| Forum Index > BW General |
|
niteReloaded
Croatia5281 Posts
I just used a mix of stop and move commands away from the center stacking point. It seems that in 3-4 seconds they are unstacked. I'm assuming this will be banned from tournaments after it's been used in one or a couple of high publicity matches. PvT is slightly skewed in P favor, this might tweak it a percent or two more. As people practice this tactic, it might make the cornerstone of some strategies. | ||
|
TT1
Canada10004 Posts
On May 16 2018 07:09 niteReloaded wrote: I tried splitting the army (pure goons, as I assumed it would ensure maximum retardation) after stacking them - which would happen in case of failed recall - and it's not hard at all. I just used a mix of stop and move commands away from the center stacking point. It seems that in 3-4 seconds they are unstacked. I'm assuming this will be banned from tournaments after it's been used in one or a couple of high publicity matches. PvT is slightly skewed in P favor, this might tweak it a percent or two more. As people practice this tactic, it might make the cornerstone of some strategies.  | ||
|
Freakling
Germany1529 Posts
Note that this does not quadruple the capacity, though, as the origin area only takes unit position (~unit centre) into account, so for Dragoons they effectively can occupy up to 5x5 tiles (25 Goons in tightest square packing, recall on central Goon). On arrival, the units have to fully fit into the destination area, though (their collision box boundaries must not overlap the search area boundaries), and search starts from the Arbiter position in the centre. So when the area is unobstructed statistically about half a unit dimension will go to waste on each edge (so one unit dimension overall). So with Dragoons you end up with a 7x7 tile square (49 goons) for unobstructed target space, as I already empirically found. Let's look at Zealots: Their collision size is 23x19 px. 129/23 = 5 (truncated), 129/19 = 6. So you can fit up to 6x7 = 42 Zealots into a recall without stacking (that's 3.5 control groups or 84 supply). Furthermore, 5.5 · 23 = 126.5 < 127 and 19 · 6.5 = 123.5 < 127, so up to 11x13 = 143 Zealots could theoretically be recalled with stacking. This obviously blows the supply limit, and the unit placement might actually not be so optimal as to allow it, but it's probably safe to say that if your whole army supply in in Zealots, you should be able to recall you whole army with this. Probes are 23x23 px in size. So with analogous calculations one gets up to 11x11 = 121 Probes in one recall. Testing this is trivial, so excuse while I do it  edit: As predicted, actual maximum recall capacities are quite a bit lower than the ideal, due to inefficiencies in the unit placement algorithm, one of them being that no unit is actually moved directly to the Arbiter position. You can actually only recall 80 Probes at once. Realistically, you can probably get about 100 Zealots, which is still a maxed out supply (without anything else!), though. | ||
|
iloveav
Poland1478 Posts
Magicbox storming... now that is bonkers :D. | ||
|
niteReloaded
Croatia5281 Posts
OK, so it's not like that anymore? : )) I'm not following closely in recent times. I know Flash is killing everyone, but are other Ts > P?? | ||
TaardadAiel
Bulgaria750 Posts
On May 12 2018 02:06 Freakling wrote: My two cents regarding that: Right now I am neither here nor there regarding whether units being preset on patrol command helps with the mine clearing. On the one hand patrol issues an instantaneous attack command to units, which would allow them to diffuse mines before they can explode. On the other hand Recall resets all orders of affected units (puts them on stop), which would cause the first attack to come out slower. Now, depending on the exact order of code execution (frame by frame), it is possible that the following happens: Units are first moved to the new location, then their order status is reset, so that in theory they could be in the old order state for a frame or so, allowing for a fast attack. I'd suggest the following setup to test this: Make a control group or two of units (Goons) to be recalled and gather them in a tight pack (no overlapping, just a normal matrix configuration for a standard recall). Make an Arbiter and get the upgrades (all this using cheats or a map editor). Have an enemy Terran minefield laid out at the intended Recall location. Safe the game. Now Alternatively put the Goons on hold position (which also allows for faster attacks) or stop (slower attacks) command before recalling them. Look how many mines get diffused, and how many goons killed. Repeat this a few times (20 times for each setup at least, to get something approaching statistical meaningfulness). Alternatively, or in addition, check the game code (OpenBW) and/or relevant data