|
Which attacks fastest? We knew what to do in BW, why has no one measured for SC2 yet?
This is the question Artosis asked here: Twitter.
I facetiously answered hold position, because I always use hold when I do a run by with lings in team games. With my limited APM, I found this most helpful to get some damage done. But at high level games with enough APM to respond quickly and either move workers or get attacking units in time to defend, it will be different.
So I accepted the challenge when Artosis replied back with a request for a link.
I decided to go into Unit Test Map by SuperFerret and test Attack, Hold, Patrol with these units:
21 Workers for all Three races on minerals.
1 DT
4 Zealots (no charge)
8 Marines (no stim)
6 Speed Lings
Each unit had 10 seconds to kill workers, this time is arbitrary and allows most units to get a few attacks off. Also the count of units is based on:
1 DT, usually keep them separated to maximize scans and your investment.
4 Zealots, enough to carry in a warp prism.
8 Marines, medivac.
6 Speed Lings, enough to not merit a large investment and a good counter when opponent is moving out.
Now the map has a different mineral patch setup than other maps, but figured with 21 workers it would create a good saturated scenario. The test has these standards:
Hold position was directly in the middle of the worker line where it maximizes amount of potential kills.
Attack move was done by moving next to base, then attack Move from left of base into the middle.
Patrol was similar to attack move, from directly left of base to patrol to the right and back in the middle of mineral line.
Below are my graphs to show which is best.
![[image loading]](http://i.imgur.com/U6AwE.png)
With this graph in place, we can come to these conclusions from the test.
1) Range units are ideal, probably 4 hydra's in an overlord will do more damage then 6 slings.
2) On ROI, 6 slings are the best, marines second, and 4 zealots third. (Based on 150 minerals for 6 lings, speed 100/100=250/100. Marines = 400+100/100=500/100)
Now this statement needs an explanation. DT's while probably having the biggest ROI potential, in a controlled environment was not the most cost effective with all the tech cost associated with getting 1 DT to a mineral line.
3) Attack Move and Patrol will generally be the best choice, now I did notice that hold position caused more units to be damaged than attack move, I do not think it's that relevant to this testing.
In closing, using attack move after getting units into position will be the best choice when dealing with ranged units and normal saturation. If an overly saturated based and melee units, hold position will be best.
This is my first informative post, so hope my organization makes sense. Any feedback much appreciated.
edit1: Changed graph to have a legend, so easier to read. Amove/Patrol are essentially the same exception of Speedlings.
edit2: I realize I am a nooblet, this does not answer Artosis question, I will now work on that as well.
|
That graph is somewhat hard to read. Not to flame, just to constructively criticize.
Because you aren't using a continuous variable across the X-axis, you should not use a connected line graph. Probably a histogram or simple dot plot would be better, and perhaps put the title in colour to match the key. Also, why not use 16 lings as that is equivalent cost to 4 zealots and 8 marines?
|
Canada13404 Posts
i dont believe artosis was necessarily referencing units in a mineral line but micro during a battle. Often move patrol move would have benefits with some units in micro situations where the same unit in move attackmove move would do worse than if patrol was used
|
I also believe he was referencing the time it takes to attack the first time. (split second)
Such as vultures fire more rapidly and slow down less when using patrol micro.
|
The picture won't load for me for some reason. =/
|
On May 17 2011 01:28 Hawkke wrote:
1) Range units are ideal, probably 4 hydra's in an overlord will do more damage then 6 slings.
I would imagine that 4 hydra's would do more damage, but they are also much, much more of an investment (just the overlord itself to bring them could cost 2/3 of what 6 lings would cost you if you lose it...).
I think testing with only 6 lings is sort of too few to be informative.
Interesting stuff though
|
You should just do separate graphs, that one is really sloppy and hard to read, it isn't at all like a traditional graph. I would do 3 graphs, one for each unit and use bar graphs for how many probes killed.
I haven't tested any of this, but from what I've seen in videos and in my experience with the game in general I would think that all of them would be the same for ranged units, and for melee units hold position would probably be the same as well if in range. People need to realize that the game is programmed much better than BW was, standards have increased a lot and error checking is a lot more automated than it used to be as well. I am sure that there are some micro tricks that will be more efficient than other, and maybe some abuses like the old 7% mineral trick that was patched out, but for things like attacking via patrol or amove I am sure blizzard made them function similarly based on distance to units.
|
Canada13404 Posts
On May 17 2011 02:08 ComaDose wrote:
I also believe he was referencing the time it takes to attack the first time. (split second)
Such as vultures fire more rapidly and slow down less when using patrol micro.
Ah yes this too but I believe this doesnt change due to the need to see the whole attack animation and also the speed decrease before speed increase being coded into the game better than it was in BW
|
Thank you for putting time into this, can you test in battle with some normal compositions? Like marines kiting zealots or something like that.
I think the problem is it comes down to the situation and player skill.
does anyone move lings to a mineral line to put them on hold position? If the probes go too far away they no longer get attacked (as you can see in the results), it seems patrol does the best job. How is Patrol differant from attack move in the first part of the path.
