EPT[SC2B] Gas Matters - Page 2
| Forum Index > News |
|
[GiTM]-Ace
United States4935 Posts
| ||
|
BBS
Germany204 Posts
| ||
|
sLiniss
United States849 Posts
Thanks for a great writeup + research! | ||
|
hacpee
United States752 Posts
On April 26 2010 19:45 Vetlock wrote: What has science done..But really a good graph and calculations,maybe your major is on math :p I see no math or science done here. All he did was collect some data, put it in excel and use data analysis. No fundamental relationship or equation was derived. | ||
|
Nah
Poland50 Posts
2 gas / second -- 3(1) 3 gas / second -- 4(2) 4 gas / second -- 6(2) These have the highest investment return. Relative 98-100%. The analysis was made for "Fast" not "Faster" speed of game, which should be noted I think. Also without any reference to other aspects of game these gas rates mean almost nothing. 16 workers mine ab. 10 minerals / second 20 workers-- 12 minerals / second 24 workers -- 13.33 minerals / second So the final minerals / gas ratio is 13.33/4 = 3.33 Ratio minerals/gas Thor -- 3 / 2 Tank -- 6 / 5 Marauder -- 4 / 1 Banshee -- 3 / 2 Viking -- 2 / 1 Stalker -- 5 / 2 Immortal -- 5 / 2 Colossus -- 3 / 2 Sentry -- 1 / 2 Phoenix -- 3 / 2 High Templar -- 1 / 3 Observer -- 1 / 2 etc. etc. I.e. Marauder takes 30 seconds to make. To make 3 marauders continously you should have 12,5 min / s , 2,5 gas / s. ( 125 * 3 / 30 , 25 * 3 / 30 ) It means that you could go with 20-24 workers at minerals and 4(2) at gas. | ||
|
Wintermute
United States427 Posts
It's interesting that you mentioned the difference in geysers depending on maps, but didn't test the actual difference in collection rates from map to map. It seems like that would have a significant effect on the outcomes and therefore the conclusions. For whatever it's worth, you get very similar/identical data using the in game "income" tool on replays. | ||
|
Chillz
Canada100 Posts
| ||
|
rockslave
Brazil318 Posts
| ||
|
ComusLoM
Norway3547 Posts
| ||
|
shoop
United Kingdom228 Posts
On April 26 2010 17:27 Arrian wrote:
While I think the OP is probably correct in its rough estimate for 3(1) as 2 gas per second and 3(2) as 2.25 gas per second, I don't think this is very clearly visible in the graph at all. A faster rate would be visible in the graph as a different slope, not just a higher or lower line. I don't think the graph shows clearly whether the slope for 3(2) is really different from the slope for 3(1); it is more as if the 3(2) line is consistently about 5 gas higher than the 3(1) line. Part of this could be explained by measurement error, part of it may be because in the 3(1) test the second and third scvs has to wait for the first two to come out of the refinery, while in the 3(2) test only the third scv has to wait a while. So I don't think this argument is altogether convincing, although I do tend to believe the conclusion. On April 26 2010 17:27 Arrian wrote: Now, a very compelling pattern emerges here, one that looks like an exponential function. Sorry, but this is nonsense. If the miners would not hold each other up, then the mean time between gas returns would obviously be inversely proportional to the number of miners; to be precise #gas = r * #miners * #time, where the mining rate r = 0.75 gas/second To calculate the time for a single return, substitute #gas = 4 to obtain #time = 4 / (r * #miners) = 5.33... / #miners. This imperfect model already fits the numbers quite well: #miners | 5.33... / #miners 1 | 5.33 2 | 2.67 3 | 1.78 4 | 1.33 5 | 1.07 6 | 0.89 Now obviously the miners do hold each other up, an effect that presumably gets worse when you increase the number of miners. Thus, in reality the mean time between gas returns will be larger than the amount of time predicted by the inverse proportional model. (This is exactly what happens: the predicted numbers are smaller than the measurements, especially for #miners equal to 5 or 6.) In contrast, in the proposed exponential model the mean time between returns drops ridiculously quickly as a function of the number of miners. For example, according to the exponential model the mean time between returns for 25 workers would be 0.00013; in other words you would collect 4/0.00013 = 31121 gas per second, while according to the inverse proportional model you would collect 25*r=18.75 gas per second, a much more reasonable figure. While an exponential model may give a reasonable fit if you just don't look at the graph beyond 6 miners, it is clearly a completely inappropriate model in this case, so you're just as well off just drawing any reasonable looking line through the data points. | ||
|
Hyldig
Denmark9 Posts
A good reminder to check the gas locations on your base, since one may be closer than the other | ||
|
Derogatory
Netherlands31 Posts
Don't have to search that out for myself now. | ||
|
DeR.Five
United States3 Posts
| ||
|
Anti
United States1113 Posts
| ||
|
Risen
United States7927 Posts
