2 million games played and running, dunno this algorithm seems stupid.
Edit: Why does it always assume I want a spawning pool, if I only want drones/ovies?
Forum Index > SC2 General |
Adeny
Norway1233 Posts
2 million games played and running, dunno this algorithm seems stupid. Edit: Why does it always assume I want a spawning pool, if I only want drones/ovies? | ||
SwampZero
Greece350 Posts
On October 22 2010 21:33 Existor wrote: But I dont want roaches (as example) and want to find fastest Build Order for 50 lings. Why it builts me Roach Warren, if I dont want roaches in any waypoint and final ? Roach Warren is by accident a genetic trait of a lot of the successful build orders that are used to generate the next one. Give it more time, theoretically it will breed itself to extinction. | ||
Existor
Russian Federation4295 Posts
Roach Warren is by accident a genetic trait of a lot of the successful build orders that are used to generate the next one. Give it more time, theoretically it will breed itself to extinction. I understand this, but I don't want to build roaches and want only zerglings. P.S. Can we change number of processors or it will crash real processor? EDIT: nvm, after some restarts there are no more Roach Warrens with lings only | ||
teemh
Canada120 Posts
| ||
voss
Australia146 Posts
thx. | ||
Existor
Russian Federation4295 Posts
What do you think about this idea? | ||
Rosvall
Sweden122 Posts
On October 22 2010 21:45 teemh wrote: It needs an "at least X after build complete" box for both minerals and gas to be added to the build requirements. Indirectly there is. You can set how many extractors to build and drones on minerals and then just count backwards. All constants are in the code. Also, it values stuff, like there's a larger chance that a BO will build one drone than one evo chamber. These are constants and doesn't change depending on your input. So there's the same chance it will build a roach warren if you only want to build lings as there is if you only want to build roaches. The difference though is when the chromosome mutates it will find that a BO without roach warren is better and discard the old build. | ||
Rosvall
Sweden122 Posts
On October 22 2010 21:49 Existor wrote: Build Order Optimizer + http://haploid.nl/sc2/build_order/ = will be FTW What do you think about this idea? Have you actually looked at the output? | ||
Existor
Russian Federation4295 Posts
Have you actually looked at the output? WHat you mean? I said, that it will be great to see both services in one: good interface, timings, queen's energy, how many minerals we have for every moment, better view for build orders, etc etc | ||
Kagami-sama
460 Posts
On October 13 2010 08:58 Lomilar wrote: It is out! http://code.google.com/p/evolutionchamber A strange game. The only winning move is not to play Zerg. How about a nice game of chess? | ||
RobiTL
France55 Posts
apparently, <= 23 roaches = only 1 Hatchery, >= 24 roaches, it decides it needs another Hatchery. Try it yourself if you want ^^ Just thought it was a bit interesting and that a 23 roaches might be the most efficient roach push off 1 base. Also, when it decides to do another hatchery, it has to make a bunch more drones in order to pay back the investment I guess. Here is the summary (might not be the best builds because I stopped them at like 10-15 minutes, but still gives a broad idea I think) + Show Spoiler + 9 roaches : 15 drones : 4:55 12SP,15E,15RW,14O,16Q,16O 10 roaches : 15 drones : 5:10 15SP,15E,15O,15O,15RW,14Q 11 roaches : 15 drones : 5:21 15SP,15E,15O,15O,15RW,14Q 20 roaches : 16 drones : 6:52 10O,16E,16SP,18O,18RW,16Q,19O,19O,19O,19O,35E 23 roaches : 19 drones : 7:18 10O,13E,14SP,17O,16Q,21O,21RW,20O,20O,20O,45O ==== additional hatchery starting from here ==== 24 roaches : 27 drones : 7:42 10O,13E,16SP,17O,18H,17Q,22RW,26O,28O,.... 25 roaches : 25 drones : 7:44 10O,13E,16SP,17O,18H,17Q,22RW,26O,30O,44O,44O,48O,62O 26 roaches : 24 drones : 7:54 10O,13E,16SP,17O,18H,17Q,21RW,26O,30O,44O,44O,50O,64O 32 roaches : 26 drones : 8:30 10O,15SP,15E,15O,17Q,19H,22RW,26O,39O,39O,41O,41O,41O,47O,47E | ||