If I tell a stalker to attack move across the map, and then patrol across the map, what traits will be differant about the movement? I know it will come back after it makes it across, but i'm talking about just the initial movement before it tries to return.
|
I am using 6 slings, as the investment of 16 lings in regards to larva(8 total) is a high cost for an unknown ROI. While 6 slings can be a great counter to a forth coming push. This is based on mineral line harrass. If Zerg use 16 slings then I will change it, but in pro match ups, generally its very few.
|
The benefit to hold position in a mineral line is that if the workers are left mining the zerglings will not go running after nearby army units (as they would under other commands).
But as ZeromuS said, I don't think this is what Artosis was referring to.
|
On May 17 2011 02:02 ZeromuS wrote:
i dont believe artosis was necessarily referencing units in a mineral line but micro during a battle. Often move patrol move would have benefits with some units in micro situations where the same unit in move attackmove move would do worse than if patrol was used
Hrmm, if this is the case, then damn it.. I need to accept challenge again. I thought he meant in regards to mineral line harrass. Guess I deserve that, for having that in my mind.
Anyone else realize they have way too many starcraft related thoughts in their minds..
|
Great post, I got the information from the map and understood what you were testing. Thanks!
|
the point of using hold position with lings is so they dont run off to attack other things.
in testing hold position with lings, of course it wont be as effective.
|
This is a great attempt and I encourage you to do these things in the future, but I'm not sure you're answering the right question.
I'm pretty sure artosis was asking which fires quickest from the command and which is most effective when you're doing stutter-step micro or other similar types of micro. So the marine test in this case might be a reasonable test, but still not quite right.
If they a-move, hold position, and patrol all have the same reaction speed then you would expect patrol to do the most damage because it's going to have to spend the least amount of time on average acquiring a new target.
|
Yeah, he was referring to the patrol micro in BW.
|
Interesting and valid question. A few points though.
Because the x axis isn't continuous, connecting the points is just confusing. Leave them as points.
But most importantly, some of the numbers are a bit hard to draw conclusions from. for instance, yes, 8 marines is a much better drop then 6 slings. But you can't really say "ranged is better" because 6 slings are worth 150 minerals, 8 marines worth 400. And yes 4 Hydras would be better then 6 slings, but 4 Hydras is a way more signifigant force then 6 slings. 12 roaches would be probably better than 4 Hydras but that doesn't mean much.
But mostly I think 6 slings is bizarrely few. If I plan on counterattacking a mineral line, I'd probably send at least 10-12. It's also much more in keeping with the investment of your other drop examples.
|
Seems like patrol is the best option in almost all cases. And now that we can remap keys patrol can be mapped to Q for easier usage.
|
I would imagine hold position goes up as saturation increases.
|
I always assumed hold position was best because when the enemies attacking units come into play don't your units switch priority to them?
|
Allright so for my first post after following for 2 years 
The answer to the question has been tested out for microing unit! (has protoss)
So moving out of your base - Patrol works better - your unit will get in a concave instead of setting their self into a blob - for kiting with range unit (stalker) move/hold works better then attack move - unit attack a bit faster and one good thing about it if there is unit without nothing to attack they will stay with the army instead of running away and they will also keep their concave if you do it correctly has for attack move still if there is unit without something to attack they will move away and get yourself into a blob once again.
Now with melee unit (DT,Zealot and i put sentry in it) attack move works better if you have chargelots - for attacking or even mineral harass.
This is pretty much my conclusion after 2 hours of testing in custom game and versus AI.
Hope it helps
|
To be honest, a bar graph with 3 bars per category would probably be a lot more effective at conveying the data than the current line graph you have.
|
The graph is a bit confusing at first glance, and I question the methodology of "10 seconds attack workers."
This is a great question, but I'm just curious if there might be a better way, perhaps high FPS time-lapse recording or something to detect frame delay in each attack...
And it would have to be done offline.
|
I had always used stop (S). Why is that not included, is it already widely known that stop is inferior?
|
Good info, but crappy graph. Should've just showed the numbers, it would've been easier to digest the data that way.
|
|
|
whats the worker kill for slings with one zealot in the mineral line?
looool....
 unless patrol ignores unit priority (:
|
I appreciate this post, its informative and looks to prove something and expand upon what we know about SC 2
Thanks
|
Also my tactic with Dts really (because I have low Apm) is rally him into a mineral while warping in and just let him go. It gets plenty of kills at my level but I will look into baby siting him more with Patrol/Attack move shift queuing as I'm sure I could get more kills.
|
Interesting that zealots do more dps as they hold position.... would have thought the opposite
|
On May 17 2011 10:18 MusiK wrote:
Interesting that zealots do more dps as they hold position.... would have thought the opposite
you need to understand the context of the experiment.
He's trying to measure which method is better for maximizing dps in general, not the mineral line (at least as I understand).
In other words, which attack is faster, hold position, stop, attack move, and is it the same across all units (ranged and melee?)
