EPTI'm going to start looking through my replays and I'll post here my results.
Metalopolis prone to close positions, why? - Page 4
| Forum Index > SC2 General |
I apologize to everyone in this thread for taking the OP seriously. My mod senses are definitely off today. -- Chill | ||
|
Deadeight
United Kingdom1629 Posts
I'm going to start looking through my replays and I'll post here my results. | ||
|
MonsieurGrimm
Canada2441 Posts
On February 16 2011 05:20 Chill wrote: Interesting. Instead of arguing about the reasonable sample size, let's either: a) Figure out how we're going to test it with a larger sample size; or, prefereably b) Figure out why Metalopolis is prone to spawning clone positions. We could make a new thread about it, and have people put their names in and have everyone do a sample of 10-20 then report back to the thread so nobody has to waste an hour on it... | ||
Chill
Calgary25986 Posts
| ||
|
MorroW
Sweden3522 Posts
they should make it like shakuras  | ||
|
Wolf
Korea (South)3290 Posts
| ||
|
philcorp
Canada32 Posts
| ||
|
eLiE
Canada1039 Posts
On February 16 2011 05:16 Liquid`Tyler wrote: You are correct to say that it's wrong to utter the words proof and conclusion at this point, but the 100 game sample size shows that it's more likely than not (>50% chance) that close positions happen more than 1/3 of the time. edit: And yeah, it's a lot higher than 50%, but since I haven't done the math I don't know what it is exactly and will leave that to someone else. Tyler, it seems like you are taking a personal interest in figuring out whether there is increased close spawning? Is this something you're experiencing? As for the results, you really can't draw anything from them because this is a sample of 100 out of freaking a shmajillion games (the population) played on metalopolis. You'd need access to blizz stats to really figure out how random the draw is. EDIT: well with that edit, I suggest we infiltrate blizz, or get a hundred people to do a hundred games, then pool dem results. | ||
|
Chill
Calgary25986 Posts
On February 16 2011 05:22 Wolf wrote: Can't you look up the spawn location percentage chances in editor? Or, better yet, isn't it the same for all maps? 100%/number of spawn locations? We know Blizzard can eliminate spawns (see: Shakuras Plateau). I think the implication is that the programming says it should be random, but testing shows it isn't. | ||
|
Vei
United States2845 Posts
On February 16 2011 05:21 MorroW wrote: i spawn close on metal like 50% of my games. maybe im just being biased cause im zerg and only remember the bad luck but still they should make it like shakuras Sums up all my thoughts :> Interesting. I wonder if metalop recently got changed, because I used to get more cross-spawn ^^ On February 16 2011 05:22 Chill wrote: I think the implication is that the programming says it should be random, but testing shows it isn't. Perhaps they never intended for all maps to be equal spawns; in Shakuras it's obviously a (what I presume to be) 50-50-0 ratio for spawns; the could've stealth patched or something Metalop to make it 50-30-20 or something (close/air/cross). They haven't necessarily said they want spawns to be X, even in the case of shakuras... unless I'm mistaken which I may be, seeing as how I don't follow blue posts except what's reposted here. | ||
|
JTouche
United States239 Posts
Thumbs up Blizz  | ||
|
Harmonious
179 Posts
On February 16 2011 04:57 emythrel wrote: The thing is that while there are 3 possibilities of positions, close ground, close air, cross. Each time you load up a game it has equal chance to roll any of them, completely independent of previous rolls, that means it will never go 33% under any sample size..... it could be 99% close ground positions in theory, thats the wonderful thing about randomness. While its not completely random, no RNG can be completely random, only as random as you can program it to be. The funny thing about true randomness is that it doesn't look as random as it could be to people. If you ask someone to write down 100 random numbers between 1 and 10 you can probably tell that it is not random since there aren't enough sequences of stuff (avoid saying patterns since there aren't patterns in randomness). Sequences would be 5,5,5,5,5 or 1,2,3,4,5,6,7 and so on. A true random sequence would have quite a few of those. Just a funny fact. In this case I think we can say with a pretty high confidence that something is up. Maybe between 60 and 80 percent (look up confidence intervals etc. if you want more details). I would certainly not bet my life on the fact that something is amiss. If TL is up for an experiment we could always flip two coins 100 times and compare results. I can pretty much guarantee that some very "skewed" results would appear. | ||
|
SiguR
Canada2039 Posts
On February 16 2011 05:21 MorroW wrote: i spawn close on metal like 50% of my games. maybe im just being biased cause im zerg and only remember the bad luck but still they should make it like shakuras I actually have had similar experiences. Before this thread was made I always felt like there were many more close position spawns occurring than otherwise. I previously would ignore the notion and put it down to the odds just working out that way. | ||
|
Twistacles
Canada1327 Posts
| ||
|
thrawn2112
United States6918 Posts
On February 16 2011 05:20 Chill wrote: Interesting. Instead of arguing about the reasonable sample size, let's either: a) Figure out how we're going to test it with a larger sample size; or, prefereably b) Figure out why Metalopolis is prone to spawning clone positions. Maybe it would be useful to bring the argument to a conclusion, just so teamliquid can set some standards for posting statistical analysis. Of course internet + arguments = fail, especially when the majority of people on here have never taken a college level stats class. If I were to ever open a thread about sc2 related statistics and see words like null hypothesis, p value, etc with values assigned and work shown.......who am I kidding lol. | ||
|
DueSs
United States765 Posts
| ||
|
Centorian
United States95 Posts
Play metalopolis against the computer for one game. Answer poll according to your result. However, it would be easy enough for people to lie, or use the poll based on their feelings and not actually doe the test. | ||
|
Draconicfire
Canada2562 Posts
On February 16 2011 05:26 Centorian wrote: A poll could work to get a lot of information very quickly. Play metalopolis against the computer for one game. Answer poll according to your result. However, it would be easy enough for people to lie, or use the poll based on their feelings and not actually doe the test. What if we started a research thread where people can post their replays on Metal and then tally it from there? The only issue I see is if the same two people post the same replay, so it would count twice. | ||
|
italiangymnast
United States246 Posts
Poll: What spawn position did you get? Close Ground (53) Close Air (23) Far (23) 99 total votes Your vote: What spawn position did you get? | ||
|
sikyon
Canada1045 Posts
On February 16 2011 05:24 Twistacles wrote: Do replay files have, hard-coded into them, some variable that stores spawn location? Because if it did, it wouldn't be too hard to code an application to go through a bunch of replay files. This is what I was thinking. SC2replayed can parse your replay by race and even BO, so can we not create some program that can parse replays by spawn position? I don't have the time to look into it but that would be pretty neat, and we could download a bunch of replays off replay sites to analyze the sample. | ||
|
hypercube
Hungary2735 Posts
On February 16 2011 05:09 Liquid`Tyler wrote: Ah, you guys aren't likely to get your PHD's when you can't even understand the OP's testing methods. He didn't examine the games that gave him the feeling that he got close positions. He got that feeling and ignored those games and loaded up 100 more games against the computer and recorded those results. That's not the whole story. It's possible that thousands of people are bothered enough by the observed frequencies to run tests. Of these only one leads to exceptional result and it's the only one that gets reported. OPs testing method is correct from his point of view, but underlying "drawer effect" should change how we interpret his result. Anyway, to add some math to the discussion, taking a binomial distribution with p=1/3 the expected number of successes is 33.3 (duh) with a standard deviation of sqrt[np(1-p)]=4.7 So OPs result deviates from the expected by 8 standard deviations, pretty much killing any discussion about publication bias or sample size. | ||
| ||
