Jan 29, 2018
-2

01/28/2018 10:21 PMPosted by Werner01/28/2018 10:11 PMPosted by BourneKilla8 Zerg out of 16, ZVZ finals FIVE ZERG foreigners qualified.

Yet in the top 4 at WCS in Leipzig over the weekend, we had 2 Protoss', a Zerg and a Terran. That to me shows excellent representation across the board of all races. I would have said something were wrong if there were like 4 Protoss' or maybe 3 Protoss' and a Zerg or something to that effect.

This is no basis for balance discussion - essentially you are saying the game is balanced on the basis of two games (if two quarters had gone Snute's and Elazer's way...as in if they'd both just won one more game each, we'd have 3 Zergs and 1 Protoss in the semis.)

01/28/2018 11:22 PMPosted by CazdogThis is no basis for balance discussion - essentially you are saying the game is balanced on the basis of two games (if two quarters had gone Snute's and Elazer's way...as in if they'd both just won one more game each, we'd have 3 Zergs and 1 Protoss in the semis.)

Except it did not and the most skilled players ended up in the top 4 - so the question is was it skill or a particularly OP race? I have a feeling that is they don't mess around with the balance of the game too much in the coming months, we will see this trend continue.

Also look at the GSL so far - what do you see? We have 4 Terrans, 4 Zerg and 6 Protoss in round of 16 so far, with one more group to go. It seems very even to me. Not one race is completely dominant. Player skill is the determinant in the matches.

So in essence, I don't really see how my point is not valid, but your opinion is your opinion and it is noted and respected. Let's leave it at that.

01/28/2018 10:19 PMPosted by WernerGood thing there were 84 games.

Good reply - those were my thoughts exactly...01/28/2018 08:44 PMPosted by BourneKillaIs Zerg so easy that they can make 40% of the GM even when Terran is OP?

No but hang on now, what is your reply to the fact that there were 84 games instead of your earlier quote of a sample size of 35? Let us stick to the point of interest here - don't change the subject.

He has no response for it. He just copied and pasted a number he found online for the sample size requirements without realizing the sample size was plenty large according to those requirements. Now he's spouting a bunch of random stats, that he has Cherry Picked because they seem in favor of his argument, to try and the detract from the fact that he done screwed up. He has no clue how statistics work.

Besides, the sample sizes used in studies are rather inapplicable in this scenario. Those sample sizes are picked for certain reasons and those reasons are irrelevant. It basically comes down to being able to achieve a certain level of confidence in your findings. ~30ish is usually what's needed to give a good confidence in your findings, e.g. 90%. It's not applicable here because you don't need 30 samples to achieve that same level of confidence.

For example, there was only a 44.54% chance that Innovation would beat creator 2-0. The odds of alive beating rogue 2-0 was 17.70%. If we take all of these into consideration and the net is less than a 10% chance of occurring, then we can say with a 90% confidence that terrans over performed.

01/28/2018 05:17 PMPosted by NerfProtossYes. YES exactly, EXACTLY!01/28/2018 05:13 PMPosted by NinjetteToo small sample size - stats are totally irrellevant for anything. There's nothing to see here, people. Please move on!

The sample size is like way too small and stuff.

But whenever a terran whiner gives a winrate that "proves" terran is underpowered, no matter how small the sample size is, you still defend it... hypocrite?

Jan 29, 2018
-2

01/28/2018 10:19 PMPosted by Werner...

Good reply - those were my thoughts exactly...

...

No but hang on now, what is your reply to the fact that there were 84 games instead of your earlier quote of a sample size of 35? Let us stick to the point of interest here - don't change the subject.

He has no response for it. He just copied and pasted a number he found online for the sample size requirements without realizing the sample size was plenty large according to those requirements. Now he's spouting a bunch of random stats, that he has Cherry Picked because they seem in favor of his argument, to try and the detract from the fact that he done screwed up. He has no clue how statistics work.

Besides, the sample sizes used in studies are rather inapplicable in this scenario. Those sample sizes are picked for certain reasons and those reasons are irrelevant. It basically comes down to being able to achieve a certain level of confidence in your findings. ~30ish is usually what's needed to give a good confidence in your findings, e.g. 90%. It's not applicable here because you don't need 30 samples to achieve that same level of confidence.

