|
On October 13 2010 18:06 Klesky wrote:Show nested quote +On October 13 2010 15:35 darmousseh wrote:On October 13 2010 15:29 Klesky wrote:
If it takes several hours to go through all the combinations and do random trials and take averages. All to determine with a high -probability- that it is 'optomized.'
Why would you write it in Java? It sounds like a very intensive operation.
C/ASM. Should bring that over-night down to a reasonable amount of time. lol, java runs just as fast 99% of the time. I laugh 99% of the time that I hear this.
Which shows how little you know about the subject. Because most applications are not CPU bound which in turn make java run just as fast as C/ASM or whatever. Most applications today are disk bound or network bound, which makes the processing speed of your code quite irrelevant.
That being said this application is most likely CPU bound which does not necessarily make java the ideal choice. In which case a similar but easy language such as C# or VB.NET can probably offer significant speed boosts.
There is no real point in making it in C or ASM due to the complexity of coding bug free and efficient code in these languages, particularly if he has any plan to release it as a community project.
But once again it all comes down to how experienced you are in a language and in which manner you compile the application. A poorly coded implementation in any language is almost always slower than a well coded application in another language. So if the creator is not very familiar with anything but java then the benefit of coding in another language might not be as big.
As for the program itself, I doubt it will make any major impact on the community. Anyone who play higher level games knows that you can't just blindly follow a build order. Not to mention most pro gamers have become quite well versed at discovering ways to cut costs in one area to speed up something else.
At most I see this program slightly speeding up discoveries like "what if I cut 1 drone here" or "what if I build my gas two drones later", etc...
|
On October 13 2010 21:47 dejavue wrote:
Seeing as there are some people here majoring in similar areas, wouldn't a collaborate effort make sense?
Just my two cents, I myself have no idea about programming of any sort, would however LOVE to see this program released...
There's no point in collaborating for a small program such as this. Too much overhead and all the co ordination would take longer than making the actual code.
|
On October 13 2010 21:54 Risen wrote:Show nested quote +On October 13 2010 21:45 archy wrote:
Is there something wrong with the mineral counter? I noticed it lists alot of wierd mineral values. In the example you have for example 51, 52, 53 or 106 minerals, and minerals are collected 5 at a time..
Would be very interested to see this released. It would be a great consultant program like Rawr is for WoW players. Minerals ARE collected 5 at a time...
Which is my point. The build order he shows does not however.
|
I don't see how there could be any moral problems with that.
It doesn't diminish creativity, it merely give pointers.
Using your current Roach build example, the application just gave us an all-in roach rush that is basically shut down by any kind of early pressure.
The nice thing is that you could give it different parameters to, say, make a BO for like 10 slings + 4roaches or something, while specifying "I want 6 lings at XX:XX minute mark because I know it's the timing for a reaper harass.
I can see this application optimizing more spefically all-in builds, like figuring out the perfect Baneling or Mutas or Hydras rush. The interesting part would be to compare the machine's BO to any BO figured out by players, and see the pros and cons, perhaps find good compromises. I can't see this as a bad thing.
|
On October 13 2010 21:47 Glacius0 wrote:
I might be wrong, but I get the feeling your build forgets that there is a limit of 3 larva where the hatchery won't spawn any additional larva. This is just my impression from having tried the build and having my 7th Roach pop @ 4:57.
I'm thinking of taking my drones off gas after reaching 7x25=175 gas and getting an expo.
Your build is slower then the program because you are not a robot basically...
|
On October 13 2010 22:15 archy wrote:Show nested quote +On October 13 2010 21:47 Glacius0 wrote:
I might be wrong, but I get the feeling your build forgets that there is a limit of 3 larva where the hatchery won't spawn any additional larva. This is just my impression from having tried the build and having my 7th Roach pop @ 4:57.
I'm thinking of taking my drones off gas after reaching 7x25=175 gas and getting an expo. Your build is slower then the program because you are not a robot basically...
