I have been trying to find out why Blizzard and other AIs do not wall off as terran or toss against zerg? Over at the B.net forums I got;
"You'd need the computer to analyze pathing and read the terrain. That can't be done in the Galaxy Editor. The coding would also need to work correctly for over dozens and dozens of different maps. That's why it wasn't done in the first place, pathing with choke detection was never coded into the game. Also, it's too complicated".
That same reply lead me to Mtops early video that seemed to just that (analyze pathing and read the terrain).
I'm not a programer but all 2 player maps have 2 starting points, 4 player 4 and so on. This gives the zerg player a hatchery 6 drones and an ovie. The terran and toss get a CC or nexis and 6 workers at the start. Why can we not create points at the ramp or choke of the main that tell the AI where it can build a wall if the enemy is Zerg. Thus elemanating analyzing and reading terrain and giving the AI a where to build and leaving the what up to the AI. This would lead to a convention that all maps would not only have starting points but also wall points.
I find it funny how your name is cliff and you're talking about walls. Anyway, it certainly sounds possible, but it would involve changing how the AI was told to think. If you can do things like create a better pathing system (no more lings running into buildings, but attacking them) or tell the AI to favor making walls in certain situations, go for it.
Well I do suppose it's possible to tell the AI where good wall places are but the main problem would be to tell it when to wall and when not to. In this case, if you tell the AI to build a wall as soon as it knows the opponent is a Zerg, then the enemy Zerg would only need to go for fast expands as it knows the AI walled itself off. To prevent this, the AI would then need to be told how to scout and predict what build the enemy is going for ( Which will require slightly more complex codings )
I would like to see the AI use a risk reward analysis. Wall = fewer resources, lack of wall = potential to fast expand = more resources but less defense. If the enemy is zerg, it's probably not worth it to try to fast expand unless the AI does something like zealot rush to defend it.
Well that's nice to hear, instead of "can't be done" or "too complicated".
Now all I have to do is find where in this forum to learn how to script it, and learn what risk reward analysis is. Just kidding.
I'm not a programer but any terran or toss AI worth it's salt must be able to wall off against a zerg. Whether it is at the ramp or some type of simcity.
I think I will be able to add a Terran walloff strategy to the EagleAI. This is because the pathing options can tell if an area is completely walled off or not. However, the standard Protoss walloff with a Zealot as gatekeeper would be almost impossible. This is because there is no way to check if the Zealot is actually blocking the entrance or not.
I think I will be able to add a Terran walloff strategy to the EagleAI. This is because the pathing options can tell if an area is completely walled off or not. However, the standard Protoss walloff with a Zealot as gatekeeper would be almost impossible. This is because there is no way to check if the Zealot is actually blocking the entrance or not.
OK the answer to this it in two parts. If 'the pathing options' see that the wall is incomplete save for a 1x1 for toss then use the army size routine to see if there is a gatekeeper. I don't know the right words, that is the check to see if you stay and fight or run and live to fight another day. If no unit, zergling run by. If there is a unit bring up range units to take it out.
I have been trying to find out why Blizzard and other AIs do not wall off as terran or toss against zerg? Over at the B.net forums I got;
"You'd need the computer to analyze pathing and read the terrain. That can't be done in the Galaxy Editor. The coding would also need to work correctly for over dozens and dozens of different maps. That's why it wasn't done in the first place, pathing with choke detection was never coded into the game. Also, it's too complicated".
That same reply lead me to Mtops early video that seemed to just that (analyze pathing and read the terrain). I'm not a programer but all 2 player maps have 2 starting points, 4 player 4 and so on. This gives the zerg player a hatchery 6 drones and an ovie. The terran and toss get a CC or nexis and 6 workers at the start. Why can we not create points at the ramp or choke of the main that tell the AI where it can build a wall if the enemy is Zerg. Thus elemanating analyzing and reading terrain and giving the AI a where to build and leaving the what up to the AI. This would lead to a convention that all maps would not only have starting points but also wall points.
Discuss flame w/e:
I find it funny how your name is cliff and you're talking about walls. Anyway, it certainly sounds possible, but it would involve changing how the AI was told to think. If you can do things like create a better pathing system (no more lings running into buildings, but attacking them) or tell the AI to favor making walls in certain situations, go for it.
Well I do suppose it's possible to tell the AI where good wall places are but the main problem would be to tell it when to wall and when not to. In this case, if you tell the AI to build a wall as soon as it knows the opponent is a Zerg, then the enemy Zerg would only need to go for fast expands as it knows the AI walled itself off. To prevent this, the AI would then need to be told how to scout and predict what build the enemy is going for ( Which will require slightly more complex codings )
I would like to see the AI use a risk reward analysis. Wall = fewer resources, lack of wall = potential to fast expand = more resources but less defense. If the enemy is zerg, it's probably not worth it to try to fast expand unless the AI does something like zealot rush to defend it.
Well that's nice to hear, instead of "can't be done" or "too complicated".
Now all I have to do is find where in this forum to learn how to script it, and learn what risk reward analysis is. Just kidding.
I'm not a programer but any terran or toss AI worth it's salt must be able to wall off against a zerg. Whether it is at the ramp or some type of simcity.
I'm not a programmer either, but it seems very possible. I would love to see this acomplished too :)
I think I will be able to add a Terran walloff strategy to the EagleAI. This is because the pathing options can tell if an area is completely walled off or not. However, the standard Protoss walloff with a Zealot as gatekeeper would be almost impossible. This is because there is no way to check if the Zealot is actually blocking the entrance or not.
OK the answer to this it in two parts. If 'the pathing options' see that the wall is incomplete save for a 1x1 for toss then use the army size routine to see if there is a gatekeeper. I don't know the right words, that is the check to see if you stay and fight or run and live to fight another day. If no unit, zergling run by. If there is a unit bring up range units to take it out.
@Cliff747: Go
The pathing options can only tell if there is a still a hole in the wall. They can't tell how big the hole actually is.