|
On September 03 2010 09:07 Toxigen wrote:One thing I'd like to point out is that the distribution in diamond (or platinum, when that was the highest league in beta) hasn't been constant. Look here: http://www.teamliquid.net/forum/viewmessage.php?topic_id=117123This is a race distribution of early phase 1 in the beta, when Zerg was much more powerful than its current incarnation due to roaches still being 1 supply and having 2 base armor. Now, you'd have to make a couple assumptions to disprove the argument that Zerg are unpopular for reasons not related to effectiveness/underpoweredness: 1. The race distribution of all leagues in phase 1 platinum isn't very different from the distribution within phase 1's platinum leagues (i.e., no race was over- or under-represented in platinum at the time). 2. The race distribution of all leagues in the beta would have been similar to the distribution of the live game had the game been released as it was balance-wise on 3/26/2010 (when this data was gathered). It seems that the non-humanoid insect-like nature of the Zerg didn't bother 29.8% of the platinum players back then. Furthermore, Terran and Protoss are closer to each other and to Zerg (30.1% and 29.2%). If it were possible to find, it would be interesting to find population data from earlier in the game's development (i.e., back when Zerg was overrepresented and considered "OP" in Korea in phase 1). While it's impossible to say either way, I'd be leaning towards the notion that Zerg population was overall was proportionally higher back then -- and maybe not just on Asian servers. Furthermore, if one accepts the two assumptions above, one could also say that a race's ease-of-use or viability DOES affect how much of the population chooses to play that race, if you compare this distribution to today's diamond league. Food for thought.
this isn't a distribution.
|
On September 03 2010 05:50 starckr wrote:
Graph obviously shows something, but, as previously stated (sort of), there are so few players at the high end that the data is less helpful about balance. A single player tanking 100 points could throw the graph off by whole percentage points . . . .
Furthermore, the correlation between popularity and balance isn't necessarily perfect. For example, because there are undeniably less Zerg players all around (i.e., if you took silver stats, there would still be less Zerg than Toss/Terran), the stats would be a lot more troubling if there were more Zerg at the top than Terran or Toss.
You and everyone that has said this in this thread has no clue what they are talking about.
It shows a trend..and it's not even a vague trend. it's a crystal clear trend that shows that at the highest level of gameplay and really the ONLY place where racial balance comes into play Terrans are dominating.
The end. It's not a sample size, it's a direct representation, know the difference.
|
this graph is also skewed. should be # of players not percentages, percentages makes it seem weird...this graph is full of fail. author probably has not taken a stats class...can OP send me the raw data as in...
@yourwhiteshadow I don't think this is true. If you read my post (aka first post aka OP's post) you will notice that:
+ Show Spoiler +a.) I did not only include the percentage but the player count in each intervall
b.) a graphic which shows what you want to look at which is quite uninteresting due to population's difference
c.) the data is available and is linked in the first post, as also stated several times in this thread and mentioned by various people
d.) the weirdness occures because you want to look at somethign different if you want it to look different do it, everything you need is given to you
e.) the normalization could be taken in account if we figured out how we should do it, this is not clear, so you may think about it (option I. normalize by percentage in each intervall II. normalize over all 3 million sc2 gamers III. normalize as difference from other leagues IV. normalize aka fit all variations to a function and only look at this functions point-density)
f.) we dont have to see a peak in a gauss curve, as it is not needed to look at the middle for certain types of questions since I. we can fit a gauss curve even if we only look at the outliers since it is only basic math we have to use and II. the assumption that we will definetly not see a gauss curve is something possible but unmotivated
g.) it is true that we don't know that the AMM works, this is the reason that I mentioned in my first post, that "Some suggested to compare the racial distribution (RD) at high skill levels as indicator for imbalance or more likely as indicator for the lack of exploitable gamestyle." and therefore the difference on high skill level could be interesting.
h.) that I included the sub 1000pts player does not mean that I am looking at them, I presented a graph with as much information as possible (the reason I excluded the sub 600 pts diamonds is that I am over 600 points and know that I am not skillfull in all aspects of the game)
i.)nternet: BTW did you every published a scientific paper? I doubt it.
k.) If you want to make something better I welcome you, because this is what I want to happen in this thread: People should make assumptions, hypothesis, look at them again, discuss them and see how the patch changes things. This could lead to a better understanding not only about racial preferences but about the communities trends as a whole.
|
On September 03 2010 09:17 Jayme wrote:Show nested quote +On September 03 2010 05:50 starckr wrote:
Graph obviously shows something, but, as previously stated (sort of), there are so few players at the high end that the data is less helpful about balance. A single player tanking 100 points could throw the graph off by whole percentage points . . . .