files (although I think this is probably found on the hard-coded level) how exactly Recall is processed internally. You be the testing guru around this thread, so I won't argue on anything related to the exact mechanism. So far, I gather, it's just an observation that's been tested independently by some pros and presumed to be true until disproven. But it's a mechanic, let's call it that way, that can have a significant impact on the meta - one of the ways to deal with a recalling protoss goes down the drain. Leaving your base as terran with current rough timings where you reach a critical mass of upgraded mech and you assume the opposition has recall capabilities is suddenly much riskier. You either have to invest resources and space in turrets, possibly screwing your timing, or you have to invest APM directly or indirectly (through minimap awareness) on vessel defense. And if the stacked unit recall, even in normal size, has improved mine clearing abilities, then a unit cap on recalls won't bring the meta back where it was (then again, it could force protoss to be pickier in choosing which units they want to recall lest they leave behind a shuttle full of templar or an observer etc.). I'm very purposefully not mentioning balance, because it is much too early to tell if it would affect balance or rather just the meta. Most of the newer posts around here are almost ethical in nature and I won't get into that, but since I can't follow streams lately, I have no idea if the protoss players have gone from testing to actually using the trick in sponsored matches, the most competitive field we have outside tournaments. If anyone has any info on that, please share. I highly approve of blizzard's decision on following the matter through and not intervening in any way. We don't even know if the trick will impact the meta, though I assume it will. The big difference would come, as several people have already pointed out, from the way it impacts the meta. If the end result is terran players happily readjusting their timings with a uber-turtly style involving dozens of turrets, making the matchup highly predicable (or by forcing all the terran players to do earlier timing pushes unless they want to lose the game), then an intervention is maybe justified. If it improves the richness of the matchup, then by all means, keep it as it is. The effects on balance are even harder to estimate - it requires statistics with a rather consistent player base (comparable skill or adjusting for ELO) and mappool as well as quite some time to build a reasonably large sample. BW has evolved way beyond the wildest developer dreams and the player base has been instrumental in that, the developers just provided an amazing foundation for creativity to thrive. They know to leave it alone disregarding any intentions they had when designing unit behaviour. Please, have faith in the scene. Tl; Dr: please don't advocate changes based on some observations from noncompetitive environment and extensive theorycrafting. We just don't have data. | ||
|
Freakling
Germany1529 Posts
On May 17 2018 05:13 TaardadAiel wrote:You be the testing guru around this thread, so I won't argue on anything related to the exact mechanism. So far, I gather, it's just an observation that's been tested independently by some pros and presumed to be true until disproven. Well, most people don't really make any effort to get over their biases, so they tend to observe what they believe anyway, not what is really true. I made multiple tests under controlled conditions, and none of them suggests that unit orders before the recall make any difference in post-recall attacks. Furthermore I looked at the game code (OpenBW) and it resets the movement state (clears all movement orders in queue and sets on stop command AFAIK) for all units before relocating them, so movement state before the recall cannot have any effect.My guess is that some combination of the following effects is responsible for a (perceived or real) higher mine clearance success rate:
But it's a mechanic, let's call it that way, that can have a significant impact on the meta - one of the ways to deal with a recalling protoss goes down the drain. Leaving your base as terran with current rough timings where you reach a critical mass of upgraded mech and you assume the opposition has recall capabilities is suddenly much riskier. You either have to invest resources and space in turrets, possibly screwing your timing, or you have to invest APM directly or indirectly (through minimap awareness) on vessel defense. And if the stacked unit recall, even in normal size, has improved mine clearing abilities, then a unit cap on recalls won't bring the meta back where it was (then again, it could force protoss to be pickier in choosing which units they want to recall lest they leave behind a shuttle full of templar or an observer etc.). Again: stacking before the recall has no effect on unit position after the recall, unit orders before the recall simply get reset. So a 20 stacked Goon recall does not behave any different from a regular 20 Goon recall. | ||
