|
On October 16 2010 04:12 hangarninetysix wrote:
This tool would probably be handy for alot of lower level players but I don't understand the concern about ruining the game, it's not like it can factor in how many zerglings you need to build to defend when you see reactor hellion vs gasless 5 rax marine all in. I think it's pretty cool for refining specific timing attacks but im not sure if its very useful past the early game since you're not trying to have some preset amount of something, but rather respond to your opponent and stay ahead of him.
I've been using the 7 roach rush he provided in 1v1 and 2v2 and it works amazingly well. I was already roach rushing but this made it so much better. The difference between having 5 and 7 roaches is huge. Theres no way I would have figured out that 10pool, 10ovie, 18warren,18 2*ovie was the optimal way to go.
And yeah it's not useful for after the early game, but you shoudl read the thread theres heaps of posts like this already and he never claims it is.
|
Give it:
Make 3 roaches before xx:xx (where xx:xx is the time ling/baneling would be getting to ur base
Then:
Get 20 roaches with +1 and speed ASAP
Then:
Trim down the number 20 until you can get them out before most players could get mutalisks.
Then:
Win all zvzs forever with a +1 speed roach push.
|
On October 16 2010 08:44 BlitZl0l wrote:
Give it:
Make 3 roaches before xx:xx (where xx:xx is the time ling/baneling would be getting to ur base
Then:
Get 20 roaches with +1 and speed ASAP
Then:
Trim down the number 20 until you can get them out before most players could get mutalisks.
Then:
Win all zvzs forever with a +1 speed roach push.
Spine crawlers? Against other roaches? I don't think it's fair to oversimplify too much.
|
On October 16 2010 08:52 Chipotle wrote:Show nested quote +On October 16 2010 08:44 BlitZl0l wrote:
Give it:
Make 3 roaches before xx:xx (where xx:xx is the time ling/baneling would be getting to ur base
Then:
Get 20 roaches with +1 and speed ASAP
Then:
Trim down the number 20 until you can get them out before most players could get mutalisks.
Then:
Win all zvzs forever with a +1 speed roach push. Spine crawlers? Against other roaches? I don't think it's fair to oversimplify too much.
that is some good methodology though, itll net a ton of ZvZ wins
|
Therefore a minmax algorithm is applicable
Are you using the term minmax correctly here? My understanding of minmax (having implemented it to solve a few two player games) is that it requires an adversary. You take a turn - they take a turn, with them trying to lower the result and you trying to raise it (hence, minimax). I don't see how it can apply to BOs which are effectively single player games?
|
are those true seconds or "faster" seconds? might be confusing considering the ingame clock is on faster.
|
On October 16 2010 18:20 StarSense wrote:
are those true seconds or "faster" seconds? might be confusing considering the ingame clock is on faster.
It is in-game seconds.
|
On October 16 2010 04:12 hangarninetysix wrote:
This tool would probably be handy for alot of lower level players but I don't understand the concern about ruining the game, it's not like it can factor in how many zerglings you need to build to defend when you see reactor hellion vs gasless 5 rax marine all in. I think it's pretty cool for refining specific timing attacks but im not sure if its very useful past the early game since you're not trying to have some preset amount of something, but rather respond to your opponent and stay ahead of him.
Its like in Chess, a good opening helps .. anyway its just the opening. Its not about simple build orders such as "FastRoach", its also about optimal build orders which transition well at a given point in time (e.g. you scout and adapt). From what i see in the current pro openings, there is still a lot of suboptimal stuff going on.
|
This will only be good for the first 7 minutes but after that ur Build Order Program will ask u so many things at one time that you would wish u had 5 people controlling ur race than just urself.
|
Tried your build order in a few 2v2 games. Lost the first two, won the other two.
I usually build more roaches after the given build order and my brother goes muta's. Works pretty good except if the opponent goes stalkers/marines (as they did in the second game).
|
On October 16 2010 18:17 Almania wrote:Are you using the term minmax correctly here? My understanding of minmax (having implemented it to solve a few two player games) is that it requires an adversary. You take a turn - they take a turn, with them trying to lower the result and you trying to raise it (hence, minimax). I don't see how it can apply to BOs which are effectively single player games?