Furthermore, the correlation between popularity and balance isn't necessarily perfect. For example, because there are undeniably less Zerg players all around (i.e., if you took silver stats, there would still be less Zerg than Toss/Terran), the stats would be a lot more troubling if there were more Zerg at the top than Terran or Toss. You and everyone that has said this in this thread has no clue what they are talking about. It shows a trend..and it's not even a vague trend. it's a crystal clear trend that shows that at the highest level of gameplay and really the ONLY place where racial balance comes into play Terrans are dominating. The end. It's not a sample size, it's a direct representation, know the difference.
Dude you are a friggin retard. Don't call others out when you clearly have no understanding of stats. I mean it's explained statistically on pretty much every page why it's not a trend. Do you really not get that as the population goes down, the error increases significantly?
So the most accurate representations are the earlier ones and they get less and less accurate as you go up. Please tell me you at least understand that much. Because from there it's easy.
Then imagine the point where they are. 35% T, 20% Z, 35% P we'll say at the lowest rating level. The error here can be anything, I'll use plus or minus 1% of what the "real" representation is, because our numbers are just a snapshot in time, not representative of the changes that occur all the time as people play.
Now the next tier up has about half the population of the lowest tier. This means that the error is almost double the error of the previous population, or in this case, around 2%. Then the next tier has half the population of that one. The error here is going to be around 4% and so on and so forth. So when you finally get to your population of 25 at the highest level, the error on what the 'real' numbers are is gigantic.
This is to say, practically, that at the top the people are moving around a lot so the volitility is very high.
The posts saying this includes all players is completely besides the point, that's not what we're comparing. All you can say with respect to that is "there are 11 terran players in the top 25." Saying that because it includes all the numbers it's accurate completely ignores the fact that this is a single snapshot in time. An equally invalid argument would be me saying that in the top 1 player, Zerg is represented 100%, Terran 0, and Protoss 0. Wouldn't that change the look of our "graph" we'll call it?
|
Tray, before you start throwing around fighting words, perhaps you should actually know what you're talking about.
You can't just claim that the results don't show a trend because "the numbers are small". That is not how statistics work. Statisticians use tests to determine significance and confidence. They can actually TELL you how confident you should be in a measurement.
Someone in this thread has already calculated significance for one part of the data. Their result was overwhelmingly in favor of there being a disproportionate amount of terrans in higher level diamond. The odds of it just being a fluke turned out to be one in a BILLION (more or less).
So either you are saying they lied or you are just ignoring the actual statistical reasoning in favor of your cargo cult wishy washy argument that relies more on jargon than mathematical reasoning.
|
Is it just me, or does anyone else think it's laughable that some people think terrans are overrepresented because that's the race you use in the campaign?
I mean, -maybe- in the bronze league there could be slightly more people playing terran, they're true SC beginners and played the campaign before trying out multiplayer. But top level diamond players? Do you REALLY think pro players choose their race because of the single player campaign and not because of which race is the most effective and earns more wins?
Plus there's the fact that we actually have the real numbers. Terran isn't even the most popular race, protoss is. LOL, I guess Artosis must have made a huge impact on people and made them want to choose protoss, right?
I think the fact that the campaign is mostly terran determined the race choice for maybe 1% of the players at best. When asked why there are so many terrans, "because of the campaign" is one of those simple answers that appears to make sense if you don't think about it... but it really isn't true at all.