|
StylishVODs
Sweden5331 Posts
On May 16 2018 22:08 niteReloaded wrote: OK, so it's not like that anymore? : )) I'm not following closely in recent times. I know Flash is killing everyone, but are other Ts > P?? nah | ||
|
funnybananaman
United States830 Posts
| ||
|
TaardadAiel
Bulgaria750 Posts
[B]On May 17 2018 05:44 Freakling wrote:Again: stacking before the recall has no effect on unit position after the recall, unit orders before the recall simply get reset. So a 20 stacked Goon recall does not behave any different from a regular 20 Goon recall. I just called the total recall a stacked unit recall, but the observed mine clearance improvement is correlated (not implying causation) to the stacking technique, right? | ||
|
Freakling
Germany1529 Posts
| ||
|
Twinkle Toes
United States3605 Posts
If I had known this 5 years ago, I'd have been B+ in ICCup! | ||
|
TaardadAiel
Bulgaria750 Posts
On May 17 2018 17:18 Freakling wrote: If there is an actual effect and not just observer bias, my guess is that it is simply correlated to the number of recalled units. So it is only correlated with stacking insofar as stacking allows you to fit more units into the target area. Ok, I might have missed it, but has this been tested? Recalling equal numbers (and compositions) of units "conventionally" and using stacking in the same mine field? | ||
|
Freakling
Germany1529 Posts
EDIT: I updated my test map to now also include a normal (25 goons) and big (50 goons) stack as well as a more sophisticated mine layout (now quite effective at blowing up all the goons). I still haven't added any other combat units or Observers into the mix. A short test run indicates that on average a normal sized recall (stacked or not) directly kills about 3 to 4 of 16 mines, whereas an enhanced recall (from the big stack) clears about 5 to 6 mines on average. EDIT 2: Might actually be more like 7 to 8 mines for the big stack recall. EDIT 3: Did some further modification to the map. Stacking is now fully automated, giving you a very reliable setup. The big stack is now 100 goons. I added a counter for recalled goons (gas value). I found that 56 goons can consistently fit into a single recall (112 supply, 4+ control groups). With those huge recalls, average mine clearing rate is about 8, double that of a conventional recall. | ||
|
zimp
Hungary951 Posts
On May 17 2018 22:31 Freakling wrote: The units will be placed without overlap upon arrival, so whether they have been stacked or not beforehand makes absolutely no difference. EDIT: I updated my test map to now also include a normal (25 goons) and big (50 goons) stack as well as a more sophisticated mine layout (now quite effective at blowing up all the goons). I still haven't added any other combat units or Observers into the mix. A short test run indicates that on average a normal sized recall (stacked or not) directly kills about 3 to 4 of 16 mines, whereas an enhanced recall (from the big stack) clears about 5 to 6 mines on average. EDIT 2: Might actually be more like 7 to 8 mines for the big stack recall. EDIT 3: Did some further modification to the map. Stacking is now fully automated, giving you a very reliable setup. The big stack is now 100 goons. I added a counter for recalled goons (gas value). I found that 56 goons can consistently fit into a single recall (112 supply, 4+ control groups). With those huge recalls, average mine clearing rate is about 8, double that of a conventional recall. good job. i can't test it now. does it matter if the mines are close to each other or separated? no terran will have time to plant 10+ mines to all areas in the base... | ||
|
Dazed.
Canada3301 Posts
| ||
|
orvinreyes
577 Posts
On May 18 2018 10:13 Dazed. wrote: So how have the results been in game, people? Is it effective? Thats all that really matters, not number crunching. Please wait for starcraft's 25th anniversary for conclusive results of the change in meta, thx. | ||
|
kogeT
Poland2037 Posts
On May 18 2018 10:13 Dazed. wrote: So how have the results been in game, people? Is it effective? Thats all that really matters, not number crunching. I did it few times in 4v4 BGH, works brilliant. | ||
|
L1ghtning
Sweden353 Posts
| ||
|
Timebon3s
Norway685 Posts
Anyone seen this in a proper game? | ||
| ||
StarCraft 2 StarCraft: Brood War Dota 2 League of Legends Counter-Strike Other Games |
|
Korean StarCraft League
CranKy Ducklings
RSL Revival
ByuN vs Cham
herO vs Reynor
FEL
RSL Revival
Clem vs Classic
SHIN vs Cure
FEL
BSL: ProLeague
Dewalt vs Bonyth
Replay Cast
Sparkling Tuna Cup
The PondCast
[ Show More ] |
|
|
EPT
