match, it wasn't this way in Splatoon 1. I know there are team matches, but it forces the player to make a quartet. If it allowed the player to make a duo and the game matches them with other duos much like Squad Battles in Splatoon 1, then it wouldn't be very one-sided in terms of Splatfest like it is being right now.

 

[LiX]

Thanks to all participating in the discussion. I've had a read and a lot of interesting facts and ideas have been said. I'd like to cmment on a few.

1) You're counting some splatfests twice. Some took place in both NA and EU or even all three regions. So, those Pearl/Marina stats aren't right. I have a datafile with Splatoon stats made when the Nike Spaltfest was the latest fest. At that time it was 15 wins for Pearl and 10 wins for Marina. That's not too weird. You could flip a coin 25 times and get 15 heads and you wouldn't question the fairness of the coin.

Technically, you are right that the splatfests that NA/EU shared are counted "twice" in a sense here. But that's why I seperated it by regions and made an attempt to calculate a quota for each region isolated. Naturally, I had to count every splatfest that (also) happened in a different area then.

However, the problem is that we don't have any region seperated data. It would be interesting to see if maybe Marina won slightly in EU but Pearl took a big win in NA, so that the numbers added up resulted in a slight win for Pearl overall in one of those "shared" splatfests. Does anyone have access to that data? It would allow for more clean calculations.

ive found the more popular looses. well from my experience

That's an interesting theory stated by multiple players here. Are there any statistics I am missing on that underline this? If so, it'd be cool to see them linked.

Why would the popularity influence the win/loss ratio in a negative way? Why would the more popular team generally have "worse" players? That doesn't seem to be set in stone.

From the "feels", it also didn't seem to be that way. I've seen 3 - 0 wins and a lot of Pearl wins in Europe appeared to be popularity + any battle category. But I'll happily accept if statistics disagree.

The problem with the current system is that it disbands the team automatically after a match, it wasn't this way in Splatoon 1. I know there are team matches, but it forces the player to make a quartet. If it allowed the player to make a duo and the game matches them with other duos much like Squad Battles in Splatoon 1, then it wouldn't be very one-sided in terms of Splatfest like it is being right now.

I definitely agree that it would be decent to be able to go in as duo like in league right now. I'm sure enough people would use that option to make it a viable implemention.

 

[MINKUKEL]

However, the problem is that we don't have any region seperated data.

That would be a rather arbitrary split in the data and reduce each region's sample size to way too small to be of any use anyway. There really is no point in looking at the results on a region by region basis.

I've seen 3 - 0 wins

Yeah, but 3-0 wins are way less common than 2-1 wins in favor of the least popular one.

 

[Flammie]

Irrelevant. There will always be the same number of matches that count played by each team. Only difference is the people playing those matches.

The Splat Power system exists purely to mitigate any difference in skill population. Population size for each team actually doesn't play too much of a factor. The game actives tries to only pair you with players of similar general skill.

Hence why if you look at the Splatfest results ever since the implementation of Splat Power in Splatoon 1 and all the Splatfests in Splatoon 2 have been significantly varies.

A team of players overall splat power divided by FOUR.
Plus most of the players may not have played with each other before.
One guy in the room may have over 2300 points compared to the others who have less than 1900, which averages around 2050 something.
So the average score rounded by 4, doesn't mean everyone in the team is the exact same level of points.

Plus the more popular team by far will have team facing against their own team choice, in odds that those players against their own splatfest team would be really good, that would increase the chances of the less popular team winning, even if it's between an increase of 1% to 10%.

 

[Gameboy224]

A team of players overall splat power divided by FOUR.
Plus most of the players may not have played with each other before.
One guy in the room may have over 2300 points compared to the others who have less than 1900, which averages around 2050 something.
So the average score rounded by 4, doesn't mean everyone in the team is the exact same level of points.

Plus the more popular team by far will have team facing against their own team choice, in odds that those players against their own splatfest team would be really good, that would increase the chances of the less popular team winning, even if it's between an increase of 1% to 10%.

I did say try, I know Team Splat Power is an average. The idea is the overall balance of the team will be of similar level. One person may be 2200, and another may only be 1800, but that ideally balances out at 2000 and ideally a similar thing is happening on the other team, maybe they have one higher skill and one lower skill player of all of them are roughly 2000. Point being, the average ability of the teams should be similar.

So if your team's average is a 2000, you won't be going against a team who's average is a 1700.

Also the popular team having matches against their own team plays very little factor. Yes, those players aren't contributing when going against their own team, but that doesn't mean some other player on your team isn't, and nor that does that affect if those players are good or bad.

 

[MINKUKEL]

in odds that those players against their own splatfest team would be really good, that would increase the chances of the less popular team winning, even if it's between an increase of 1% to 10%.

You can say the same for a really bad player instead of a really good player.

 

[LiX]

That would be a rather arbitrary split in the data and reduce each region's sample size to way too small to be of any use anyway. There really is no point in looking at the results on a region by region basis.

How is a continental split arbitrary and not of any use? This whole thread is based on the idea of looking at splatfests results in the three big regions (NA, Europe, Japan) and comparing/discussing the possible preferences of characters or their topics. Heh.

 

[Flammie]

You can say the same for a really bad player instead of a really good player.

Let us say we have 100 people on one team, and they have a skill level between 1 and 100, and each of them have a different number.
The other team have 80 people with skill level between 1 and 100,(So that mean every fifth person is skipped so 1 2 3 4 then 6 7 8 9 etc. til you get 100. If you pin all 80 against 80 selected random from the 100, are the chances of winning increased or decreased for the less amount team?

If also what you are saying is true, why aren't there more 3 - 0 Splatfest scores?

 

[MINKUKEL]

are the chances of winning increased or decreased for the less amount team?

Neither.

If also what you are saying is true, why aren't there more 3 - 0 Splatfest scores?

That's weird, and I don't know. But that has nothing to do with this. It even includes the relation between Solo and Team outcomes, which opens up a whole nother can of worms.

How is a continental split arbitrary and not of any use? This whole thread is based on the idea of looking at splatfests results in the three big regions (NA, Europe, Japan) and comparing/discussing the possible preferences of characters or their topics.

If there was any reason for the regions to behave differently when it comes to Pearl/Marina wins/loses it wouldn't be arbitraty, but I see no reason to believe that.