For example, there was only a 44.54% chance that Innovation would beat creator 2-0. The odds of alive beating rogue 2-0 was 17.70%. If we take all of these into consideration and the net is less than a 10% chance of occurring, then we can say with a 90% confidence that terrans over performed.

Generally the population size has to be more than 35, not the incident size.

I JUST brought up where Zergs DOMINATED in a similar fashion. OH! except this was before the Raven got deleted.

The Terrans are doing well in GSL. There are other large tournaments going RIGHT NOW. Terran is not doing so hot. What's your excuse for cherry picking data?

01/28/2018 11:57 PMPosted by WernerAlso look at the GSL so far - what do you see? We have 4 Terrans, 4 Zerg and 6 Protoss in round of 16 so far, with one more group to go. It seems very even to me.

How many players there are in a round has no relevance in a vacuum. It's what the odds are of that occurring that has relevance. To know that, you have to know what the previous round was and whether or not more players of a certain race advanced more frequently than they should have.

The number of XYZ players in a tournament has no meaning. How those players perform does have meaning.

Imagine a theoretical tournament that was 60% terran. Now imagine if in the next round, the number of terrans dropped to 5%. If the game were fair, the next round should have had roughly 60% terran but instead we got 5%. That means terrans dramatically underperformed in that round. Likewise if the round was 5% Terran and in the following round it was 25% it would mean terran was overperforming.

Think of it like a bag of red, green and blue marbles. A red marble has a 33% chance of being drawn and the same for green and blue. But let's suppose that there are more red marbles in the bag. That means the collective odds of selecting a red marble will be greater because their numbers are greater. The odds of an individual red marble being selected is still the same, 33%, but red marbles in general are more likely to get picked because there are more of them. The same can be said if there are fewer marbles of a certain color.

You can Define what the odds are of picking a particular color, and if the selections that you draw from the bag deviate from the odds you calculated, and the sample size is large enough, you can say that, with confidence, there is another variable in the equation other than A) the odds of an individual marble of a particular color being selected, and B) the number of marbles of that color.

Suppose that we have 100 red marbles in a bag of 1000. That means the odds of selecting a red marble should be roughly 10%. Let's suppose we reach into the bag and pull out 20 marbles, and that 10 of them are red. That means the odds of selecting a red marble was 10 in 20 or 50%. 50% is much larger than 10% we're just want the odds should be for selecting a red marble. We can therefore say that there is something else going on such as all the red marbles being on top of the marble pile so when you reach in your much more likely to select a red marble. How is it fair that the red marbles only make up 10% of the bag, but have a 50% odds of being drawn?

Now this scenario is analogous to a tournament round. If you reach into the tournament bag of marbles and pull out more terrans then should be expected it means Terran is over performing. The longer this trend continues, aka the larger the sample size, the more confident you can be in this result. When that result reaches greater than 90%, you can say Terran is overpowered and would only have a 10% chance of being wrong.

01/29/2018 08:10 AMPosted by tEhbAtZ01/28/2018 11:57 PMPosted by WernerAlso look at the GSL so far - what do you see? We have 4 Terrans, 4 Zerg and 6 Protoss in round of 16 so far, with one more group to go. It seems very even to me.

How many players there are in a round has no relevance in a vacuum. It's what the odds are of that occurring that has relevance. To know that, you have to know what the previous round was and whether or not more players of a certain race advanced more frequently than they should have.

The number of XYZ players in a tournament has no meaning. How those players perform does have meaning.

Imagine a theoretical tournament that was 60% terran. Now imagine if in the next round, the number of terrans dropped to 5%. If the game were fair, the next round should have had roughly 60% terran but instead we got 5%. That means terrans dramatically underperformed in that round. Likewise if the round was 5% Terran and in the following round it was 25% it would mean terran was overperforming.

Think of it like a bag of red, green and blue marbles. A red marble has a 33% chance of being drawn and the same for green and blue. But let's suppose that there are more red marbles in the bag. That means the collective odds of selecting a red marble will be greater because their numbers are greater. The odds of an individual red marble being selected is still the same, 33%, but red marbles in general are more likely to get picked because there are more of them. The same can be said if there are fewer marbles of a certain color.

