|
I apologize to everyone in this thread for taking the OP seriously. My mod senses are definitely off today.
-- Chill |
On February 16 2011 05:47 hypercube wrote:Show nested quote +On February 16 2011 05:40 TBO wrote:
My bad should read have read the thread more careful. However my statement is still partly true, at least if he has not been the only person doing the test. But just the only person posting it because he got this extreme results. And the results isn't too extreme btw, cumulated probability for close spots being 72 when probability for close spot being 1/3 in a sample size of 100 is 0,1066.
Can you check your math? I got a z value of >8 (EV=33, SD=4,71), haven't looked up the actual probability but it must be really, really low. Someone posted 1:10^1 6 which sounds about right.
Really shouldn't do math while learning for an linguistics exam.... been totally wrong of course.
http://www.wolframalpha.com/input/?i=72 from 100 with probability 1/3
there you have it:
probability to get 72 or more is 3.22x10^-15.
|
On February 16 2011 05:53 TBO wrote:Show nested quote +On February 16 2011 05:47 hypercube wrote:On February 16 2011 05:40 TBO wrote:
My bad should read have read the thread more careful. However my statement is still partly true, at least if he has not been the only person doing the test. But just the only person posting it because he got this extreme results. And the results isn't too extreme btw, cumulated probability for close spots being 72 when probability for close spot being 1/3 in a sample size of 100 is 0,1066.
Can you check your math? I got a z value of >8 (EV=33, SD=4,71), haven't looked up the actual probability but it must be really, really low. Someone posted 1:10^1 6 which sounds about right. Really shouldn't do math while learning for an linguistics exam.... been totally wrong of course. http://www.wolframalpha.com/input/?i=72 from 100 with probability 1/3there you have it
Isn't that the probability of exactly 72? That doesn't seem right to me.
|
On February 16 2011 05:49 aristarchus wrote:Show nested quote +On February 16 2011 05:47 hypercube wrote:On February 16 2011 05:40 TBO wrote:
My bad should read have read the thread more careful. However my statement is still partly true, at least if he has not been the only person doing the test. But just the only person posting it because he got this extreme results. And the results isn't too extreme btw, cumulated probability for close spots being 72 when probability for close spot being 1/3 in a sample size of 100 is 0,1066.
Can you check your math? I got a z value of >8 (EV=33, SD=4,71), haven't looked up the actual probability but it must be really, really low. Someone posted 1:10^19 which sounds about right. The calculator at http://stattrek.com/Tables/Binomial.aspx gave me a p value of about 10^-15. (That's the probability of getting 72 or more close positions on 100 tries, and I believe it's done with precise binomial distribution calculations rather than approximations. Definitely the number should be some tiny decimal of that sort.
Yeah, I actually remembered the number incorrectly. Anyway, as far as I'm concerned there are only two possible conclusions:
A) OP made up his numbers
B) the spawn positions were not assigned with an equal and random probability
|
On February 16 2011 05:53 Zelniq wrote:Has nobody even questioned whether or not the OP is telling the truth? smells like bs/trolling. can someone check his match history for 100 games on metal recently? he claims he's HUARGH (http://sc2ranks.com/eu/717496/HUARGH) on EU, division Scion Uncle http://sc2ranks.com/div/67191/division-scion-unclehis most recent match history (http://eu.battle.net/sc2/en/profile/717496/1/HUARGH/matches) doesn't show any metalopolis games. sure seems he's full of it. also I really doubt someone actually hosted a game, waited 10 seconds for it to start, scouted close positions (sent OL to close-air), or instead left and loaded replay, rinse and repeat 100 times.
He could've done it via single player not multiplayer custom game.
|
Straight outta Johto18973 Posts
On February 16 2011 05:48 philcorp wrote:Show nested quote +On February 16 2011 05:43 motbob wrote:
The sample size is plenty big enough to show a statistically significant conclusion. I'll run the relevant t-test later. It really isnt. Go and ask a full lecture hall full of first year undergraduates to go and flip a coin 100 times. By the central limit theorem you expect a gaussian distribution about a mean of 50%. If your class is big enough you will get people who flipped a coin 80 or even 100 heads in a row. Funnily enough, if you actually carry out this experiment the students will fake the data and the distribution ends up being much more sharply peaked than it should.