For the vast majority of people, they don't consider the campaign at all. They try the races out, and stick with the one they like better. And what do people like? Ease of use, fun to use units, variety of units... and most importantly, how powerful the race feels.
Terran is all over those categories. Protoss can be player-friendly too... fewer units to control at once, more powerful individual units, and very strong when in large groups.
Zerg just feels like nothing special... no high damage units, no hard-to-counter units, and the "swarminess" from SC1 is gone. It's pretty easy to see why there are barely half as many zerg players as there are for other races.
|
On September 03 2010 09:27 ReplayArk wrote:Show nested quote +this graph is also skewed. should be # of players not percentages, percentages makes it seem weird...this graph is full of fail. author probably has not taken a stats class...can OP send me the raw data as in... @yourwhiteshadow I don't think this is true. If you read my post (aka first post aka OP's post) you will notice that: + Show Spoiler +a.) I did not only include the percentage but the player count in each intervall
b.) a graphic which shows what you want to look at which is quite uninteresting due to population's difference
c.) the data is available and is linked in the first post, as also stated several times in this thread and mentioned by various people
d.) the weirdness occures because you want to look at somethign different if you want it to look different do it, everything you need is given to you
e.) the normalization could be taken in account if we figured out how we should do it, this is not clear, so you may think about it (option I. normalize by percentage in each intervall II. normalize over all 3 million sc2 gamers III. normalize as difference from other leagues IV. normalize aka fit all variations to a function and only look at this functions point-density)
f.) we dont have to see a peak in a gauss curve, as it is not needed to look at the middle for certain types of questions since I. we can fit a gauss curve even if we only look at the outliers since it is only basic math we have to use and II. the assumption that we will definetly not see a gauss curve is something possible but unmotivated
g.) it is true that we don't know that the AMM works, this is the reason that I mentioned in my first post, that "Some suggested to compare the racial distribution (RD) at high skill levels as indicator for imbalance or more likely as indicator for the lack of exploitable gamestyle." and therefore the difference on high skill level could be interesting.
h.) that I included the sub 1000pts player does not mean that I am looking at them, I presented a graph with as much information as possible (the reason I excluded the sub 600 pts diamonds is that I am over 600 points and know that I am not skillfull in all aspects of the game)
i.)nternet: BTW did you every published a scientific paper? I doubt it.
k.) If you want to make something better I welcome you, because this is what I want to happen in this thread: People should make assumptions, hypothesis, look at them again, discuss them and see how the patch changes things. This could lead to a better understanding not only about racial preferences but about the communities trends as a whole.
/rant
so here is my issue, you can't just look at the tail end of a distribution curve for a few reasons:
1). you have to idea where the average lays
2). because of 1, you no longer know how many std deviation units you are from the avg
3). looking at 1/3 of the picture leaves out a ton of information
we expect to see a bell curve, which is what we would assume would be indicative of a "balanced" game. a balanced game would let 'nature' decide, and ultimately let skill decide which player is better, right? so then why not represent the data as a distribution function of race. most people would expect to see a bell curve if we assume that the game is fair and balanced. however, that is something we don't see, and so we jump to the conclusion, "IMBAAAAA!!!!!!". yet, if we look at the graph again, we clearly see that it isn't a good representation of the data. let's say hypothetically, there are 2k protoss and terran players each, and only 50 zerg players. the way that graph is set up, it would make you think that the game is "IMBA" because the zerg bars are ALWAYS going to be lower than the others...unless magically the top 20 players are zerg and they are the only ones above 1500 (even then, it'd seem weird).
to summarize, what i'm trying to say is that we expect to see a bell curve, but we do not. there is a reason for it. the data is simply NOT normalized. so we really SHOULDN'T be looking for a bell curve. population distribution...well we didn't do it properly...based on my math it looks like everything is well imba, and without getting too technical with errors, deviations, etc, it seems that the races are well represented if you take into account the number of players playing each race.
btw, i play toss, and as much as i hate terran and want to believe the game is imba towards terran, it looks like its not:
https://spreadsheets.google.com/ccc?key=0AhUL1GZVJIr2dHpWVmppNkhjZVNWcW9FQ0E4MkM4dWc&hl=en
|
Tray, ad hominem is generally frowned upon in rational discussion. Calling someone a "friggin retard" does little to support your point to a discerning audience, and tends to shut down the debate. Please stop. You are free to argue your points, just not shout people down and insult them.
