More accurate fact check: Zerg wins every premier tournament

General Discussion
1

So far there has 15 premier tournaments played in 2019, out of those 15 tournaments zerg won 9 tournaments with 5 different winners.

As far as i can tell there has never been a year so dominated by one race when it not only comes to ammount of tournaments won but also ammount of different winners.

The reason i chose to focus on premier tournaments and not include major and minor tournaments is to avoid tournaments where the skill descrepency is so big it impacts the results, wich we most likely wont find in a premier tournament.

28 Likes
2

What evidence do you have to support this claim? Did you go through every premier and non-premier tournament, calculating the variability of skill in each, and showed that premier tournaments had a smaller skill discrepancy? Or is this purely an assumption?

Let me introduce you to the law of law numbers:

https://en.wikipedia.org/wiki/Law_of_large_numbers

What this law states is that you can’t define an average for a small sample. Think of it like this. What is the average of a single coin flip? It’s either 100% heads or 100% tails. Does a sample of 1 define the rate of heads and tails? Nope! You have a bad sample that is too small to get a good average.

The same applies to premier tournaments and SC2. 15 is not enough to get a good average. You need at least 30, preferably more. Does this mean zerg is OP? No. Does this mean zerg is UP? No. Does this mean zerg is balanced? No. What this means is that the listed evidence is incapable of supporting any claim because the sample size is too small.

We can perform this experimentally. Here is a website that flips coins for you:

https://www.random.org/coins/?num=10&cur=60-usd.0025c-nj

Here is my first flip: 4 tails, 11 heads. 36%/63%.
Second flip: 7 tails, 8 heads. 46% / 54%.
Third flip: 9 tails, 6 heads. 60%/40%.

See? 15 is not large enough to give an accurate average. Now lets average all three sets, for a total of 45: 20 tails, 25 heads. 45%/55%. As you can see, even with a sample of 45 our result still varies by a difference of 10%. That’s why with a statistical test, you need a >90% confidence (or even better, >99%) because it minimizes the odds that you are wrong. To get a higher confidence, you need a sample even larger than 45.

15 is just too small of a sample to have any meaning. This evidence is insufficient to support your claims. You either need many more premier tournaments, or you need to include non-premier tournaments. The variability with small sample sizes is too large to be able to measure anything.

3 Likes
3

I’m sorry to disappoint you but the law of large numbers is completely unrelated to this situation. Think about what youre saying. To apply the law of large numbers you need to specify a random variable - which in your case is a model for the distribution Zerg wins in trournaments (you have provided no such model, it seems like you are implicitly assuming that its a biased coin i.e. a bernoulie distribution which seems incredibly silly to me). Now law of large numbers says that if you take the average over N tournaments as N goes to infinity this converges with probability 1 to the mean of the random variable. There is no actual uniform number N in the (weak) law of large numbers which says that if you did N trials and then the average is accurate. This is because this “actual number” depends on the variance of the distribution something you haven’t taken into account. Its true that using the central limit theorem you can show that for a biased coin you’d need ~100 coin flips to be 90% sure about the bias of the coin. But this DEPENDS ON THE DISTRIBUTION. In particular the variance. So now you need to convince everyone that the distribution of Zerg’s winning a tournament is most accurately modelled by a bernouli random variable. This is complete nonsense. Its like modelling wind velocity with coin flips - its much to complicated to be modelled by this stupid thing. If anything the distribution of Zerg winning a tournament would have significantly less variance because players are pretty consistent and this game is decided mostly by skill (barring any balance issues) which is also pretty stable factor, at least throughout the span of a single year. This will lower the variance significantly. Which would also lower the number of trials required to be sure about the “bias”.

I’m not gonna do this calculation for you because I think all of this is also completely irrelevant because anyone with half a brain looking at this game can see that something is slightly broken about Zerg at the moment. The number of tournament wins only reinforces this fact.

16 Likes
4

Wooow…what an answer to the douchebag trying to negate zerg is over the top right now.

Come on rabiDrone…that missed math class…

6 Likes
5

First of all, I really don’t understand why people complain about this specific topic so much. By now, one would think that if a race was overly weighted they would be making major changes as opposed to smaller ones.

Hmm. Though, I admit, if Blizzard wanted to let fungals shoot from Overlords when we overcharged them or to randomly just give us a floating Zerg turd that would be able to roll around, but freeze while firing, and just launch out bile that did 400 AOE damage to biological units in the next patch, I would not be against it. :joy:

To Evaluate Races:
I guess the easiest way to do it would probably be to do what Blizzard is doing right now. Taking note of the ladder, pro players and their opinions of the races while making small changes every now and then.

Would it not be the best way to do this though if Blizzard evaluated the skill of a large set of players, based off a large criteria (giving certain skills that aren’t uneven balance [if possible] according to the races abilities that may not equivalent out to each other), then to assess the players of equal skill based on how they did on the ladder?

