|
On November 29 2010 07:32 jdseemoreglass wrote:The difference between a 1/17 drone and a 16/17 drone is clearly 15 SECONDS. Those have to be weighted more when measuring the prorated value of a drone in production.
Yes, the difference has to be weighted. I'm saying you're weighting it wrong. (well, actually I'm saying you're measuring the wrong thing, and possibly compensating for that inaccurately)
For this to have any relevance to real builds, the point of the drone count has to be a measure of how many units the build is producing. The relevant comparison statistic has to be something like "On average, build X has produced Y more units than build Z."
The statistic you're actually computing is loosely correlated, but I assert it introduces a lot of unnecessary error into the comparison. The hypothetical examples I mentioned were just to demonstrate that your ad hoc statistic measures a lot of irrelevant things.
|
Reading this makes me cringe. You cannot PROVE anything using the scientific method, only DISPROVE. You are really just gathering data to support which theory you believe is correct.
|
On November 29 2010 08:18 PackAttack wrote:
Reading this makes me cringe. You cannot PROVE anything using the scientific method, only DISPROVE. You are really just gathering data to support which theory you believe is correct.
|
On November 29 2010 08:18 PackAttack wrote:
Reading this makes me cringe. You cannot PROVE anything using the scientific method, only DISPROVE. You are really just gathering data to support which theory you believe is correct.
I have removed the word PROVE from my OP... perhaps now you can submit a comment that is actually helpful or constructive to the discussion?
|
On November 29 2010 08:09 Hurkyl wrote:Show nested quote +On November 29 2010 07:32 jdseemoreglass wrote:The difference between a 1/17 drone and a 16/17 drone is clearly 15 SECONDS. Those have to be weighted more when measuring the prorated value of a drone in production. Yes, the difference has to be weighted. I'm saying you're weighting it wrong. (well, actually I'm saying you're measuring the wrong thing, and possibly compensating for that inaccurately) For this to have any relevance to real builds, the point of the drone count has to be a measure of how many units the build is producing. The relevant comparison statistic has to be something like "On average, build X has produced Y more units than build Z." The statistic you're actually computing is loosely correlated, but I assert it introduces a lot of unnecessary error into the comparison. The hypothetical examples I mentioned were just to demonstrate that your ad hoc statistic measures a lot of irrelevant things.
But resources mined is just as relevant, if not more, than current number of drones. Take a look at the top two builds right now. The hatch first build results in fewer drones and overlords, but because the drones mined longer on two bases, it has higher total resources mined. The overpool build may produce more units, but so much early gathering is sacrificed by staying on 10 drones for so long and also oversaturating the main before the expansion finishes.
You cannot just use a simple measure of supply and larva... That is why so many threads before this one failed to produce meaningful results.
|
I've been wanting to do something like this myself for quite a while, but haven't had the time. So, props to you for taking this on. Can I suggest, though, that you add two additional pieces of data to your results:
The time the spawning pool finishes.
The time the expansion hatchery finishes.
While I understand that you want to test solely based on economy, in large part because that is something you can look at rather objectively, the reality is that your ability to hold off early rushes also matters. What these times will show is how much early "safety" a build is giving up for its benefit in economy. Both the spawning pool and the hatchery finishing make you much safer. Being able to say" I am sacrificing 20 seconds on my hatchery to gain 2 drones" is much more useful than just "I get 2 extra drones, but who knows how unsafe I am." While I'm not saying optimize based on these points, I think we do want to know what the tradeoffs are.
|
On the philosophy of science and SC2:
Take your silly "science can only disprove" logic back to biology and physics where things are actually complicated. You can't objectively "prove" what the "best" openings are because you can't predict where the metagame will be next week/month/etc, but given specific opponent openings it's perfectly possible to prove what the most economical responses are, which seems to be what this thread is about. Give me a break.