Ofc it applies. In chess minmax is applicable only with 2 players, because there is no game progress with one player ^^.
With one player RTS it works like:
1) at time T you have two options.
2) Branch (so you have two game simulations). One with action A done, other with option B done
3) do (2) for 10 time ticks creating more branches
4) at time T+10 run your weighting function, remove weakest branches
5) continue with (2)
the longer you can afford to create branches, the weaker the minmax weighting function is, (=computing time) the more likely you will find "sneaky" BO's, which will look bad for some 30 seconds but suddenly get well after some 60 seconds (e.g. early queen). Those BOs are likely missed by a human beeing.
|
Alright. I think I got a somewhat functional pre-alpha terrible looking UI for this application.
I'm looking for a group of people I can chuck this app to, and have them provide feedback, as well as pwn the ladder. Anyone who knows at least what software testing is, and can provide details.
Skype me. Username is Lomilar.
|
I dont have skype but I'd love to try it!
|
the longer you can afford to create branches, the weaker the minmax weighting function is, (=computing time) the more likely you will find "sneaky" BO's, which will look bad for some 30 seconds but suddenly get well after some 60 seconds (e.g. early queen). Those BOs are likely missed by a human beeing.
Still sounds nothing like minimax to me  .
|
On October 17 2010 10:30 Lomilar wrote:
Alright. I think I got a somewhat functional pre-alpha terrible looking UI for this application.
I'm looking for a group of people I can chuck this app to, and have them provide feedback, as well as pwn the ladder. Anyone who knows at least what software testing is, and can provide details.
Skype me. Username is Lomilar.
I'm interested but I don't want to use skype t.t.
Any other way I can try out your program?
|
It's actually quite scary for 2v2
Me and my friend can combine two builds from the program:
16 banes for one player
7 roaches for the other
They both finish at the same time(around 5 mins).
Almost impossible to stop.
|
An alternative build which I've found faster (your mileage may vary):
10 Spawning Pool
10 Extractor => 1 drone mining on completion
10 Overlord
10 Roach Warren
10 Add another drone to the extractor
10 Overlord
7 x Roach.
|
Baneling bust:
10pool
9gas
10 Extractortrick - Drone
11 Ovie
11Ling speed
11Banesnest
11 7*lings
18 6*banes
Time: 4:30
Units:
8 Lings
6 Banes
|
edit: pasted wrong code sry
|
On October 16 2010 20:44 Schnullerbacke13 wrote:Show nested quote +On October 16 2010 18:17 Almania wrote:Therefore a minmax algorithm is applicable Are you using the term minmax correctly here? My understanding of minmax (having implemented it to solve a few two player games) is that it requires an adversary. You take a turn - they take a turn, with them trying to lower the result and you trying to raise it (hence, minimax). I don't see how it can apply to BOs which are effectively single player games? Ofc it applies. In chess minmax is applicable only with 2 players, because there is no game progress with one player ^^. With one player RTS it works like: 1) at time T you have two options. 2) Branch (so you have two game simulations). One with action A done, other with option B done 3) do (2) for 10 time ticks creating more branches 4) at time T+10 run your weighting function, remove weakest branches 5) continue with (2) the longer you can afford to create branches, the weaker the minmax weighting function is, (=computing time) the more likely you will find "sneaky" BO's, which will look bad for some 30 seconds but suddenly get well after some 60 seconds (e.g. early queen). Those BOs are likely missed by a human beeing.
Uh, so this is not minimax search. You're describing some sort of tree search with pruning and a heuristic evaluating function (wtf is a "minmax weighting function"?), but not minimax, which requires an adversary who is trying to minimize your score (i.e. chess), or at least a model which creates a zero-sum game (i.e. modeling player vs nature, and trying to minimize the worst possible outcome, which again does not make sense here since all actions aside from opponents actions (which we aren't modeling) are within player's control).
Thank you for post OP, looks great, will keep checking this.
|
|
|
|
|
|
EPT