To address your actual points: there is no reason to assume that the larger sample size is the accurate one when different portions of the population are under different conditions. For example, let's say we have a town with 10,000 slaves and 10 slaveowners. If we were to measure the average (mean) income in the town, we would get a number between 0 and the income of a slaveowner. Your method is suggesting that we measure the income of the slaves only, because they are the larger sample size. That would get us an incorrect mean (but a correct mode, which is not what we are looking for). In the situation with Terrans in Diamond League, we have a general trend of there being more Terrans the farther up we go. Even before Terran starts dominating the ladder, we still have that trend of more and more Terrans showing up. It is unreasonable to assume that a trend that appears even in the larger sections of the Diamond League must suddenly cut off as soon as we pass the 1200 point mark. You are examining statistics in isolation, which isn't very helpful at all. Here is the core argument that many, many people have repeated throughout this thread: if Terran is overpowered, then Terran players of less skill will rise through the rankings to their respective 50% winrate zones. However, players on the upper end of the spectrum will inevitably "max out" at some point and reach the upper limit of the ranking system. At that point, the Terran players will start to fill up the upper end of the ranking system. This will show itself as a disproportionate representation of Terrans at the upper end of Diamond League, with an increase in Terran players the farther up one goes. That is exactly what this graph shows us. About your "snapshot in time" theory, I'd be glad to examine further if you can produce some more data from earlier since release. However, as things stand, we have a single set of very strong data.
Finally, please don't use ad absurdum. I believe everyone knows that there is a significant difference between one person and twenty, and twenty people and 340. The fact remains that the top Zerg is 50 points ahead of the next closest opponent, a Terran, and has a much lower win/loss ratio than said Terran. It's not that important who ladders more.
|
It's really no surprise that Terran is so high like that. Everyone loves to follow flavors of the month. Sure Terran is pretty powerful right now, but I'm willing to bet that the bigger impact has been everyone complaining about Terran being OP everywhere for the past month or so. Most people on BNet I'm sure want to win, not improve at the game, and so of course they will pick the race that gives them the biggest advantage to win - even if they've only heard about the imbalance but haven't experienced it for themselves.
|
nice stats, didnt know there was so many terrans in top diamond
|
A picture is worth a thousand words. And a graph is worth 10,000. Props to the OP. Hope this gets around to Blizzard.
|
On September 03 2010 09:36 blacktoss wrote:
Tray, before you start throwing around fighting words, perhaps you should actually know what you're talking about.
You can't just claim that the results don't show a trend because "the numbers are small". That is not how statistics work. Statisticians use tests to determine significance and confidence. They can actually TELL you how confident you should be in a measurement.
Someone in this thread has already calculated significance for one part of the data. Their result was overwhelmingly in favor of there being a disproportionate amount of terrans in higher level diamond. The odds of it just being a fluke turned out to be one in a BILLION (more or less).
So either you are saying they lied or you are just ignoring the actual statistical reasoning in favor of your cargo cult wishy washy argument that relies more on jargon than mathematical reasoning.
Yes I can claim that. It's EXACTLY what I'm claiming and it's true. The person you're referring to did the worst math and completely ignored the fact that Terran isn't expected to be 33% representative. It should be relative to the total population of players that play terran if skill is evenly distributed between the three races, which yes, we can assume.
So his one in a billion was not even remotely accurate. In fact these numbers near the tail end are without any doubt are statistically insignificant. He didn't lie, he just didn't know what he was doing.
|
@Savio I think that Blizzard is doing much more in depth analysis of the game balance, they even have the individual player's win rate versus each race and the explicit build order. I think they are in fact running Chi-Square tests on their assumptions, and run algorithm over a set of ten thousand proxy rush games to see when the first Zealots are attacking stuff and if the five second delay will work, before they are explicitly paying tester to look if the balance change will do as they wish.
