Loading...
  OR  Zero-K Name:    Password:   
Title: 4hundred's Battle
Host: AUrank4hundred
Game version: Zero-K v1.5.11.1
Engine version: 104.0.1-75-g341c771
Started: 6 years ago
Duration: 47 minutes
Players: 1
Bots: True
Mission: False
Watch Replay Now
Manual download

Team 1

AUrank4hundred
Team 2

AI: Hard (1) <CircuitAIHard32>
AI: Brutal (2) <CircuitAIBrutal32>
AI: Brutal (1) <CircuitAIBrutal32>
Spectators

Show winners



Preview
Filter:    Player:  
Page of 2 (21 records)
sort

6 years ago
2 brutal + 1 hard: Titan Duel 2
+0 / -0


6 years ago
Cool now do it without abusing terraforming, which bots don't understand
+0 / -0


6 years ago
Make bots understand terraforming.
+5 / -0
Skasi
Just play vs Godde, rANDY and Drone or some other top-10-ers, then you should have the brutal 1 vs 3 you want. :P
(replace any of them with another good player if you just want the third one to be a hard player instead of a brutal one)
+0 / -0


6 years ago
quote:
Make bots understand terraforming.


But there is so much low hanging fruit that still needs picking..
+0 / -0
Skasi
6 years ago
Yeah. First they need to be taught about lobsters, clowns and pluking.
+0 / -0


6 years ago
I think bots understanding and countering terraforming is the lowest hanging fruit for improving the player experience.
+0 / -0


6 years ago
quote:
I think bots understanding and countering terraforming is the lowest hanging fruit for improving the player experience.


I think Circuit understanding the importance of high value targets, like a pene or a singu, over low priority targets, like a solar, would be a good start (currently it doesn't distinguish).

Also the ability to use bombers with some kind of intelligence, which would allow it to snipe most of its problem counters like trollcom, Behe, Singu.

Understanding terraform in the sense 'turret on top of a spike = more dangerous and longer range' would be helpful, but I don't think it will improve competitive performance as much as either of the above.
+0 / -0
The intermediate requirements for understanding how to use and counter terraform, however, would be brutal.
+0 / -0


6 years ago
USrank[GBC]1v0ry_k1ng I didn't say that terraform is the lowest hanging fruit for improving competitive performance. Circuit is good enough that I think all that the best players would notice an improvement in its ability to prioritize targets. I have watched somewhat new players play circuit and it gives the impression of responding to their actions. It sneaks raiders around their defense, it uses widow on their heavies.

I think some sort of handling of terraform would be a significant improvement because it would let people use terraforming without the niggling feeling that they are breaking the AI. Imagine if Circuit never made anti-air. It would be fairly easy to defeat this circuit with heavy gunships. Some players, especially new players who are still messing around, would not care that circuit is so weak to heavy gunships. However, once they get a bit of experience I think that players would wish that Circuit knew how to deal with aircraft because that would allow them to experience a richer air-game, one with counterplay. Singleplayer currently lacks access to higher level of the terraforming game.

I had fun playing coop Warcraft 2 by rushing walls (a multiplayer-only feature that AIs cannot deal with) and hiding behind them while building an army (or burrowing through trees). This was fun for a while, but, after doing this a few times it would have been more fun for the AI to actively be trying to break through out walls while we tried to defend them.
+0 / -0
quote:
I think some sort of handling of terraform would be a significant improvement because it would let people use terraforming without the niggling feeling that they are breaking the AI


What behaviours or features did you have in mind, triggered why what circumstance?

The simplest approach would be for Circuit to detect that it has eg. 2/3 of the map (handling already exists), and that enemy has more than x cost of terraform that Circuit has scouted.

Circuit would respond by rushing superweapon and nuking/DRP/stoning/bertha-ing turtle to death.
+0 / -0


6 years ago
quote:
more than x cost of terraform that Circuit has scouted

Requires game to somehow tell the AI which bits are terraform and which are weapon deformation damage. On some maps, small terraform may be sufficient to achieve great effects, which would avoid this detection. Cheap terraform can often achieve significant effects as-is.
+0 / -0
I think its going to be difficult and unrewarding to make AI understand terraform to any meaningful level. It doesn't even understand play on a flat map yet in many respects. I think a few work-around responses (detect turret on a spike -> consider it to have more range/greater threat, detect when player is doing a lot of terraform -> make superweapons) would probably be enough. If circuit could actively detect when terraform prevents a factories units reaching the enemy and stop building from that factory, it might even behave that way organically as it looks for somewhere else to spend its metal.
+0 / -0
The key to making AI know how to deal with terraform is teaching AI how to terraform.
AI should terraform actively, not just reactively.
This includes not caring about original map height, only about current map height.

Example for simple terraform logic:
If 2 points with a small aerial distance have a long pathfinding distance and you have the funds, add a terraform job to connect them directly.

Gather points using results from units that find themselves doing a longer-than-expected route. Limit the number of points per frame if necessary. Reduce priority of terra jobs that the enemy can fire on, etc.

quote:
Realizing that the terrain has changed...

Exactly the opposite of this, check the terrain continuously to see the best thing at the moment, the past is irrelevant
+0 / -0


6 years ago
Realizing that the terrain has changed and changing factory usage appropriately is a first step. All this would require is an occasional pathfinding poll similar to the terrain check it does at the start of the game. If it starts making spiders and superweapons then I think that is some level of successful terraform handling.

The next level is to make some countermeasures. Perhaps restore bad terrain or send Seismics.
+0 / -0
The next version of circuit should handle the situation seen in this game a lot better in any case. Until now it has not been adjusting group size to account for the threat of enemy statics, only enemy mobiles - soon it will react to porc by no longer sending small groups, only huge sledgehammer balls :3

quote:
Realizing that the terrain has changed and changing factory usage appropriately is a first step. All this would require is an occasional pathfinding poll similar to the terrain check it does at the start of the game. If it starts making spiders and superweapons then I think that is some level of successful terraform handling.


Sounds achievable.
+0 / -0


6 years ago
quote:
I think its going to be difficult and unrewarding to make AI understand terraform to any meaningful level.

Difficult? Yes. Unrewarding? Umad? :P
+1 / -0


6 years ago
So where is the new version of KGB with terraform feature? :3
+0 / -0

6 years ago
the terraform is only used to get a little distance on the gauss, which are used to survive imapalers (which they always do). only other terraform is to bury mexes and nanos, for the same reason. doing it without the terraform would not be a massive issue i think.

Note, i only won this one because they spammed crabes, i think they had like 10 in a clump running around.
+0 / -0


6 years ago
quote:
doing it without the terraform would not be a massive issue i think.


Would be interested to see!
+0 / -0
Page of 2 (21 records)