I wonder if the program calculates the idle drone time when the game starts and when drones first hatch.
|
Great idea, but flawed.. for anything other than tier 1 units it sounds pretty average to me.
i.e. Quickest way to Ultras? the program will tell you how to do it as quick as possible but it doesn't take into account the opponent (only slightly important...lol).
|
Interesting idea, but I don't think you have anything to worry about.
I don't see it getting any real usage other than finding the best possible way to execute cheese.
And even then, you will need good micro to win against defenders advantage, especially if they scout you. There's nothing game breaking here IMO.
It really can't take into account how valuable more drones are vs. more units in a given situation for example. That needs to be your decision because it depends on your ability to scout and choose correctly based on what your opponent is doing.
In short, this will help you if you plan to win all your games within 7 minutes or so, but otherwise not so much.
(Though I'm sure the code is quite interesting, props. CS is a big part of my major, so I'm naturally interested.)
|
@Lomilar
Appreciate the effort on possibly making a UI for this build optimizer.
But it would be nice if you could just release the source code for now so that the community can benefit from your work and also contribute back. Thanks!
|
It's interesting that it thinks its better to 10 overlord. Isn't 9 overlord supposedly more efficient?
This program could prove insanely useful. Thanks! Definitely let it see the light of day.
|
On October 13 2010 23:11 Wr3k wrote:
It's interesting that it thinks its better to 10 overlord. Isn't 9 overlord supposedly more efficient?
Your question is flawed.
First define efficient then define better, then find the scope. The program is only interested in 7 fastest roaches.
|
On October 13 2010 23:15 SwampZero wrote:Show nested quote +On October 13 2010 23:11 Wr3k wrote:
It's interesting that it thinks its better to 10 overlord. Isn't 9 overlord supposedly more efficient? Your question is flawed. First define efficient then define better, then find the scope. The program is only interested in 7 fastest roaches.
True, I believe 9 overlord gets you ~15 more minerals around the 4-5 min mark, but may very well be slower for roaches since you are 10overpooling. If you plan to overpool I suppose 10 overlord is always going to be better. If its more of a macro build where the pool is later I'm sure the program will tell us to 9overlord.
What I'm curious about is whether or not such a program would allow you to calculate something like:
Get:
12 speedlings by 4:15
+1 carapace
10 muta by 7:30
+1 air carapace
All while maximizing resource gathering and spending any excess on drones.
As much as I like knowing the fastest way to get X, I can see it being quite limited. Would it be possible to say... get 16 mutas, and also have 5 roach and 20 lings before 8mins? Can the program accurately handle a build where a zerg fast expands? How does it deal with drone saturation? Does it assume optimal? What about maynarding workers?
|
Nergasm. This is an awesome awesome application. I think it's a bit of a time saver but not going to ruin anything. Correct me if I'm wrong but don't toss and terran get to know the fast build orders by virtue of their more linear growth? I don't think that difference is thought of when they balance the game, else zerg would be weak at start and strongest later.
|
|
|
i'm kind of torn on this, it seems like your moral dilema is very justified. On the one hand, it helps everyone (assuming a protoss and terran equivalent are made) However, it takes a very integral part of the game away. Honing build orders is a MAJOR part of the game, and one of the reasons the top pros make their way to the top. I personally dont think it should ever see the light of day, however if it did i'd probably still use it 
|
I've got my own version of this program and it's been helping a ton.
Mine's a bit different from the OP's though. I basically place constraints at certain time values and otherwise max econ.
For example, I know that a 4gate all-in will appear at time x+rushDistance with units y. I know from experience that it takes a units of b composition to defeat that. My program then finds the best econ build to get a units of b composition at x+rushDistance time. I've also made it so you can put in multiple compositions and the program figures out which is best.
I've also made it so you can have multiple sets of constraints. So I can essentially say "I want to fast expand against fast Hellions and then transition into Muta asap. What's the best opening to do so?" - this is important in my mind because often times the best way to counter Hellions [or whatever], doesn't provide a smooth transition into anything else, it's just good for Hellions.