@yourwhiteshadow Thanks for doing some work, I would like to mention that your normalization is chipped because you watch the player with more than 600 points, why do you do this? There are much more player, if you include the sub 600 points diamonds you will get 64k people. So this is one thing I mentioned in my first respond to you. If you want to normalize - and we surely can do a little stuff with it - we can't simply cut the player off at one end. We should motivate why we do it.
I think you have a good point if you say that to only watch the outliners is bad, but if you got the values: + Show Spoiler +
4,1 8,9E-005
4,2 5,8E-005
4,3 3,8E-005
4,4 2,4E-005
4,5 1,6E-005
4,6 1,0E-005
What I further think is a nice idea you have is the hypothesis that the race's should be fitting a curve over various skill level (if we take league's and points as skill), but I think if we wait and look at the data after the patch we will be able to tell a lot more about some of our assumptions.
|
On September 03 2010 09:57 Acritter wrote:
Tray, ad hominem is generally frowned upon in rational discussion. Calling someone a "friggin retard" does little to support your point to a discerning audience, and tends to shut down the debate. Please stop. You are free to argue your points, just not shout people down and insult them.
To address your actual points: there is no reason to assume that the larger sample size is the accurate one when different portions of the population are under different conditions. For example, let's say we have a town with 10,000 slaves and 10 slaveowners. If we were to measure the average (mean) income in the town, we would get a number between 0 and the income of a slaveowner. Your method is suggesting that we measure the income of the slaves only, because they are the larger sample size. That would get us an incorrect mean (but a correct mode, which is not what we are looking for). In the situation with Terrans in Diamond League, we have a general trend of there being more Terrans the farther up we go. Even before Terran starts dominating the ladder, we still have that trend of more and more Terrans showing up. It is unreasonable to assume that a trend that appears even in the larger sections of the Diamond League must suddenly cut off as soon as we pass the 1200 point mark. You are examining statistics in isolation, which isn't very helpful at all. Here is the core argument that many, many people have repeated throughout this thread: if Terran is overpowered, then Terran players of less skill will rise through the rankings to their respective 50% winrate zones. However, players on the upper end of the spectrum will inevitably "max out" at some point and reach the upper limit of the ranking system. At that point, the Terran players will start to fill up the upper end of the ranking system. This will show itself as a disproportionate representation of Terrans at the upper end of Diamond League, with an increase in Terran players the farther up one goes. That is exactly what this graph shows us. About your "snapshot in time" theory, I'd be glad to examine further if you can produce some more data from earlier since release. However, as things stand, we have a single set of very strong data.
Finally, please don't use ad absurdum. I believe everyone knows that there is a significant difference between one person and twenty, and twenty people and 340. The fact remains that the top Zerg is 50 points ahead of the next closest opponent, a Terran, and has a much lower win/loss ratio than said Terran. It's not that important who ladders more.
Your example is not even remotely close to what is happening here. I explained it correctly. No need to create another analogy. It's explained on every page in this thread. If you're still having a hard time understanding it, google "confidence interval" or "statistical analysis and normal distribution.
The problem people are having is that Race representation should be considered independently and measured against how it changes as the rating goes up. The problem is as the rating goes up, the sample goes down, and error increases exponentially.
Anyway your analogy is completely off. I'm not doing anything remotely close to what you're talking about and my theory about the snapshot is not a theory, it's a statistical fact. Take a look at the spreadsheet above your post there idiot.
|
The problem is as the rating goes up, the sample goes down, and error increases exponentially. @Tray the first and last time I will talk to you, since I like to talk to mannered people and if various claims are brought to you that you might be offensive don't take it as given, but think about it somedays. If you say the error increases exponentially I would like to see the formular you use. What is your method to get the error? If you want you can PM me the latex term and everything I need as reference to understand your point, but still now I didn't get the thing you want to say with this post.
|
On September 03 2010 10:17 Tray wrote:
Yes I can claim that. It's EXACTLY what I'm claiming and it's true. The person you're referring to did the worst math and completely ignored the fact that Terran isn't expected to be 33% representative. It should be relative to the total population of players that play terran if skill is evenly distributed between the three races, which yes, we can assume.
