|
Let me start by saying I'm a complete noob, so forgive me if what I'm writing sounds retarded, but it seems logical to me from what knowledge I have from watching VODs.
Since I don't have a beta key and the current AI is pretty bad, I started doing random little tests on damage, how many shots it takes to kill what, etc. and I started testing the optimal number of drones for harvesting. I found a couple of things out, the more obvious of which is that mineral nodes which are closest to your HQ can support only 2 harvesters whereas the ones that are a space farther can support 3 with each worker being slightly less efficient as a result of the timing. So optimal saturation should be 2xclose nodes + 3xfarnodes which is usually 21 or 20. You can put more on, but beyond this point the mining rate of your worker will drop approximately 20 fold.
I figured I would try some tests to figure out the average income with various HQ positions in combination with certain numbers of harvesters. Long story short I found that placing an expo 1 space farther away from the minerals actually allows you to mine out a mineral field more quickly at full saturation for an increase of only 3 workers. The downside is obviously that your workers have to travel slightly longer to return home with their cargo, meaning that before you approach the saturation point you will be making less than you would with normal placement, making it pretty impractical for early expos.
I realized though that it could be applied to high yield expos in mid/late- game for a couple of reasons. First, the high yield fields don't have as many nodes as the normal ones, meaning the number of workers needed for full saturation will drop by around 6. Second, by the time you're expoing at the high yield you should have enough workers to fully saturate the nodes immediately when the HQ finishes, skipping the part of the curve where you would be mining more slowly than an HQ placed 1 square closer. Finally, because the high yields offer double minerals the disparity between the average intake of the differently placed HQs is also doubled accordingly.
So if you get a high-yield expo up, place it 1 space farther from the nodes than you normally would, and immediately saturate it with 18 workers, you will be able to mine it out a few minutes faster than someone with a properly placed HQ.
You can obviously test this really easily yourselves, but here are the basic numbers I came up with for anyone who is too lazy. I'm assuming that the income counter on replays is measured in units per minute: 1 worker mining at minimum proximity can collect around 40.5 resources per minute (gas and minerals mine at the same rate) 2 workers mining any node will acquire fully double what one would, but if a node 1 space farther is given a third worker, this third worker will increase your MPM(minerals per minute) by roughly 30 instead of 40, because there is a brief period of waiting for the previous worker to finish. Dancing workers (a third worker on a proximity node) only increase your tally by roughly 2.5 MPM, after which point the increases get exponentially smaller. Full saturation of a normal mineral field yields approximately 800MPM, whereas saturation with a misplaced HQ yields roughly 850-860MPM with a ceiling of about 900 if you take 3/4 of that (x2)(for 6 nodes as opposed to 8) this is about 1200MPM with normal placement on a high yield with full saturation, meaning you would get minerals at an increase of about 100-120MPM.
May not sound like much but one extra zealot a minute and having your high-yield mined out a couple minutes before your opponent can make a big difference in a close game.
I see very little downside to this if you are able to saturate the mineral field immediately. Other than harassment becoming slightly more effective. What do you people think?
|
On April 20 2010 10:19 Aether wrote: What do you people think?
I think the search function could have saved you hours of your life.
User was banned for this post.
|
well, not really. I was messing around with this stuff to learn about the game because I don't have access to b.net. I would've been doing something along these lines anyway.. Why, is this common practice?
|
Very interesting and nice work. I'm not sure anyone has done this so in depth before, so I'm not sure how he wasted his time Vexx?
|
On April 20 2010 10:39 Vexx wrote:I think the search function could have saved you hours of your life.
I disagree, nothing I found using the search function for "worker saturation" yielded anything like this. It's definitely something I'd never heard of before - everyone always thinks of optimal saturation if your nexus/cc/hatchery is as close as possible to the mineral fields, but nobody's yet seen whether your intake speeds up as Aether describes: increasing saturation past the normal optimal level and counterbalancing by adjusting the mining distance slightly. I'd be interested to see what the optimum combination of distance and saturation would be.