tetracycloide
295 Posts
| ||
teemh
Canada120 Posts
| ||
RobiTL
France55 Posts
On October 22 2010 22:25 teemh wrote: It keeps trying to use the 10/10 overlord then extractor trick. This doesn't seem right to me. You lose 6 minerals by canceling. Even assuming 11/10 drone was hatched fast enough to gather one mineral, that's only 5 a net loss of 1. With this in mind, you can argue that it's just front loading minerals, more earlier less later but I wonder if we'd get more realistic builds out if the extractor trick wasn't used? I think it's in order to lose the least larva possible (if you have 3 larvaes, you're basically LOSING larvas, you should count for that). So a 10over extractor trick is better than 10over no trick. However, to compare with the 9over, that's a bit tougher… edit : (also the program doesn't take into account the movement time for the drone to go build the extractor, and come back, so in practice, the program might also be wrong about the 10over extractor trick) | ||
poppahorse
2 Posts
| ||
chumpchous
68 Posts
I love giving it an initial waypoint that initially favors something like a 6 pool, you know, just get me 6 lings by 4 minutes. Then make the following waypoint something that demands some early economy. At first it goes "oh great! 6 pool, easy: spawning pool, lings" and your outputted drone count is really low. Then over time it ads more and more drones to your first waypoint as it realizes it gets more success in the next waypoint that way. This thing is awesome. Does it know how to remove drones from gas? It seems that it generally likes to just leave them on, even when it is clearly no longer beneficial. It would be nice if we could manually seed the generator with what we think is the most optimal build first. For example, if I want the fastest build that gets me 14 drones I know said build will never include a spawning pool but no queens so it would be nice if I could relay that information as an input because the algorithm takes a lot longer to make that determination than I do. Well, you can to a certain degree. If you just require to have two hatcheries by the 4 minute mark as one of your first waypoints, it will utilize the hatchery in meeting your following waypoints. | ||
Deleted_143
Australia256 Posts
| ||
Ahelvin
France1866 Posts
I believe the evaluation function is taking into account the completion time, but also the number of mineral/gas remaining. Thus if you want to make only drones and overloard, you will stockpile minerals eventually, minerals that could be used to make a spawning pool. Thus the program could chose to favour a scenario in which there is a spawning pool, because the fitness would be higher. Correct me if I'm wrong OP? | ||
icezar
Germany240 Posts
On October 22 2010 22:38 poppahorse wrote: this build order site is better - http://www.starcraft2buildorder.net/ Boy, that is a really bad site! I just can`t believe you wrote that, anyway..... | ||
BurningSera
Ireland19621 Posts
| ||
| ||
StarsWar
11: Korea Closed Qualifier Day 1
Solar vs GuMihoLIVE!
ByuN vs herO
[ Submit Event ] |
StarCraft 2 StarCraft: Brood War Bisu 2692 Dota 2Larva 903 Mini 892 Snow 454 Hyuk 337 ZerO 283 ggaemo 247 hero 212 Soulkey 209 Stork 148 [ Show more ] Counter-Strike Other Games Organizations StarCraft 2 StarCraft: Brood War StarCraft 2 StarCraft: Brood War
StarCraft 2 • aXEnki StarCraft: Brood War• intothetv • Gussbus • Kozan • IndyKCrew • LaughNgamez Trovo • Laughngamez YouTube • Migwel • Poblha |
Big Brain Bouts
BSL
TerrOr vs XuanXuan
Dark vs JDConan
Korean StarCraft League
StarsWar
WardiTV Invitational
CSO Cup
ForJumy Cup
BSL
Zhanhun vs WolFix
Dienmax vs Cross
Sparkling Tuna Cup
StarsWar
[ Show More ] WardiTV Invitational
Online Event
ESL Open Cup
Afreeca Starleague
StarsWar
ESL Open Cup
ESL Open Cup
Afreeca Starleague
StarsWar
Club NV x Duckling Show…
GSL Code S
Stats vs SHIN
Cure vs GuMiho
ESL Pro Tour
|
|