|
Honestly, the only time I put any effort into splitting is when the mineral patches are laid out in a 3-2-3 pattern like this:
![[image loading]](http://i.imgur.com/At0EN.png)
You build worker first and then quickly split them to the clusters of 3/3. All 6 workers will return twice before your first worker is done, unlike other splits where you usually have a quarter of a second delay before the second one builds. That's literally the only advantage that splitting will ever give you... Building another worker 1/4 second earlier.
It's almost so trivial as to be meaningless. It's basically the same as APM spam: if the first minute of the game wasn't so hopelessly boring then I wouldn't bother with it at all.
|
On February 03 2011 04:25 nickwtf wrote:Show nested quote +On February 03 2011 03:18 kzn wrote:
...
Thus there is an interval equal in time to d every 17 seconds where Build->Split has a mining rate advantage of K. Thus, every 17 seconds Mtb "catches up" by Kd.
It is obvious that Split->Build has an advantage in total minerals early on. Thus we need to know when Build->Split breaks even with Split->Build:
...
I'm sure you understand the math a lot better than I, so I'm not arguing against this, but can you explain this in a little more detail? I just don't understand why you basically get the time advantage of d compounded every time a new worker comes out. If you (build -> split) and I (split -> build), you are always only one worker ahead of me for a time duration of d. If you have 12 workers, you are saying you would count d six times (for the 6 you produced)? I see that 6 workers will obviously produce more minerals in d time than one, but when you get your 6th worker, I still have 5 producing during d, so its still a 1 worker lead for a duration of d... I don't pretend to be any kind of math expert, so if I'm missing something obvious, please just explain it. I'm just confused about the d advantage being compounded. Thanks
In strict terms this isn't a compounding advantage, but it works like this:
Build->Split:
T = 0, First Worker Starts
T = d, Initial Workers Mining
T = 17, First Worker Completes, 7 Workers Mining
T = 34, Second Worker Completes, 8 Workers Mining
and so forth
Split->Build:
T = 0, Initial Workers Mining
T = d, First Worker Starts
T = 17 + d, First Worker Completes, 7 Workers Mining
T = 17 + 17 + d, Second Worker Completes, 8 Workers Mining
And so forth.
Because you can only produce one worker at a time, if you start building later and everything is constantly producing, your final worker will finish later, by the same delay.
Basically, you start your worker d seconds later than mine, so I have a period of d seconds where I have 1 more worker than you every time one of my workers completes, until something stalls a worker completion. Every time this occurs, I get dK minerals as a result of that.
It doesn't compound, its not like there's 2d delay for the 8th worker, but worker production is bottlenecked because you can only produce one at a time.
[edit] Might have misunderstood you. The reason its 6dK by 12 workers is because I had 1dK for every probe 7->12
|
I didn't mention that this is done on specific maps actually, and thank you sevia for pointing that out. XNC, Scrap, Steppes, Metal, are maps that I know have the 3-2-3 mineral formation, which is where the 6-1-1 split is superior to other mineral splits. This is based on the irrefutable fact that all 6 probes are on the closest mineral patches, and that mineral patch distances are not all equal. Therefore, if you have 4 on the closest and 2 on the farther ones, your income will be logically lower than a player who has their probes on the closest mineral patches, it's simple math.
The 6-1-1 split works because of the natural spread the mining AI makes when 6 workers target the middle mineral patch. On every 3-2-3 mineral formation, if you target the middle mineral patch, and select the end probes to mine the outermost mineral patches, your 4 remaining probes will select the close mineral patches.
If you don't care about the extremely minor advantage this gives you, don't do it. However, not caring and trying to refute factual evidence that at the smallest degree that this split is better (however minor that advantage is) than other splits are different things.
|
On February 03 2011 06:10 tehemperorer wrote:
I didn't mention that this is done on specific maps actually, and thank you sevia for pointing that out. XNC, Scrap, Steppes, Metal, are maps that I know have the 3-2-3 mineral formation, which is where the 6-1-1 split is superior to other mineral splits. This is based on the irrefutable fact that all 6 probes are on the closest mineral patches, and that mineral patch distances are not all equal. Therefore, if you have 4 on the closest and 2 on the farther ones, your income will be logically lower than a player who has their probes on the closest mineral patches, it's simple math.
The 6-1-1 split works because of the natural spread the mining AI makes when 6 workers target the middle mineral patch. On every 3-2-3 mineral formation, if you target the middle mineral patch, and select the end probes to mine the outermost mineral patches, your 4 remaining probes will select the close mineral patches.
If you don't care about the extremely minor advantage this gives you, don't do it. However, not caring and trying to refute factual evidence that at the smallest degree that this split is better (however minor that advantage is) than other splits are different things.
It is insufficient to argue this purely on theory. We are simplifying things everywhere, including assumptions about probe AI, assumptions about the nature of ones mineral count as a function of time and workers, and all sorts of other stuff.
If you can do said splits and actually show an increase in income, fine. But "logically, it must be this way" is not actually the same as factual. I've tried a number of times to find a benefit from splitting in tests and I cannot.
|
On February 03 2011 05:31 kzn wrote:
Might have misunderstood you. The reason its 6dK by 12 workers is because I had 1dK for every probe 7->12
Alright it makes sense to me now. What each new worker is making in d time is additive, but d itself is not. I was thinking about this all wrong. all wrong...
Thanks
|
On February 03 2011 06:45 kzn wrote:Show nested quote +On February 03 2011 06:10 tehemperorer wrote:
I didn't mention that this is done on specific maps actually, and thank you sevia for pointing that out. XNC, Scrap, Steppes, Metal, are maps that I know have the 3-2-3 mineral formation, which is where the 6-1-1 split is superior to other mineral splits. This is based on the irrefutable fact that all 6 probes are on the closest mineral patches, and that mineral patch distances are not all equal. Therefore, if you have 4 on the closest and 2 on the farther ones, your income will be logically lower than a player who has their probes on the closest mineral patches, it's simple math.
The 6-1-1 split works because of the natural spread the mining AI makes when 6 workers target the middle mineral patch. On every 3-2-3 mineral formation, if you target the middle mineral patch, and select the end probes to mine the outermost mineral patches, your 4 remaining probes will select the close mineral patches.
If you don't care about the extremely minor advantage this gives you, don't do it. However, not caring and trying to refute factual evidence that at the smallest degree that this split is better (however minor that advantage is) than other splits are different things. It is insufficient to argue this purely on theory. We are simplifying things everywhere, including assumptions about probe AI, assumptions about the nature of ones mineral count as a function of time and workers, and all sorts of other stuff. If you can do said splits and actually show an increase in income, fine. But "logically, it must be this way" is not actually the same as factual. I've tried a number of times to find a benefit from splitting in tests and I cannot.
Fair enough, when I get a sec I will try my best to organize a controlled approach to testing my theory. I'm not sure at this point what it will entail, but I will think about it to make sure it is direct and without holes.
|
|
|
|
|
|
EPT