Under a t-test, you allow for randomness due to sample size with your degrees of freedom. While 100 heads in a row is still possible, it is so unlikely that it is more feasible that the coin is simply biased or that there is something wrong with your flipping technique.
If you'd like to reduce all of that to simple RNG, then here's a useful applet for you to run.
http://www.math.uah.edu/stat/applets/BinomialCoinExperiment.xhtml
Just for completeness (referring back to my initial Binomial test), I did a simple t-test using the sample data. I considered again the comparison of close positions against close air and cross-map as an indicator function.
For H0: mean=50, my test statistic comes out to be 4.875 which is sufficient to reject the hypothesis at even α=0.0005. Also, because I am considering only a two scenario case, it is already allowing for inherent skew towards close positions.
Hope that helps.
|
On February 16 2011 05:52 hypercube wrote:Show nested quote +On February 16 2011 05:48 philcorp wrote:On February 16 2011 05:43 motbob wrote:
The sample size is plenty big enough to show a statistically significant conclusion. I'll run the relevant t-test later. It really isnt. Go and ask a full lecture hall full of first year undergraduates to go and flip a coin 100 times. By the central limit theorem you expect a gaussian distribution about a mean of 50%. If your class is big enough you will get people who flipped a coin 80 or even 100 heads in a row. Funnily enough, if you actually carry out this experiment the students will fake the data and the distribution ends up being much more sharply peaked than it should. FFS, just run any statistical test, instead of relying on your (flawed) intuition. If you took statistics in college or university you have the tools, so just use them.
All a test like the p-test will tell you is how far in the tail of the gaussian you are. Unlikely things happen, deal with it.
|
On February 16 2011 05:56 MoreFaSho wrote:Show nested quote +On February 16 2011 05:53 TBO wrote:On February 16 2011 05:47 hypercube wrote:On February 16 2011 05:40 TBO wrote:
My bad should read have read the thread more careful. However my statement is still partly true, at least if he has not been the only person doing the test. But just the only person posting it because he got this extreme results. And the results isn't too extreme btw, cumulated probability for close spots being 72 when probability for close spot being 1/3 in a sample size of 100 is 0,1066.
Can you check your math? I got a z value of >8 (EV=33, SD=4,71), haven't looked up the actual probability but it must be really, really low. Someone posted 1:10^1 6 which sounds about right. Really shouldn't do math while learning for an linguistics exam.... been totally wrong of course. http://www.wolframalpha.com/input/?i=72 from 100 with probability 1/3there you have it Isn't that the probability of exactly 72? That doesn't seem right to me.
a bit lower down there is the cumulated ones (72 or more successes)
|
On February 16 2011 05:56 Deadeight wrote:Show nested quote +On February 16 2011 05:53 Zelniq wrote:Has nobody even questioned whether or not the OP is telling the truth? smells like bs/trolling. can someone check his match history for 100 games on metal recently? he claims he's HUARGH (http://sc2ranks.com/eu/717496/HUARGH) on EU, division Scion Uncle http://sc2ranks.com/div/67191/division-scion-unclehis most recent match history (http://eu.battle.net/sc2/en/profile/717496/1/HUARGH/matches) doesn't show any metalopolis games. sure seems he's full of it. also I really doubt someone actually hosted a game, waited 10 seconds for it to start, scouted close positions (sent OL to close-air), or instead left and loaded replay, rinse and repeat 100 times. He could've done it via single player not multiplayer custom game.
