EPTPerfect micro with Phoenixes - Page 3
| Forum Index > SC2 General |
|
ne4aJIb
Russian Federation3209 Posts
| ||
|
TheoMikkelsen
Denmark196 Posts
| ||
|
shin_toss
Philippines2589 Posts
| ||
|
skorched
United States81 Posts
![[image loading]](http://33.media.tumblr.com/tumblr_ls286xzAXv1r07nhuo1_400.gif) User was warned for this post | ||
|
endy
Switzerland8970 Posts
| ||
|
kramuti
4 Posts
Essentially treat each of the corruptors as point sources, and obtain the envelope of all the added spherical waves (well circular in this case). aka, waves obey superposition. | ||
|
uh-oh
Hong Kong135 Posts
 | ||
|
Xiphias
Norway2223 Posts
User was warned for this post | ||
|
Mortal
2943 Posts
| ||
|
Goibon
New Zealand8185 Posts
today is one of those days | ||
|
k10forgotten
Brazil260 Posts
Awesome! | ||
|
Sholip
Hungary422 Posts
On July 25 2014 08:02 fmod wrote: Cool, what is your background, like education wise? I'm currently studying as mechatronics engineer, will start my 3rd semester in September.  On July 25 2014 08:40 pigmipigmi wrote: Why not take a different tack entirely? Instead of looking for an absolutely PERFECT mathematical answer by making simplifying assumptions (in this case, no microing of the corruptors) which cause the practical application to be necessarily inexact, look for a PERFECT (or much closer to perfect!) practical result by refusing to make simplifying assumptions and actually get in there with a necessarily inexact multidisciplinary approach using psychology and game theory (to model micro choices) AND math (to generate the data on which to base the game theory)? Well, that seems a bit complicated. How could I guess how the opponent decides to micro their Corruptors? Anyway, I think their only option is to pull them away, in which case you can simply follow them with the Phoenixes and attack them. Once they turn around again, you can continue microing in circles. If they don't turn around, then you can follow them until you have killed all of them (or eaten a Fungal in which case all your Phoenixes die .)On July 25 2014 09:40 Anacreor wrote: Is the ratio of the angle 45,98/45 equal to the ratio of the movement speeds, or maybe when it's squared? Actually, the angle is acos(u/v), where acos is the inverse cosine function (or maybe it's arccos in English? I can't remember). On July 25 2014 10:00 MavivaM wrote: but instead of retreating and attack again like I'd do, rather escaping by 45 degrees left/right or the pursuing unit (possibly the opposite direction where infestors/thors are). Am I right or there's more that I have missed? The 45 degrees only applies to Corruptors. Other units have different speed and range values, so the angle will be different, as described in the linked pdf. On July 25 2014 15:03 uh-oh wrote: I just learned about the rotation matrix along with many other things you used in your derivation last semester, it's nice to see that it can actually be applied to productive pursuits like SCII Same for me actually. On July 25 2014 14:26 kramuti wrote: The solution to multiple corruptors is given by Huygen's Principle. Essentially treat each of the corruptors as point sources, and obtain the envelope of all the added spherical waves (well circular in this case). aka, waves obey superposition. Could you explain that a bit more detailed, please? | ||
|
Sholip
Hungary422 Posts
On July 25 2014 09:44 [PkF] Wire wrote: You may want to correct c_1 = u and not 0 on page 3 -hope I'm not making a fool of myself, it's actually pretty late and I could have misread ^^. Of course, you're right! I'll correct it, thanks! Looks like you read it through really thoroughly. | ||
|
Dingodile
4135 Posts
On July 25 2014 07:40 Sholip wrote: Then you have to adjust accordingly to another circle. It shouldn't cause you to lose Phoenixes because the Corruptor can only attack your Phoenixes if your opponent recognizes the pattern in your movement and intecepts the Phoenixes at another point of the circle. Worst thing to happen is that you can't kill the Corruptor. Oh I was under impression that phoenix dont get a single attack from corrupter as you showed in that video. If this is possible too when opponent microes his corrupter. | ||
|
Clazziquai10
Singapore1949 Posts
| ||
|
kramuti
4 Posts