You can Define what the odds are of picking a particular color, and if the selections that you draw from the bag deviate from the odds you calculated, and the sample size is large enough, you can say that, with confidence, there is another variable in the equation other than A) the odds of an individual marble of a particular color being selected, and B) the number of marbles of that color.

Now this scenario is analogous to a tournament round. If you reach into the tournament bag of marbles and pull out more terrans then should be expected it means Terran is over performing.

Terran IS overperforming. NOT surprising, when you account for people like Ragnarok qualifying.

I'm gonna ask you why YOU think think so many Zerg foreigners qualified in the first place? But you keep cheery picking data.

THE minimum you would need to be statistically relevant, would be 35 TVZs.

Answer the question batz: Is Terran OP? If so does that mean that Zerg is just so amazingly ez that it can DOMINATE GM while Terran is the REALLY overpowered race?

01/29/2018 08:25 AMPosted by BourneKillaI'm gonna ask you why YOU think think so many Zerg foreigners qualified in the first place? But you keep cheery picking data.

It's impossible to answer that question and it's entirely irrelevant. You could have a thousand terrans sign up for a tournament or only one Terran and it wouldn't mean anything. In order to Define an observation as unusual you have to have an expected value to compare it against. If a tournament is 10% Terran, then Terran should have roughly a 10% chance of advancing to the next round which means 10% of the terrans should advance to the next round. If the observation deviates from this then we can say something is up, but only because we can Define what is expected.

Who signs up for a tournament has no meaning because you can't Define how many people you expect to show up for each race. You could have more terrans or fewer terrans or no terrans or a thousand zergs but it has no meaning because you can't define an expected value to compare your observation to. If you can't define an expected value then you cannot prove that the observation was unusual. You also can't prove that it was normal, either. The only thing you can prove is that there is not enough information to figure out what is going on.

01/29/2018 08:31 AMPosted by tEhbAtZ01/29/2018 08:25 AMPosted by BourneKillaI'm gonna ask you why YOU think think so many Zerg foreigners qualified in the first place? But you keep cheery picking data.

It's impossible to answer that question and it's entirely irrelevant. You could have a thousand terrans sign up for a tournament or only one Terran and it wouldn't mean anything. In order to Define an observation as unusual you have to have an expected value to compare it against. If a tournament is 10% Terran, then Terran should have roughly a 10% chance of advancing to the next round which means 10% of the terrans should advance to the next round. If the observation deviates from this then we can say something is up, but only because we can Define what is expected.

Who signs up for a tournament has no meaning because you can't Define how many people you expect to show up for each race. You could have more terrans or fewer terrans or no terrans or a thousand zergs but it has no meaning because you can't define an expected value to compare your observation to. If you can't define an expected value then you cannot prove that the observation was unusual.

So, to reiterate:

1. the population who QUALIFIES (not signs up--I can't just drive to Seoul and tell them I wanna play in the GSL. I saw it on youtube and I think I have what it takes to beat the best people on earth) DOES NOT MATTER

2. the WCS WORLD CHAMPIONSHIP DOES NOT MATTER

3. any other tournaments currently ongoing DO NOT MATTER

4. holding over 40% of the Grandmaster League, while Terran has only 25% DOES NOT MATTER

5. Zerg leading in winrate across the ENTIRE LADDER DOES NOT MATTER!

Terran doing well against Zerg, in a population so small that Noregret being sick and barely taking a game off ANYONE would PLUMMET their win rate. THIS DOES MATTER!

01/29/2018 08:38 AMPosted by BourneKillaSo, to reiterate:

1. the population who QUALIFIES (not signs up--I can't just drive to Seoul and tell them I wanna play in the GSL. I saw it on youtube and I think I have what it takes to beat the best people on earth) DOES NOT MATTER

2. the WCS WORLD CHAMPIONSHIP DOES NOT MATTER

3. any other tournaments currently ongoing DO NOT MATTER

4. holding over 40% of the Grandmaster League, while Terran has only 25% DOES NOT MATTER

5. Zerg leading in winrate across the ENTIRE LADDER DOES NOT MATTER!

Terran doing well against Zerg, in a population so small that Noregret being sick and barely taking a game off ANYONE would PLUMMET their win rate. THIS DOES MATTER!