On April 26 2010 21:52 shoop wrote: Hm. I have some objections. While I think the OP is probably correct in its rough estimate for 3(1) as 2 gas per second and 3(2) as 2.25 gas per second, I don't think this is very clearly visible in the graph at all. A faster rate would be visible in the graph as a different slope, not just a higher or lower line. I don't think the graph shows clearly whether the slope for 3(2) is really different from the slope for 3(1); it is more as if the 3(2) line is consistently about 5 gas higher than the 3(1) line. Part of this could be explained by measurement error, part of it may be because in the 3(1) test the second and third scvs has to wait for the first two to come out of the refinery, while in the 3(2) test only the third scv has to wait a while. So I don't think this argument is altogether convincing, although I do tend to believe the conclusion. Sorry, but this is nonsense. If the miners would not hold each other up, then the mean time between gas returns would obviously be inversely proportional to the number of miners; to be precise #gas = r * #miners * #time, where the mining rate r = 0.75 gas/second To calculate the time for a single return, substitute #gas = 4 to obtain #time = 4 / (r * #miners) = 5.33... / #miners. This imperfect model already fits the numbers quite well: #miners | 5.33... / #miners 1 | 5.33 2 | 2.67 3 | 1.78 4 | 1.33 5 | 1.07 6 | 0.89 Now obviously the miners do hold each other up, an effect that presumably gets worse when you increase the number of miners. Thus, in reality the mean time between gas returns will be larger than the amount of time predicted by the inverse proportional model. (This is exactly what happens: the predicted numbers are smaller than the measurements, especially for #miners equal to 5 or 6.) In contrast, in the proposed exponential model the mean time between returns drops ridiculously quickly as a function of the number of miners. For example, according to the exponential model the mean time between returns for 25 workers would be 0.00013; in other words you would collect 4/0.00013 = 31121 gas per second, while according to the inverse proportional model you would collect 25*r=18.75 gas per second, a much more reasonable figure. While an exponential model may give a reasonable fit if you just don't look at the graph beyond 6 miners, it is clearly a completely inappropriate model in this case, so you're just as well off just drawing any reasonable looking line through the data points. Your math is wrong. He clearly states in his article that he doesn't use his regression for unreasonable combinations, like oh... 25 workers on gas. Based on his assumption that the graph holds up to 3 workers per gas, and allowing for some variance, the exponential regression is just fine. Take your math major and shove it up your ass, the first thing people do in "real" life is make a set of assumptions for each situation. So no, YOU are making complete nonsense. | ||
|
Risen
United States7927 Posts
On April 26 2010 23:34 Anti wrote: is it just me or is the first graph actually missing the 3(1) chart Read the entire thing next time? He says you can't see the 3(1) because the 4(1) lays over it. | ||
|
Anti
United States1113 Posts
I was thinking about gas usage the other day though, thanks for the article. | ||
|
Archerofaiur
United States4101 Posts
But seriously is there any game on the face of the earth that recieves this kind of dedicated research. Im sure Halo doesnt have statistical analysis on a professional level. | ||
|
Shikyo
Finland33997 Posts
On April 26 2010 23 begin_of_the_skype_highlighting 26 2010 23 end_of_the_skype_highlighting:43 Archerofaiur wrote: ESCIENCE! But seriously is there any game on the face of the earth that recieves this kind of dedicated research. Im sure Halo doesnt have statistical analysis on a professional level. SSBM, GGXX, any competitive fighting game. And this specific kind of analysis is ultimately quite pointless. | ||
|
Markwerf
Netherlands3728 Posts
First of all the method and data is nice and all but really this could be summarized much easier. It is pretty safe to assume beforehand to that rate of gas income is constant while the gas geyser is running as we all know this from playing experience. Thus just putting 1, 2 and 3 workers on a geyser and just timing for a minute each and calculating how much you gas income you would get would suffice, doing the 2nd gas as well is completely unneccesary as it's obviously the same as the first. So 4(2) is really the same as 2x 2(1)??, everyone could think of that beforehand.. The only interesting thing in here is how much less efficient the 3rd worker is then the first 2 ones using the geyser. The raw gas income is also slightly interesting. The inverse graph is completely useless and add's nothing to the whole issue, that entire paragraph should be scrapped really. Also the investment and income table at the end hardly makes sense as it counts the costs of workers which you would be having anyway. It neglects the oppurtunity cost of making the refinery as well which though different for each race can't just be neglected either. The whole writing and setup of this piece is just not good. Needlessly trying to do complicated things which add nothing to the whole subject... | ||
| ||