Edit: small typo.
|
I read through those threads, and people did touch on things like the fact that maximum saturation is 3 nodes + distance mining can be more efficient, but no one really put it all together into a direct in-game application.
|
I've got to agree there, I'd never heard of this before either. I find this to be pretty useful tbh.
|
Wow, interesting... thanks for the information.
When you say place the hq one square away, does that mean one square in both directions (assuming a corner high yield base), or one square in just one direction?
For example, if I spawn at 9 on LT and I am expanding to the high yield just below my natural (in the bottom-right, about 4 or 5 o'clock), the minerals make a corner formation. Do I place the hq one space left, or one space left AND one space up?
|
one square on both sides if the minerals are on a corner.
|
This topic will be closed after my post, you will see why
No need to thank me, i didn't make the graph.
User was banned for this post.
|
right, then why when people expand do they not place their Nexus/Hatchery/CC 1 space farther from the high yields? I've never seen it done in any replays... Is this not relevant information?
This chart also doesn't address the disparity between HQs placed in different locations. This graph is a small part of the overall puzzle. I think some people are just seeing the numbers for a saturated mineral patch and saying "yeah yeah, we know" without considering the extended implications which is what I'm actually focused on.
It is a direct in-game application of this information that could be potentially useful, especially for people with highly macro-oriented styles.
|
On April 20 2010 11:29 ooni wrote:This topic will be closed after my post, you will see why + Show Spoiler +No need to thank me, i didn't make the graph. Thanks (for not reading the OP and making yourself look like an ass).
|
On April 20 2010 11:29 ooni wrote: This topic will be closed after my post, you will see why
No need to thank me, i didn't make the graph.
Have you even read the OP ? This has nothing to do with it..
|
Wait, are you saying 18 workers mining at a base placed 1 space away mines faster than 18 workers mining at a base placed in the correct spot?
|
Sorry if you didn't understand the graph. I'll explain it thoroughly.
He said: Since I don't have a beta key and the current AI is pretty bad, I started doing random little tests on damage, how many shots it takes to kill what, etc. and I started testing the optimal number of drones for harvesting. I found a couple of things out, the more obvious of which is that mineral nodes which are closest to your HQ can support only 2 harvesters whereas the ones that are a space farther can support 3 with each worker being slightly less efficient as a result of the timing. So optimal saturation should be 2xclose nodes + 3xfarnodes which is usually 21 or 20. You can put more on, but beyond this point the mining rate of your worker will drop approximately 20 fold.
The optimisation according to graph is at 24.5 not 21 or 20, you can see from the graph gradient at 24.5 is not much lower than at 20 or 21.
He calculated in terms of 21 or 20, i.e. a case where the minerals aren't saturated. When saturated the distance make little to no difference.
misplaced saturation (this graph is for automatic mining) is 900-1000; It make little to no difference, where the workers are placed. This is because once you have 18+ peons start to dance towards an empty patch automatically. Either he miscalculated or tested with only two matches which brought an unstable result.
|
On April 20 2010 11:57 RifleCow wrote: Wait, are you saying 18 workers mining at a base placed 1 space away mines faster than 18 workers mining at a base placed in the correct spot?
yes, essentially.
The reason is because if the mineral node is 3 spaces from the nexus, the minimum allowed, only 2 workers can fit into the cycle, and there's a brief period when the patch is not being mined. If you over-saturate the patch, the workers do not wait at the patch the way they do with a vespene guyser unless the current worker is nearly done mining. This causes them to wander around from node to node since they are all saturated and his actual mining time is massively decreased.
if the nodes are 4/5 spaces away instead of 3/4, all nodes can accommodate a full cycle of 3 workers.
So instead of 14 workers at 100% + 4 workers at around 5% efficiency versus 18 workers all averaging around 95% of their maximum capacity.
On April 20 2010 12:11 ooni wrote:
The optimisation according to graph is at 24.5 not 21 or 20, you can see from the graph gradient at 24.5 is not much lower than at 20 or 21.