Stop putting words in my mouth. If you want to see what I said then read my posts. You clearly didn't. Either that or you didn't understand them either way you need to reread them.

01/28/2018 04:53 PMPosted by AmorphousQualified for Code S:9Terran /11Protoss /12Zerg

Terran vs Zerg70.6% vs 29.4%

Zerg vs Protoss54.8% vs 45.2%

Protoss vs Terran57.9% vs 42.1%

Qualified for top 16:4Terran /4Zerg /6Protoss

----------------------

Predicting Her0(P) and Maru(T) to make it out from group H.

I was hoping that no balance changes would take place before this season ended. Would have been interesting to see Terran win it all despite the outcries.

01/28/2018 05:27 PMPosted by tEhbAtZ01/28/2018 05:21 PMPosted by DrFuzzyGloveIt is a small sample. Even innovation by himself skews this. He is one of the most consistent and solid players of the game. If he played z or p it would skew it that way.

Despite this indicating the game is actually alright, it has to be taken lightly due to sample size.

Lots of people playing the sample size card, claiming it's too small. How many of you have done a confidence test to prove if it is too small or not, or did you just assume it's too small because it fits your presumptions?

01/28/2018 05:50 PMPosted by BourneKilla...

35 is generally the minimal sample size to to be considered statistically relevant. That's why most studies are 42-45, in case some people drop out

Good thing there were 84 games.

So, what do I have wrong here.

If I'm putting words in your mouth, then give me a summary of your conclusions.

Sarcasm and tonality in text can throw everything off. By all means, seriously, give me what your final conclusion is based on the data.

01/29/2018 09:02 AMPosted by BourneKillaIf I'm putting words in your mouth, then give me a summary of your conclusions

It is not my job to prove a negative. Anyone can read your post and see you are pulling things out of thin air, such as your accusation that I said WCS finals don't matter. I didn't say that and I encourage you to prove otherwise. It is your job to prove it since you are the one claiming it. Or you could just stop putting words in my mouth. What are you, 10 years old?

01/29/2018 09:11 AMPosted by tEhbAtZ01/29/2018 09:02 AMPosted by BourneKillaIf I'm putting words in your mouth, then give me a summary of your conclusions

It is not my job to prove a negative. Anyone can read your post and see you are pulling things out of thin air, such as your accusation that I said WCS finals don't matter. I didn't say that and I encourage you to prove otherwise. It is your job to prove it since you are the one claiming it. Or you could just stop putting words in my mouth. What are you, 10 years old?

I gave you a chance to prove what you're ACTUALLY saying. You decided to insult me. Because I did summarize what you were saying?

You have a chance to tell everyone what conclusions you ARE drawing from the data: What conclusion is it (you know, a positive)?

You don't have to prove what I'm saying about you isn't true. You just have to prove what conclusions you are coming up with. What are they? Because I've made all these points before in online debate. You looked at the ladder stats and you called them "a fart in the wind" (so, too small a sample size). You looked at these players and you certainly seem to think they have statistical relevance.

What conclusions did you draw? Or are you full of !@#$ and I already explained your position perfectly?

01/29/2018 08:10 AMPosted by tEhbAtZ01/28/2018 11:57 PMPosted by WernerAlso look at the GSL so far - what do you see? We have 4 Terrans, 4 Zerg and 6 Protoss in round of 16 so far, with one more group to go. It seems very even to me.

How many players there are in a round has no relevance in a vacuum. It's what the odds are of that occurring that has relevance. To know that, you have to know what the previous round was and whether or not more players of a certain race advanced more frequently than they should have.

The number of XYZ players in a tournament has no meaning. How those players perform does have meaning.

Imagine a theoretical tournament that was 60% terran. Now imagine if in the next round, the number of terrans dropped to 5%. If the game were fair, the next round should have had roughly 60% terran but instead we got 5%. That means terrans dramatically underperformed in that round. Likewise if the round was 5% Terran and in the following round it was 25% it would mean terran was overperforming.