  • by the way, the bold part of this is more or less being done now by Google’s AI. Everyone take a moment to thank Google please, as they will forever be helping to bring an end to this conversation. :grin: Then once all the AI’s are at the same level, have them play against each other?
6

Hmmmm, maybe you missed all the commentary on the new balance update, but most people consider at least some of those changes to be pretty major. I don’t remember anyone of any race participating in the update threads saying “Pffft, this is all really minor stuff.”

1 Like
7

Oh really?

Oh, so it IS relevant. Got it! When you cannot write even one paragraph without contradicting yourself, it is time for a new argument.

I actually did. I talked about variance quite a bit in my first post. If you claim otherwise then you didn’t read my post.

Uh oh, someone failed their statistics class. A Bernoulli trial is a statistical test in which there are only two options, success or failure. Guess what happens when a player plays sc2? They win or lose. The properly normalized sum of bernoulli trials is a binomial distribution. That’s a mathematical fact.

Yep, we have a wikipedia copypasta troll. He doesn’t actually know what a sum of bernoulie trials is.

1 Like
8

You saying you talked about variance demonstrates exactly the point that you have no idea what variance actually is. The same applies to your use of the term “bernoulli trial”. Thanks for making my arguments for me. As for the rest of your dribble I’m sorry I’m too lazy to respond, I have already said all that needs to be said, I don’t feel there’s any point in arguing with you anymore I mainly wrote this for the benefit of all the people on this forum wasting their time arguing with you so that everyone could just chill out and realize how silly your arguments actually are. I will not respond any further, sorry if that disappoints you in any way. Good luck with your future endeavors.

3 Likes
9

Im reasonably certain that batz has just read a Wikipedia article on statistics, because every time he tries to apply them, somebody else comes in and explains why he’s wrong, and then all he can do is condescendingly dismiss them as not actually understanding statistics or data without actually being able to demonstrate any particular error on their part.

5 Likes
10

Lmao. No, it proves you didn’t read my post in the slightest. The entire point of needing to sample and calculate an average of the sample is because the results vary. I even linked to the law of large numbers, and went through examples and showed how variance requires a large sample. It is factually incorrect to claim that I did not talk about variance.

Kid, you didn’t refute even a single point that I made. You even agreed with me on the majority of my points, but you say you proved how silly my argument is? That’s laughable.

11

I’m really starting to feel sorry for you so I’m violating my promise not to respond. To understand what variance is you should read about the following topics in the following order:

  1. Probability space

Edit: After some discussion I realized the following topics are crucial for you to understand the flaws in your arguments but weren’t in the original list:
1.a Events
1.b Independence (VERY IMORTANT! A lot of your mistakes are due to this).

  1. Random Variable
  2. Probability Distribution
  3. Mean (expectation value)
  4. Variance

Wikipedia is certainly your friend but try not to cut any corners and work out some simple examples (like bernouli/binomial distributions) to make sure you understand. Good luck :slight_smile: .

2 Likes
12

Aka, you lost the debate before it even started. Thanks for another easy win. Next time make sure you use basic statistical terms correctly.

13

Case in point. 20 chars

14

blizz forum is more entertaining than every tv show ever :smiley:

2 Likes
15

He said that a sum of tournament wins is a bernoulli distribution:

That’s wrong. It’s a binomial distribution. This guy either failed his stats class or was copying/pasting wikipedia while having zero understanding of what he was copying. From the wiki page on binomial distributions:

In probability theory and statistics, the binomial distribution with parameters n and p is the discrete probability distribution of the number of successes in a sequence of n independent experiments, each asking a yes–no question, and each with its own boolean-valued outcome: success/yes/true/one (with probability p) or failure/no/false/zero (with probability q = 1 − p). A single success/failure experiment is also called a bernoulli trial; a single trial, i.e., *n* = 1, the binomial distribution is a bernoulli distribution.

In my post I took the sum of multiple coin flips. That’s a binomial distribution. He has no idea what he’s talking about. Most of the stuff he copied and pasted agrees with me, but he was convinced he beat my argument. He’s just clueless.

16

Excellent! Now you should be more careful about what you are reading

So its a sequence of independent experiments. Now go read about what does it mean for random variables (or just events in a probability space) to be independent. Then come back here and tell me if you think we should model “Zerg winning tournament A” and “Zerg winning tournament B” for say 2 consecutive tournaments A and B, as 2 independent events.

2 Likes
17

Either way you are still dead wrong. If the draws aren’t independent, then it’s a hypergeometric distribution, still not a bournilli distribution.

18

Wow, I give up. You are amazing man, when are are you running for president?

2 Likes
19

I commend you for making an effort, but you really had no chance at fooling anyone who has ever taken a class on statistics when you can’t use basic terminology correctly. There have been far more knowledgable trolls than you, and I still caught them.

20

Why which he means that there are more patient posters than you, and eventually he gave up trying to trick anybody and declared victory before running off shouting “WHOOSH!” pretending to fly.