So his one in a billion was not even remotely accurate. In fact these numbers near the tail end are without any doubt are statistically insignificant. He didn't lie, he just didn't know what he was doing.
You are right, he should have used ~40% rather than 33%. But the rest of his analysis was correct. The odds of this distribution being caused by chance are way past explainable by random chance. He provided numbers, while you have merely asserted that the sample numbers are too small. "In fact these numbers near the tail end are without any doubt are statistically insignificant". You have no numbers to support this assertion.
|
On September 03 2010 10:24 ReplayArk wrote:Show nested quote +The problem is as the rating goes up, the sample goes down, and error increases exponentially. @Tray the first and last time I will talk to you, since I like to talk to mannered people and if various claims are brought to you that you might be offensive don't take it as given, but think about it somedays. If you say the error increases exponentially I would like to see the formular you use. What is your method to get the error? If you want you can PM me the latex term and everything I need as reference to understand your point, but still now I didn't get the thing you want to say with this post.
Maybe you should just look it up for yourself and get smarter instead of being like everyone else in here and demanding I teach you all how statistics work.
This is 5 seconds of google. http://en.wikipedia.org/wiki/Sample_size
A typical example would be when a statistician wishes to estimate the arithmetic mean of a quantitative random variable (for example, the height of a person). Assuming that they have a random sample with independent observations, and also that the variability of the population (as measured by the standard deviation σ) is known, then the standard error of the sample mean is given by the formula:
σ/sq.root(n)
As N becomes smaller and smaller, where N is the population of the sample size, the error increases.
if n = 25, error = std/5 = 20% (std = 1 for simplicity, doesn't matter what you use, it's all relative)
if n = 50, error = std/7.07 = 14%
if n = 100, error = std/10 = 10%
if n = 1000, error = std/31.6 = 3%
And I'm going to go on because some of you guys really need the education.
So this means that we can with a 95% certainly that the real number of zerg representation at n= 25 is about 20% (or whatever it is) plus or minus 20%. This means that we're 95% sure the 'real' value is between 0-40%, but 5% of the time it will be outside that range.
This is the correct way to analyze this data including sampling error, which in our case takes into account the variability of player ratings.
Also this is why the normalized numbers are all fairly close relative to total population that plays that race at the n = big numbers, but why it's highly skewed at n = small numbers.
|
On September 03 2010 07:54 cursor wrote:
I'm not sure if its 100% the "strength of the units" type of imbalance. I may favor the argument that fundamentally, Zerg is just far more difficult to play. The scouting, the macro, the fragility of the units, the need to creep- the need to have More workers and More units than a Terran with Mule- it's just harder mechanically.
Don't forget the Terran's dominant versatility in terms of weapon ranges and splash damage compared to the other races. 
|
Yes I can claim that. It's EXACTLY what I'm claiming and it's true. The person you're referring to did the worst math and completely ignored the fact that Terran isn't expected to be 33% representative. It should be relative to the total population of players that play terran if skill is evenly distributed between the three races, which yes, we can assume.
So his one in a billion was not even remotely accurate. In fact these numbers near the tail end are without any doubt are statistically insignificant. He didn't lie, he just didn't know what he was doing.
Here's the kicker genius: You're fighting a straw man. The claim made was that Terran make up a disproportionate number of high level diamond players such that the deviation from an even distribution is not due to random chance.
The claim is CORRECT. The math is CORRECT. You are the one coming in here and putting words in peoples' mouths and then saying "haha no u r rong I will now insult you".
I don't think you know how statistics works at all. The hypothesis tested was the hypothesis that Terran should make up 33% of the racial distribution in high diamond league (I am not sure if he took into account Random). The statistical test he used showed with high confidence that this hypothesis was false. End of story that is all the reasoning used.
You talk about statistical insignificance but your criticism does not address the point's validity, it attacks it on the basis that "it is not the right hypothesis". Sorry, but that does not say anything about the conclusion drawn.