Could you explain that a bit more detailed, please? It may be easier to look at it using the position and velocity of the center of mass. The optimal angle to minimize the damage should be the same as if you calculated the velocity of the center of mass and used this as pathing for the velocity (if you have multiple phoenix, treat them the same way)...essentially you are treating X corrupters as 1 with an weighted average of position/velocity Vcm = Sum of (mi*vi) / Sum of (mi). where mi is the mass of each, and vi is the velocity of each (so the denominator is just the total mass). Using a mass of 1 should be ok... It should give you the same answers as this (i think): Since you are both on a circle of radius r, essentially treat each object to be on a wavefront (crest of the wave is the line cut out by the respective circles. To solve the problem for multiple objects, you have several overlapping circles. The answer is obtained by using superposition (waves add like vectors, so you can use matrix operations). The first way is probably the cleanest way to think about it, and looks like is would be a bit easier to use in your derivation. This, in general, will not keep the phoenix from being damaged, but it should minimize it. If you wanted a time evolution (aka, corruptors and phoenix coming from the bases to the fight, you 'simply' evolve your system, and either use circular wave addition, or using the velocity of the center of mass. One thing that was intersting to me is that you never seemed to account for the projectile speed, or the animation speed of the firing. Are both of these assumed to be the same? and are they? I suppose even if you use this, the answer could be obtained using a characteristic velocity. Maybe it doesn't matter to some extent if the corruptor is always out of range...hmmm... This is still a little vague, and I have given no proofs. I am pretty sure this should work though. Hope is at least gives you and idea on how to look at multiple objects. There may also be some cases that could be found in which there are actually better optimal paths that minimize the number of units lost, rather than minimizing the total damage taken. I would have to think it through more, and probably have to put pen to paper. Edit: to fake micro, have the corrupters do a (weighted )random walk. where there is a good chance the corruptors will go towards the phoenix, but not always. It won't be perfect, but you can start to characterize what are the best ways to move the phoenix given particular corruptor spacings, and the like (better players will keep their units in the optimal positions, moving in the optimal direction). I would still probably start with using center of mass values of things, and then tinkering with the individual units' motion while keeping the center of mass values fixed. | ||
|
ZenithM
France15952 Posts
| ||
|
Sholip
Hungary422 Posts
As for projectiles, though, I think their speed doesn't matter. As long as the Corruptor can't attack, its projectile is obviously out of the picture, and the Phoenix's attack will hit the target sooner or later. | ||
|
kramuti
4 Posts
On July 26 2014 00:12 Sholip wrote: @kramuti Well, that indeed is a bit vague, I have to admit. As for projectiles, though, I think their speed doesn't matter. As long as the Corruptor can't attack, its projectile is obviously out of the picture, and the Phoenix's attack will hit the target sooner or later. Maybe this is clearer. (with a link for a pic!) Since the mass is irrelevant, the velocity of the center of mass of N corruptors is simply the average of their velocities. Vcm = (v1 +v2+...vN.) / N The position of the center of mass is the average of all the individual postions. Rcm = (r1+ r2+...rN) / N. Use vectors Vcm, and Rcm, and apply constraints appropriately. You will get a circle of a new radius. This radius can be obtained using basic wave theory as well. The average velocity of any corruptor in your derivation is zero. (note so the average velocity of the center of mass is also zero) So it can be treated as if it is still at the center of the circle. If you add in another corruptor. Draw an identical ring for the phoenix's path around it. The path the phoenix should now take is given by drawing the envelope that surrounds both circles. http://s3.amazonaws.com/minglebox-static/img.1330326294889.47e3d31f.gif Here each of the points on the inner circle are the corruptors. The small rings are the ones the paths a phoenix would take in a one-on-one fight. The outer circle is the path a phoenix should take to minimize damage, from all these corruptors...aka a bigger circle where you connect all of the dots of all the tangent points. The phase should just be the addition of the indidual phases (I think). Again, this would minimize the damage taken, but I don't think it would maximize kill ratio. You probably really need a numerical model to give optimal trajectories under specific constraints. I also think this is where fire rates, animation times, acceleration rates (assuming the general solution is NOT symmetric...aka you will take damage for anything more than 1 on 1) will start to apply. Edit: clarification and correction. | ||
|
Cheticus
United States83 Posts
| ||
| ||