Experience is really the key. You have to be able to extrapolate from "I'm going to be facing this, I need that to not die". You also need to be able to recognise what your opponent is doing to be able to apply the right build. It also requires knowledge of the metagame, "I expect to face these openings, what builds are safe against them?" I constantly tweak the numbers of units required as my ability to micro them changes, I might find that I can do just as well with one less Roach - which changes the entire build.
|
On October 13 2010 23:18 Wr3k wrote:Show nested quote +On October 13 2010 23:15 SwampZero wrote:On October 13 2010 23:11 Wr3k wrote:
It's interesting that it thinks its better to 10 overlord. Isn't 9 overlord supposedly more efficient? Your question is flawed. First define efficient then define better, then find the scope. The program is only interested in 7 fastest roaches. True, I believe 9 overlord gets you ~15 more minerals around the 4-5 min mark, but may very well be slower for roaches since you are 10overpooling. If you plan to overpool I suppose 10 overlord is always going to be better. If its more of a macro build where the pool is later I'm sure the program will tell us to 9overlord. What I'm curious about is whether or not such a program would allow you to calculate something like: Get: 12 speedlings by 4:15 +1 carapace 10 muta by 7:30 +1 air carapace All while maximizing resource gathering and spending any excess on drones. As much as I like knowing the fastest way to get X, I can see it being quite limited. Would it be possible to say... get 16 mutas, and also have 5 roach and 20 lings before 8mins? Can the program accurately handle a build where a zerg fast expands? How does it deal with drone saturation? Does it assume optimal? What about maynarding workers?
Based on what he has answered in this thread. Yes on all but the "optimal"-question where he claimed he added arbitrary delays to somethings.
If 12 speedlings by 4:15 is met before your desired time, you would probably be able to add an amount of drones to the goal until the algorithm returns 12 speedlings @ 4:15
|
As zerg is a completely reactive race, I don´t know if it could change the esports scene, except if you could it on the fly (which I doubt would be viable ...)
|
On October 13 2010 23:33 10or10 wrote:Show nested quote +On October 13 2010 23:18 Wr3k wrote:On October 13 2010 23:15 SwampZero wrote:On October 13 2010 23:11 Wr3k wrote:
It's interesting that it thinks its better to 10 overlord. Isn't 9 overlord supposedly more efficient? Your question is flawed. First define efficient then define better, then find the scope. The program is only interested in 7 fastest roaches. True, I believe 9 overlord gets you ~15 more minerals around the 4-5 min mark, but may very well be slower for roaches since you are 10overpooling. If you plan to overpool I suppose 10 overlord is always going to be better. If its more of a macro build where the pool is later I'm sure the program will tell us to 9overlord. What I'm curious about is whether or not such a program would allow you to calculate something like: Get: 12 speedlings by 4:15 +1 carapace 10 muta by 7:30 +1 air carapace All while maximizing resource gathering and spending any excess on drones. As much as I like knowing the fastest way to get X, I can see it being quite limited. Would it be possible to say... get 16 mutas, and also have 5 roach and 20 lings before 8mins? Can the program accurately handle a build where a zerg fast expands? How does it deal with drone saturation? Does it assume optimal? What about maynarding workers? Based on what he has answered in this thread. Yes on all but the "optimal"-question where he claimed he added arbitrary delays to somethings. If 12 speedlings by 4:15 is met before your desired time, you would probably be able to add an amount of drones to the goal until the algorithm returns 12 speedlings @ 4:15
Ahhh, so if you added drones to get 12 speedlings @ 4:15, could you also tell it to go from there and calculate the fastest way from that point to get 12 mutas?
|
On October 13 2010 21:45 archy wrote:
Is there something wrong with the mineral counter? I noticed it lists alot of wierd mineral values. In the example you have for example 51, 52, 53 or 106 minerals, and minerals are collected 5 at a time..
Would be very interested to see this released. It would be a great consultant program like Rawr is for WoW players.
Probs because of extractor trick. If not that than because income is not always in interval of 5.
Just look at the income tab in game. It's almost never exactly a multiple of 5.
|
|
|
|
|
|
EPT