Think of it like a bag of red, green and blue marbles. A red marble has a 33% chance of being drawn and the same for green and blue. But let's suppose that there are more red marbles in the bag. That means the collective odds of selecting a red marble will be greater because their numbers are greater. The odds of an individual red marble being selected is still the same, 33%, but red marbles in general are more likely to get picked because there are more of them. The same can be said if there are fewer marbles of a certain color.

You can Define what the odds are of picking a particular color, and if the selections that you draw from the bag deviate from the odds you calculated, and the sample size is large enough, you can say that, with confidence, there is another variable in the equation other than A) the odds of an individual marble of a particular color being selected, and B) the number of marbles of that color.

Suppose that we have 100 red marbles in a bag of 1000. That means the odds of selecting a red marble should be roughly 10%. Let's suppose we reach into the bag and pull out 20 marbles, and that 10 of them are red. That means the odds of selecting a red marble was 10 in 20 or 50%. 50% is much larger than 10% we're just want the odds should be for selecting a red marble. We can therefore say that there is something else going on such as all the red marbles being on top of the marble pile so when you reach in your much more likely to select a red marble. How is it fair that the red marbles only make up 10% of the bag, but have a 50% odds of being drawn?

Now this scenario is analogous to a tournament round. If you reach into the tournament bag of marbles and pull out more terrans then should be expected it means Terran is over performing. The longer this trend continues, aka the larger the sample size, the more confident you can be in this result. When that result reaches greater than 90%, you can say Terran is overpowered and would only have a 10% chance of being wrong.

Nice misuse of law of large number here. Unfortunately it only works when you do have large numbers.

01/29/2018 09:39 AMPosted by DogOP wants to talk about stats and let's do some real stats: Do a two tailed t-test with %in round of 16 in terran vs non-terran, got p-vaue of 0.96. It is like the very definition of statistically insignificant.

A two tailed test is inapplicable single tailed would work fine and you have to take into account the skill of the players when defining your expected value. Innovation for example will have like a 75% chance to beat leenock.

01/29/2018 09:44 AMPosted by DogNice misuse of law of large number here. Unfortunately it only works when you do have large numbers.

It's not being misused. Feel free to calculate just how many trials you need to have a 90% confidence with these results. If the number of trials needed is larger than observed, then your point would be true.

ITT: People applying statistics that only work with random samples to a non random sample.

Jan 29, 2018
-1

01/29/2018 09:48 AMPosted by tEhbAtZ01/29/2018 09:39 AMPosted by DogOP wants to talk about stats and let's do some real stats: Do a two tailed t-test with %in round of 16 in terran vs non-terran, got p-vaue of 0.96. It is like the very definition of statistically insignificant.

A two tailed test is inapplicable single tailed would work fine and you have to take into account the skill of the players when defining your expected value. Innovation for example will have like a 75% chance to beat leenock.01/29/2018 09:44 AMPosted by DogNice misuse of law of large number here. Unfortunately it only works when you do have large numbers.

It's not being misused. Feel free to calculate just how many trials you need to have a 90% confidence with these results. If the number of trials needed is larger than observed, then your point would be true.

Interesting how you ignore questions that upset you.

Seriously, what changed for TVZ since the WCS Global finals?

I bet it was the Raven. Let's get the old raven back so Zerg can have a chance!

Jan 29, 2018
-2

01/28/2018 05:17 PMPosted by NerfProtossYes. YES exactly, EXACTLY!01/28/2018 05:13 PMPosted by NinjetteToo small sample size - stats are totally irrellevant for anything. There's nothing to see here, people. Please move on!

The sample size is like way too small and stuff.

01/18/2018 08:35 AMPosted by NerfProtosshttp://aligulac.com/periods/latest/

PvT = 70.00%

Leading race Protoss (3%)

Lagging race Terran (5%)

inb4 'muh sample size'

Mirrors : 21 PvP,4 TvT, 23 ZvZ

nerfing stalkers is irresponsible, we need a forum vote!

inb4 'muh sample size'

The hypocrisy is strong with our Terran friends, isn't it?