On topic:
Most of these threads/posts seem to be assuming that an opponent doesn't exist, and for the most part don't seem to be factoring in larvae/zergling timings, and what your opponent is likely to do based on your opening. For example if I FE vs protoss he might cannon or zealot rush me, if he doesn't FE himself. If I do a 13 pool expand, he's likely to do something totally different. Removing all that context might make for an interesting mathematical debate, but seems pretty useless as far as strategy goes.
|
On November 29 2010 08:18 PackAttack wrote:
Reading this makes me cringe. You cannot PROVE anything using the scientific method, only DISPROVE. You are really just gathering data to support which theory you believe is correct.
i wish people would stop nitpicking. the point of this is to look for the best opening, while OBVIOUSLY balancing risk. noone is going to actually suggest 22hatch 21pool or something.
stop fucking splitting hairs and arguing over semantics, its not constructive. YOU KNOW what the OP is asking you, so stop expecting him to babysit you and hold your hand and draw pictures. People who are trying to act intelligent in this thread are actually sounding like idiots, i.e. like you know nothing about the game/don't understand the OP.
please stop trolling by trying to sound clever.
|
This actually showed me something a lot more interesting. When you play at slower speeds you can actually manage your drones perfectly (getting that 100% mining uptime). I ended up with 1100 minerals (a lot more than the current "optimal" build) and 40 drones with 2 larva ready and 8 injected larva almost done with just a regular 14 hatch 15 pool build on my 1st try.
|
My whole point is this: You cannot say at the beginning "omg dont worry about other player just build" then later on write "no you idiot we have to do builds that are part of the metagame".
You are telling me first of all to not worry about what the other player is doing because we need to "work on just general BOs", and then you say that we cant do 6 hatch 6 hatch 6 hatch 6 hatch 6 hatch on every expansion on the map then just drone because its impractical vs another player? Be consistant. Either care about the other player's actions or dont. I submit that the most economic build order if you give an unlimited time is 16 hatch 16 hatch 16 hatch across the map then make spawning pool and a queen for each hatch then drone to 200. Thats the most economic build. Of course its not practical, but you said we arent worrying about the other player. If we ARE worrying about the other player, than the entire thread is moot, because there isn't one general build you can do vs everything! you dont even factor in scouting workers! I bet 13 pool 16 hatch loses to a 6 pool rush or a 2 gate proxy.
So dont tell me that 16 hatch across the map is unrealistic because it doesnt take into account the other player, then turn around and tell me that we only want to focus on "general builds" or "sound builds". You are just telling me I can only pick builds from the metagame, and test those.
On the secondary note of the science thing, I'm just saying if you want to do this in the scientific method, which I took to be OP's original intent, then we must realize there is no way all of these builds account for anything. Without factoring in scouting workers, for instance, every one of these builds is useless to use every time, because while you may be sitting on 42 drones and a queen at 6 minutes, a 6 pool rush kills you way before that.
You are pretty much asking "Hey what's the best chess opening?"
|
On November 29 2010 08:54 SpadeT wrote:
I've been wanting to do something like this myself for quite a while, but haven't had the time. So, props to you for taking this on. Can I suggest, though, that you add two additional pieces of data to your results:
The time the spawning pool finishes.
The time the expansion hatchery finishes.
While I understand that you want to test solely based on economy, in large part because that is something you can look at rather objectively, the reality is that your ability to hold off early rushes also matters. What these times will show is how much early "safety" a build is giving up for its benefit in economy. Both the spawning pool and the hatchery finishing make you much safer. Being able to say" I am sacrificing 20 seconds on my hatchery to gain 2 drones" is much more useful than just "I get 2 extra drones, but who knows how unsafe I am." While I'm not saying optimize based on these points, I think we do want to know what the tradeoffs are.
Sure, I will add this data in the OP to help people make informed decisions.
|
On November 29 2010 09:20 Guerrilla705 wrote:
My whole point is this: You cannot say at the beginning "omg dont worry about other player just build" then later on write "no you idiot we have to do builds that are part of the metagame".