Maybe in singleplayer you spawn differently than in multiplayer.
|
In any case, we should do another trial of 100 games and see if they can get the same result.
|
On February 16 2011 05:53 Zelniq wrote:Has nobody even questioned whether or not the OP is telling the truth? smells like bs/trolling. can someone check his match history for 100 games on metal recently? he claims he's HUARGH (http://sc2ranks.com/eu/717496/HUARGH) on EU, division Scion Uncle http://sc2ranks.com/div/67191/division-scion-unclehis most recent match history (http://eu.battle.net/sc2/en/profile/717496/1/HUARGH/matches) doesn't show any metalopolis games. sure seems he's full of it. also I really doubt someone actually hosted a game, waited 10 seconds for it to start, scouted close positions (sent OL to close-air), or instead left and loaded replay, rinse and repeat 100 times.
Was thinking the same thing myself, with numbers just a little too outrageous. There's no reason for blizzard to do this, and random number generators are good enough to pick 12 different spawn configurations with reasonable distribution. Happy trolling.
|
On February 16 2011 05:56 MoreFaSho wrote:Show nested quote +On February 16 2011 05:53 TBO wrote:On February 16 2011 05:47 hypercube wrote:On February 16 2011 05:40 TBO wrote:
My bad should read have read the thread more careful. However my statement is still partly true, at least if he has not been the only person doing the test. But just the only person posting it because he got this extreme results. And the results isn't too extreme btw, cumulated probability for close spots being 72 when probability for close spot being 1/3 in a sample size of 100 is 0,1066.
Can you check your math? I got a z value of >8 (EV=33, SD=4,71), haven't looked up the actual probability but it must be really, really low. Someone posted 1:10^1 6 which sounds about right. Really shouldn't do math while learning for an linguistics exam.... been totally wrong of course. http://www.wolframalpha.com/input/?i=72 from 100 with probability 1/3there you have it Isn't that the probability of exactly 72? That doesn't seem right to me.
The page gives you both, but interestingly, it is more likeley to get 72 of 100 than getting 73 or more of 100, this means that the z value must be enormous.
|
I don't know if it was said this has gone up to a lot of pages but using the Chi Square equation for statistical relevance.
Chi squared equals 67.888 with 2 degrees of freedom.
The two-tailed P value is less than 0.0001
By conventional criteria, this difference is considered to be extremely statistically significant
So whether you arbitraly aprove of it or not, it's statistically relevant.
|
On February 16 2011 05:57 philcorp wrote:
All a test like the p-test will tell you is how far in the tail of the gaussian you are. Unlikely things happen, deal with it.
Yes, but when they do it might be worth to examine the assumptions which lead you to believe they were unlikely in the first place.
|
On February 16 2011 05:56 MoreFaSho wrote:Show nested quote +On February 16 2011 05:53 TBO wrote:On February 16 2011 05:47 hypercube wrote:On February 16 2011 05:40 TBO wrote:
My bad should read have read the thread more careful. However my statement is still partly true, at least if he has not been the only person doing the test. But just the only person posting it because he got this extreme results. And the results isn't too extreme btw, cumulated probability for close spots being 72 when probability for close spot being 1/3 in a sample size of 100 is 0,1066.
Can you check your math? I got a z value of >8 (EV=33, SD=4,71), haven't looked up the actual probability but it must be really, really low. Someone posted 1:10^1 6 which sounds about right. Really shouldn't do math while learning for an linguistics exam.... been totally wrong of course. http://www.wolframalpha.com/input/?i=72 from 100 with probability 1/3there you have it Isn't that the probability of exactly 72? That doesn't seem right to me.
Yes. The probability of getting exactly 72 out of 100 with p=1/3 is 2.6*10^-15
the probability of getting 72 or more out of 100 with p=1/3 is 3.*10^-15
source: stats function on my voyage 200
|
On February 16 2011 05:59 Ihle wrote:
The page gives you both, but interestingly, it is more likeley to get 72 of 100 than getting 73 or more of 100, this means that the z value must be enormous.
You mean, it's more likely you get exactly 72 than 73 or more?