He calculated in terms of 21 or 20, i.e. a case where the minerals aren't saturated. When saturated the distance make little to no difference.
misplaced saturation (this graph is for automatic mining) is 900-1000; It make little to no difference, where the workers are placed. This is because once you have 18+ peons start to dance towards an empty patch automatically. Either he miscalculated or tested with only two matches which brought an unstable result.
I believe you're misreading your own graph. The intersection point is actually exactly at 21 workers on that graph, which is exactly where I calculated it to be... Your graph supports my evidence. What you're saying WOULD be true if workers didn't move away from nodes that are occupied and waited at them, but they are not able to do this if the node is 3 spaces away. Therefor, you need a wandering probe to randomly hit the open mineral patch during the brief period for which it is available. Since the chances of this are very low, the probe spends about 95% of it's time wandering around and only 5% mining.
|
Actually just checked, the optimal differentiation point in the graph, at the point is between 22~23. It does seem like 20~21 is ideal because if you look
20 brings 53 minerals p/m 21 brings 52 minerals p/m
Suddenly drops to 22 brings 50 minerals p/m
So 20~21 must be optimal right?
nope, with 23
23 bring 49 minerals p/m
Thus making 22~23 optimal
So what's the difference if it's 22+? when it's 22+ no matter what the peons still bounce between minerals, that means even you put 3 per minerals they will still bounce to an emptyfield. Though it does not look optimised that peons bounce around the minerals, the income wise it is beneficial. I believe you can benefit from optimising the nodes 2-3 only if you have low peon numbers or have full satuation and microing the peons constantly.
See the problem is I'm and everyone else on this thread is generalising workers. Drones are on creep, probes have different acceleration and SCVs are just SCVs.
|
Can this be the case with SC1 also because I remember hearing a comment on a replay video about a Russian SC1 player who always placed his Nexus 1 space away from what is considered closest, but the commentator had no idea way.
Anyways, nice job.
|
what that fails to incorporate is the fact that if a mineral node is occupied, the drone that arrives at it DOES NOT WAIT. This is important.
Imagine just one mineral node with a probe. This probe has a certain capacity for mining which is the distance it must move divided by its movement speed plus the time it takes to physically mine the minerals. To make it simple, we will say that the probe travels 1 square per second and the HQ is 32 spaces away from the mineral patch, and takes 36 seconds to mine the minerals. the probe takes a certain amount of time to complete 1 revolution, in this case exactly 100 seconds. 64 seconds back and forth plus 36 seconds mining time. If we want to add more probes to this mineral patch, we need to account for the fact that probe #1 is coming back in exactly 64 seconds, so if probe #2 consumes 36 seconds mining his minerals, this leaves only 28 seconds for probe #3 to mine before the first probe returns. Upon returning he says: "Hey! This guy is going to be mining for 8 more seconds? I would wait for 2-3 seconds, but I'm not waiting for that!" and wanders off to find a different mineral patch. If he would wait there the mineral node would be operating at full capacity and the drones would not, unfortunately they don't do this.
the solution to this problem is not to pack as many workers in as possible so that there are 5 of them waiting to mine a patch with a 20 second window, the solution is to INCREASE the length of a single rotation to better accommodate the probe's mining capacity. By increasing the distance from 32 spaces from the minerals up to 36 spaces, we create a cycle of exactly 108 units. Now 3 probes can each take their 36 seconds without interfering with one another as they are taking a few extra seconds to bring the minerals back.
After all of that, if you still don't understand what I'm saying, load up a game on an empty map, build 40 or so drones, 1 CC/Nex/Hatch in the usual spot at the high yield, and another one 1 space away from where you normally would, send 18 drones to each one and see which one finishes first. If you don't believe my logic, test it out. I guarantee the one that is placed 1 space away will finish at least a full minute before the one in the normal spot.
|
What if you force 3 workers to mine a close proximity patch? The third worker would wait for a while, but would it not mine at the same rate as 3 workers mining a far patch?
|
|
|
|