You are telling me first of all to not worry about what the other player is doing because we need to "work on just general BOs", and then you say that we cant do 6 hatch 6 hatch 6 hatch 6 hatch 6 hatch on every expansion on the map then just drone because its impractical vs another player? Be consistant. Either care about the other player's actions or dont. I submit that the most economic build order if you give an unlimited time is 16 hatch 16 hatch 16 hatch across the map then make spawning pool and a queen for each hatch then drone to 200. Thats the most economic build. Of course its not practical, but you said we arent worrying about the other player. If we ARE worrying about the other player, than the entire thread is moot, because there isn't one general build you can do vs everything! you dont even factor in scouting workers! I bet 13 pool 16 hatch loses to a 6 pool rush or a 2 gate proxy.
So dont tell me that 16 hatch across the map is unrealistic because it doesnt take into account the other player, then turn around and tell me that we only want to focus on "general builds" or "sound builds". You are just telling me I can only pick builds from the metagame, and test those.
On the secondary note of the science thing, I'm just saying if you want to do this in the scientific method, which I took to be OP's original intent, then we must realize there is no way all of these builds account for anything. Without factoring in scouting workers, for instance, every one of these builds is useless to use every time, because while you may be sitting on 42 drones and a queen at 6 minutes, a 6 pool rush kills you way before that.
You are pretty much asking "Hey what's the best chess opening?"
I am simply trying to find the most efficient build possible. That certainly seems like relevant data, even if the specific build is not viable to the game. If you really want me to try 6 hatch 6 hatch 6hatch I will, but I can almost guarantee you will reach the 6 minute mark with almost nothing to show for it. You are making an assumption about what is most economical and we are trying to break open the assumptions here.
|
I am simply trying to find the most efficient build possible. That certainly seems like relevant data, even if the specific build is not viable to the game. If you really want me to try 6 hatch 6 hatch 6hatch I will, but I can almost guarantee you will reach the 6 minute mark with almost nothing to show for it. You are making an assumption about what is most economical and we are trying to break open the assumptions here.
If we are trying to break open assumptions why does it have to be 6 minutes? You cant claim "X is completely and utterly proven to be the best BO because it has the most drones and minerals at 6 minutes". 6 minutes is a completely useless number unless you also test each one of these at 0:00-5:59, and 6:01-9999 9. The whole experiment is just based on one arbitrary time, we need to accept the implications of this. It means: yes, this may prove the best economic strategy is you are playing a 6 minute game. It does NOT prove anything about the most economic strategy as a whole.
To my understanding, and correct me if I'm wrong, you are trying to find a balance for the most economic opening while still being able to do this practically and not have to worry about unit production in non-cheese games. I submit that to figure this out we need to find a very specific time for when the majority of openings finish, and I bet its not a pretty number like 6 minutes. To do what you are trying to do will take a much more in depth sequence of experiments than is presented here. In the true manner of science you need to make an experiment that is only testing 1 very speicfic thing and can come up with evidence of a conclusion one way or another. So you need to refine this experiment to somehow take into account all final times of all builds as well as all possible opponents builds. Or, you need to specify that this experiment is NOT proving what is the most economic opening for zerg; rather, this experiment reasonably demonstrates to us which build produces the most drones and minerals at the 6 minute mark.
tl:dr, refine your variables, its just too broad of an experiment to do.