Surely, that can't be right 
|
DID NOBODY READ THE BIG RED BANNER AT THE TOP OF THIS THREAD?
Please....you guys are destroying this... if you don't have a contribution on how to solve the actual issue here then don't post.
|
United States7166 Posts
On February 16 2011 05:58 Ludwigvan wrote:Show nested quote +On February 16 2011 05:56 Deadeight wrote:On February 16 2011 05:53 Zelniq wrote:Has nobody even questioned whether or not the OP is telling the truth? smells like bs/trolling. can someone check his match history for 100 games on metal recently? he claims he's HUARGH (http://sc2ranks.com/eu/717496/HUARGH) on EU, division Scion Uncle http://sc2ranks.com/div/67191/division-scion-unclehis most recent match history (http://eu.battle.net/sc2/en/profile/717496/1/HUARGH/matches) doesn't show any metalopolis games. sure seems he's full of it. also I really doubt someone actually hosted a game, waited 10 seconds for it to start, scouted close positions (sent OL to close-air), or instead left and loaded replay, rinse and repeat 100 times. He could've done it via single player not multiplayer custom game. Maybe in singleplayer you spawn differently than in multiplayer.
it's true he could have done it in singleplayer and it wouldn't show in match history apparently, it also avoids the 10 second countdown timer so it makes sense. I'm still very doubtful that he took the time to load 100 games and load 100 replays (to see spawn location, or instead wait to scout his opponent's base). I think we should assume bullshit unless he provides replays, they should autosave
|
motbob
United States12546 Posts
So, people who actually know what they're talking about agree: simple statistical tests show that the OP's result is ridiculously improbable to have happened by chance, if we accept a 1/3 chance of getting close positions as the null hypothesis.
However, this doesn't rule out the possibility that the OP is full of it.
|
On February 16 2011 06:05 Zelniq wrote:Show nested quote +On February 16 2011 05:58 Ludwigvan wrote:On February 16 2011 05:56 Deadeight wrote:On February 16 2011 05:53 Zelniq wrote:Has nobody even questioned whether or not the OP is telling the truth? smells like bs/trolling. can someone check his match history for 100 games on metal recently? he claims he's HUARGH (http://sc2ranks.com/eu/717496/HUARGH) on EU, division Scion Uncle http://sc2ranks.com/div/67191/division-scion-unclehis most recent match history (http://eu.battle.net/sc2/en/profile/717496/1/HUARGH/matches) doesn't show any metalopolis games. sure seems he's full of it. also I really doubt someone actually hosted a game, waited 10 seconds for it to start, scouted close positions (sent OL to close-air), or instead left and loaded replay, rinse and repeat 100 times. He could've done it via single player not multiplayer custom game. Maybe in singleplayer you spawn differently than in multiplayer. it's true he could have done it in singleplayer and it wouldn't show in match history apparently, it also avoids the 10 second countdown timer so it makes sense. I'm still very doubtful that he took the time to load 100 games and load 100 replays (to see spawn location, or instead wait to scout his opponent's base). I think we should assume bullshit unless he provides replays, they should autosave
I also believe this is bullshit. OP should pay attention to the thread and gives feedback or replies...the fact that there isn't much of it makes it seem like he posted and is now waiting for the "chaos"...
go zelniq!
edit: and like motbob said...its clear if op is telling the truth this is enough to prove it T_____T
i personally have never had a feeling that close positions were this common and felt that it is around 1/3
|
On February 16 2011 06:05 Zelniq wrote:
it's true he could have done it in singleplayer and it wouldn't show in match history apparently, it also avoids the 10 second countdown timer so it makes sense. I'm still very doubtful that he took the time to load 100 games and load 100 replays (to see spawn location, or instead wait to scout his opponent's base). I think we should assume bullshit unless he provides replays, they should autosave
You can join as spectator with 2 bots and it's much faster. It doesn't really take that much time to go through 100, shouldn't take more than 30seconds per game with this method.
|
|
|
|
|
|
EPT