What he found was that a zealot on hold position had better dps than if he was running around over 10 seconds on a more or less fully saturated mineral line.
I don't think anyone should make the logical leap that you should hold position on your zealot if you attack a mineral line.
edit: talking to hawkke on bnet as I wrote this, that is in fact what he's recommending based on this data.
At the very least I'm super duper skeptical of the viability of this unless it's late game.
In general it sounds like an atrocious idea.
|
So by this graph, patrol attacking is the best? Not sure if I'm reading it right
|
A line graph is possibly the worst choice for this type of analysis. A bar graph would be much better, especially since the x-axis is not even numerical.
|
wouldn't the differences be considered a bug? i never played broodwar, but this sounds like it's an unintended advantage for those that have the knowledge+skill.
a unit should patrol until a unit is in range, then attack.
a unit should move until a unit is in range, then attack.
a unit should hold position until a unit is in range, then attack.
there should be no animation lag or anything. maybe there is, but there shouldn't be, imo. it's kinda cool that it can add another element to the gameplay i guess, but it just seems kinda... bad to me.
|
On May 17 2011 02:21 ZeromuS wrote:Show nested quote +On May 17 2011 02:08 ComaDose wrote:
I also believe he was referencing the time it takes to attack the first time. (split second)
Such as vultures fire more rapidly and slow down less when using patrol micro. Ah yes this too but I believe this doesnt change due to the need to see the whole attack animation and also the speed decrease before speed increase being coded into the game better than it was in BW
well that is what he wanted to be sure of I think
|
@OP: How many repeats have you done for this test, for fairness?
|
|
|
On May 17 2011 02:27 Hawkke wrote:Show nested quote +On May 17 2011 02:02 ZeromuS wrote:
i dont believe artosis was necessarily referencing units in a mineral line but micro during a battle. Often move patrol move would have benefits with some units in micro situations where the same unit in move attackmove move would do worse than if patrol was used Hrmm, if this is the case, then damn it.. I need to accept challenge again. I thought he meant in regards to mineral line harrass. Guess I deserve that, for having that in my mind. Anyone else realize they have way too many starcraft related thoughts in their minds..
While I love the effort you put into this, I think the above poster was correct and you completely misinterpreted what Artosis was actually asking.
In BW, if you took a vulture and were trying to micro with it on the move, it would attack faster by using patrol commands than if you tried to use attack-move, stop, or hold position when you were ready to fire. There were other similar examples.
And the question Artosis was asking is whether equal differences between different commands exist in Starcraft 2.
Personally, I don't think they do, but it may prove that one offers a slight difference.
But a good analysis of mineral line harass regardless!
|
Canada13404 Posts
On May 17 2011 14:08 yomi wrote:Show nested quote +On May 17 2011 02:21 ZeromuS wrote:On May 17 2011 02:08 ComaDose wrote:
I also believe he was referencing the time it takes to attack the first time. (split second)
Such as vultures fire more rapidly and slow down less when using patrol micro. Ah yes this too but I believe this doesnt change due to the need to see the whole attack animation and also the speed decrease before speed increase being coded into the game better than it was in BW well that is what he wanted to be sure of I think
Yeah me too though someone posted here a little while ago about target acquisition speed. Perhaps one of the patrol, hold position and attack move commands minimizes something like overkill when a unit has a ranged projectile based attack? That would be amazing to learn that they do or dont :p I cant test now but someone who is awake and reads this and has time it would be cool to know if this is the case at all
|
On May 17 2011 11:52 Meatt wrote:
wouldn't the differences be considered a bug? i never played broodwar, but this sounds like it's an unintended advantage for those that have the knowledge+skill.
uhhh.. isnt that like.. what the whole game is about?
|
what? 6 speedlings? you test with 8 marines = 400 minerals, 4 zealots = 400 minerals, 1 dt which @ 1 gas = 2 minerals is 375, but 6 speedlings = 300?
an extra 2 speedlings could have made a significant difference given that you say you didnt count the injuries
were the workers scvs?
On May 17 2011 11:52 Meatt wrote:
wouldn't the differences be considered a bug? i never played broodwar, but this sounds like it's an unintended advantage for those that have the knowledge+skill.
a unit should patrol until a unit is in range, then attack.
a unit should move until a unit is in range, then attack.
a unit should hold position until a unit is in range, then attack.
there should be no animation lag or anything. maybe there is, but there shouldn't be, imo. it's kinda cool that it can add another element to the gameplay i guess, but it just seems kinda... bad to me.
holding position doesnt really fit with the other two there, in a mineral line scenario
and attack move is somewhat more aggressive than patrol in terms of chasing
|
Guys, it's not about which is "the best". In BW patrol, attack and hold postition had different effects on units and their kiting micro. With some commands units would fire quicker than they would with others.
Artosis wants to know whether attack, patrol or hold position results in faster attack response whilst kiting, making micro more efficient. I've personally tried them and didn't see signifcant differences, but maybe someone could take more of a scientific approach
|
|
|
|
|
|
EPT