|
On November 29 2010 09:37 Guerrilla705 wrote:Show nested quote +I am simply trying to find the most efficient build possible. That certainly seems like relevant data, even if the specific build is not viable to the game. If you really want me to try 6 hatch 6 hatch 6hatch I will, but I can almost guarantee you will reach the 6 minute mark with almost nothing to show for it. You are making an assumption about what is most economical and we are trying to break open the assumptions here. If we are trying to break open assumptions why does it have to be 6 minutes? You cant claim "X is completely and utterly proven to be the best BO because it has the most drones and minerals at 6 minutes". 6 minutes is a completely useless number unless you also test each one of these at 0:00-5:59, and 6:01-9999 9. The whole experiment is just based on one arbitrary time, we need to accept the implications of this. It means: yes, this may prove the best economic strategy is you are playing a 6 minute game. It does NOT prove anything about the most economic strategy as a whole. To my understanding, and correct me if I'm wrong, you are trying to find a balance for the most economic opening while still being able to do this practically and not have to worry about unit production in non-cheese games. I submit that to figure this out we need to find a very specific time for when the majority of openings finish, and I bet its not a pretty number like 6 minutes. To do what you are trying to do will take a much more in depth sequence of experiments than is presented here. In the true manner of science you need to make an experiment that is only testing 1 very speicfic thing and can come up with evidence of a conclusion one way or another. So you need to refine this experiment to somehow take into account all final times of all builds as well as all possible opponents builds. Or, you need to specify that this experiment is NOT proving what is the most economic opening for zerg; rather, this experiment reasonably demonstrates to us which build produces the most drones and minerals at the 6 minute mark. tl:dr, refine your variables, its just too broad of an experiment to do.
You are wrong in suggesting that 6 minutes is an arbitrary or unreliable standard. I could explain that long-term economy is necessarily predicated upon consistently maximizing short-term economy, beyond the point at which necessary investments are overcome by their payoff, due to the exponential-multiplier effects which an economy naturally assumes, but until I provide evidence to support it, it is a meaningless assertion.
Please, just submit whatever build you want and see how it stands up to the current leaders and stop wasting time theorycrafting, as I specifically requested in the OP that it be excluded from the discussion. If you want to extend your build to a later time, go ahead, but if I can produce a build from those listed here which surpasses it then I hope you would put the discussion to a rest.
|
OP has been shown time and again that they need to add more variables to their question if they want an answer: either a time constraint or a min+drone+hatch+queen restraint.
They aren't giving it.
Therefore, this is a troll thread - it is here to get Zerg players to spend time and energy on it when they could be elsewhere talking about something constructive, and generating ill-will when their hard work in attempting to get enough information to actually help is met with derision.
I'm done here, and my fellow Z, you should be too.
|
ty mcmasters. this experiment is far too broad to work. I respect your attempts OP, but please make it more confined next time please.
And I'm not arguing that short term economy is important, I am arguing that claiming short term economy = 6:00 is a rash assumption. Why not 5:46?
|
I can't believe people are complaining about increasing the amount of data we have available to draw conclusions.
How else do you think we'll ever work it out? Of course it's not perfect. Welcome to real science, where you go and test things and it's messy but you do your best and it seems to work. Bashing the investigation for being complicated is fruitless! Of course it's complicated. But we've solved harder things in science.
The timing question is in interesting one, and it makes for a nice followup investigation - can an opening that appears sub-optimal at 5.46 become superior at 6.00? Exploring that would help us interpret this data a lot more.
|
On November 29 2010 10:06 Guerrilla705 wrote:
ty mcmasters. this experiment is far too broad to work. I respect your attempts OP, but please make it more confined next time please.
And I'm not arguing that short term economy is important, I am arguing that claiming short term economy = 6:00 is a rash assumption. Why not 5:46?
So I need to make it more confined but 6:00 is too confined? lol...
Go back and choose 5:45 if you like... I bet you will find the some builds come out in the same exact order. The variables aren't defined? I clearly stated that minerals mined and drone count were the primary variables being tested. I think that is what 99% of us mean when we talk about economy. The methods here are sound. If you have a problem with the constraints, then put out a build that tests the constraints and by all means we will analyze it. Otherwise put the whining to a rest or get off the thread.
|
I may try and do the 13 pool, 16 hatch build myself, or at least look at the replay. EvoChamber has predicted this to be very solid, so I'm surprised it's not a top contender.
|
I haven't read all the comments but I think you should allow for the fact people will try and block hatch, so once your pool spawns you should dedicate at least 1 larva for lings. Take it into consideration and see how it affects going like 10 pool compared to 15 hatch
|
|
|
|
|
|
EPT