Maybe you should just look it up for yourself and get smarter instead of being like everyone else in here and demanding I teach you all how statistics work.
This is 5 seconds of google.
A typical example would be when a statistician wishes to estimate the arithmetic mean of a quantitative random variable (for example, the height of a person). Assuming that they have a random sample with independent observations, and also that the variability of the population (as measured by the standard deviation σ) is known, then the standard error of the sample mean is given by the formula:
σ/sq.root(n)
As N becomes smaller and smaller, where N is the population of the sample size, the error increases.
if n = 25, error = std/5 (std = 1) = 20%
if n = 50, error = std/7.07 = 14%
if n = 100, error = std/10 = 10%
if n = 1000, error = std/31.6 = 3%
Once again, you are full of shit. You come in here saying you are better than anyone else, throw around a few big words, and then cite google and say "hurdy hur".
It doesn't matter if the variance is high, because when YOU DO THE ACTUAL TEST (instead of bullshitting with jargon), you find that p is very high. So high in fact that you MUST discard the hypothesis that "Terran is not overrepresented in high diamond". You can say "WELL SIGMA IS HIGH" but it doesn't matter. Statistical tests are robust. They take sigma into account.
I am beginning to feel like you don't know anything about statistics at all.
|
On September 03 2010 10:40 blacktoss wrote:Show nested quote +Yes I can claim that. It's EXACTLY what I'm claiming and it's true. The person you're referring to did the worst math and completely ignored the fact that Terran isn't expected to be 33% representative. It should be relative to the total population of players that play terran if skill is evenly distributed between the three races, which yes, we can assume.
So his one in a billion was not even remotely accurate. In fact these numbers near the tail end are without any doubt are statistically insignificant. He didn't lie, he just didn't know what he was doing. Here's the kicker genius: You're fighting a straw man. The claim made was that Terran make up a disproportionate number of high level diamond players such that the deviation from an even distribution is not due to random chance. The claim is CORRECT. The math is CORRECT. You are the one coming in here and putting words in peoples' mouths and then saying "haha no u r rong I will now insult you". I don't think you know how statistics works at all. The hypothesis tested was the hypothesis that Terran should make up 33% of the racial distribution in high diamond league (I am not sure if he took into account Random). The statistical test he used showed with high confidence that this hypothesis was false. End of story that is all the reasoning used. You talk about statistical insignificance but your criticism does not address the point's validity, it attacks it on the basis that "it is not the right hypothesis". Sorry, but that does not say anything about the conclusion drawn. Show nested quote +Maybe you should just look it up for yourself and get smarter instead of being like everyone else in here and demanding I teach you all how statistics work.
This is 5 seconds of google.
A typical example would be when a statistician wishes to estimate the arithmetic mean of a quantitative random variable (for example, the height of a person). Assuming that they have a random sample with independent observations, and also that the variability of the population (as measured by the standard deviation σ) is known, then the standard error of the sample mean is given by the formula:
σ/sq.root(n)
As N becomes smaller and smaller, where N is the population of the sample size, the error increases.
if n = 25, error = std/5 (std = 1) = 20%
if n = 50, error = std/7.07 = 14%
if n = 100, error = std/10 = 10%
if n = 1000, error = std/31.6 = 3% Once again, you are full of shit. You come in here saying you are better than anyone else, throw around a few big words, and then cite google and say "hurdy hur". It doesn't matter if the variance is high, because when YOU DO THE ACTUAL TEST (instead of bullshitting with jargon), you find that p is very high. So high in fact that you MUST discard the hypothesis that "Terran is not overrepresented in high diamond". You can say "WELL SIGMA IS HIGH" but it doesn't matter. Statistical tests are robust. They take sigma into account. I am beginning to feel like you don't know anything about statistics at all.
Dude you are incredibly stupid. My analysis is correct. Assuming Terran should make up 33% of the distrubtion at the top is NOT correct. It should be EQUAL to the total % of players that play terran. Period. It's not debatable.
As the sample drops, the error in the 'real' value increases and becomes less accurate. Period.
Stay in school.
|
|
|
|
|
|
EPT
