Page 1 of 10 123 ... LastLast
Results 1 to 10 of 92

Thread: [Tips & Tricks] Defensive Mechanics against New Wall-Breaker AI

  1. #1
    Forum Veteran Pingfao's Avatar
    Join Date
    Jan 2013
    Location
    Partying up in DTP
    Posts
    1,720

    Post [Tips & Tricks] Defensive Mechanics against New Wall-Breaker AI

    Foreword: Many AI enthusiasts, such as Hannibal, Junliang, Daddy, Flammy, and myself, have been actively testing various wall configurations to efficiently defend against the new revamped Wall-Breaker AI, since the release of Patch 3.54. I will continuously update this guide as more mechanics are discovered. Special thanks to everyone who contributed in these researches.
    _________________________________________________

    The new AI ignores most decoys as soon they are "open." Therefore, the idea of these mechanics is not to completely nullify Wall-Breakers' usage via decoys, but to maximize the numbers of Wall-Breakers required to breach the design. They are, more or less, more efficient ways to utilize walls for "boxes inside of boxes inside of boxes."

    X-Junctions: Avoid joining T-junctions to minimize the new blast radius' effectiveness; requires higher numbers of Wall-Breakers to "open up" individual compartments.


    (5)


    Avoid X-Junctions by creating staggered T-Junctions

    The next mechanic has 2 unique effects, depending on the Wall-Breakers' relative deployment locations. The idea is to create unnatural, lengthened paths, which separate the Wall-Breakers from the army, making them more vulnerable.

    Partially enclosed compartment: L-shaped wall strip that will attract Wall-Breakers when they are deployed away from the open gap. This also works as a funnel. (1)

    Maze strip: When Wall-Breakers are deployed close to the gap of the mentioned partially enclosed compartment, the horizontal strip acts as a maze to lengthen the Wall-Breakers' path, increases the chance of Wall-Breakers being sniped down before reaching the target walls.


    Partially enclosed compartment/Maze strip creating unnatural and lengthened Wall-Breaker paths

    Replay compilation demonstrating these mechanics:



    Wall Cushion - Double perimeter layer, which forces Wall-Breakers away from certain important compartments. (1)


    Double-layered walls force Wall-Breakers to go to thinner layers of other compartments


    Another variation of Wall Cushion

    Demonstrative replays showcasing clever applications of Wall Cushion:



    Decoy Island - Satellite enclosed compartment that attracts Wall-Breakers away from the main base; similar effect of "spikes." (2)


    Enclosed, individual building lures Wall-Breakers away from the main base

    Detailed discussion with demonstration of Decoy Island: http://youtu.be/6zVLG97dmd0 (2)

    Additional Mechanics - Coming soon.
    _________________________________________________

    Sources

    (1) Flammmy - http://youtu.be/jEjOSEN9Llo
    (2) Daddy - http://youtu.be/6zVLG97dmd0
    (3) Soki - http://youtu.be/FkZxey6RZVw
    (4) TimTheEnchanter - http://youtu.be/6gj5yI-lzTU
    (5) Junliang
    Last edited by Pingfao; April 29th, 2013 at 04:21 PM.

  2. #2
    Super Member Hannibal's Avatar
    Join Date
    Dec 2012
    Posts
    959
    Ping, that's a nice video. I agree completely with the T junction part. I think that's a particularly big deal right now when invading forces are going to be breaching outer layers quickly, and you want to slow their movement (via wbs or direct attack) through the base. And as your examples nicely illustrate, given the increased wb radius, they need to be spaced by two tiles rather than the weak T of a one tile offset.

    However, it's worth noticing that the wallbreakers display very different behavior towards your strip in each clip. In the first attack it essentially replicates the hole widening bug. It's hard for me to tell what they're targeting in the second attack as they die so quickly to your large (♥♥♥) bombs. It may be the hole widening bug again. But in the third their activity seems consistent with the path length theory. Same with the fourth where from a different spawn point yet they ignore the strip completely.

    Your video perfectly illustrates my problem with wb exploits right now -- they're based on buggy behavior. And the bugginess is dependent on where the wb is spawned from. Two attacks from the same side of the base may not produce the same bugginess depending on the square the wb is launched from.

    This presents two problems. First, defensively if you can't rely on the wb to act reproducibly to your tile pattern, it's hard to evaluate the value you're getting for each wall tile in the pattern. Secondly, I'm very much expecting the bugs to be fixed. At least, SC will both surprise and disappoint me if they don't.
    With apologies, I don't give out my Gamecenter ID or respond to PM solicitations to do base design critiques.

  3. #3
    Pro Member
    Join Date
    Jan 2013
    Posts
    568
    Quote Originally Posted by Hannibal View Post
    Ping, that's a nice video. I agree completely with the T junction part. I think that's a particularly big deal right now when invading forces are going to be breaching outer layers quickly, and you want to slow their movement (via wbs or direct attack) through the base. And as your examples nicely illustrate, given the increased wb radius, they need to be spaced by two tiles rather than the weak T of a one tile offset.

    However, it's worth noticing that the wallbreakers display very different behavior towards your strip in each clip. In the first attack it essentially replicates the hole widening bug. It's hard for me to tell what they're targeting in the second attack as they die so quickly to your large (♥♥♥) bombs. It may be the hole widening bug again. But in the third their activity seems consistent with the path length theory. Same with the fourth where from a different spawn point yet they ignore the strip completely.

    Your video perfectly illustrates my problem with wb exploits right now -- they're based on buggy behavior. And the bugginess is dependent on where the wb is spawned from. Two attacks from the same side of the base may not produce the same bugginess depending on the square the wb is launched from.

    This presents two problems. First, defensively if you can't rely on the wb to act reproducibly to your tile pattern, it's hard to evaluate the value you're getting for each wall tile in the pattern. Secondly, I'm very much expecting the bugs to be fixed. At least, SC will both surprise and disappoint me if they don't.
    I wouldn't hold your breath on bug fixing, but I agree with your assessment entirely.

    Prime examples of how the AI is not "fine". Consistent behavior would be nice.

  4. #4
    Forum Veteran Pingfao's Avatar
    Join Date
    Jan 2013
    Location
    Partying up in DTP
    Posts
    1,720
    Quote Originally Posted by Hannibal View Post
    Ping, that's a nice video. I agree completely with the T junction part. I think that's a particularly big deal right now when invading forces are going to be breaching outer layers quickly, and you want to slow their movement (via wbs or direct attack) through the base. And as your examples nicely illustrate, given the increased wb radius, they need to be spaced by two tiles rather than the weak T of a one tile offset.

    However, it's worth noticing that the wallbreakers display very different behavior towards your strip in each clip. In the first attack it essentially replicates the hole widening bug. It's hard for me to tell what they're targeting in the second attack as they die so quickly to your large (♥♥♥) bombs. It may be the hole widening bug again. But in the third their activity seems consistent with the path length theory. Same with the fourth where from a different spawn point yet they ignore the strip completely.

    Your video perfectly illustrates my problem with wb exploits right now -- they're based on buggy behavior. And the bugginess is dependent on where the wb is spawned from. Two attacks from the same side of the base may not produce the same bugginess depending on the square the wb is launched from.

    This presents two problems. First, defensively if you can't rely on the wb to act reproducibly to your tile pattern, it's hard to evaluate the value you're getting for each wall tile in the pattern. Secondly, I'm very much expecting the bugs to be fixed. At least, SC will both surprise and disappoint me if they don't.
    Thank you and great analysis, Hannibal. I did notice a lot of inconsistencies as well, especially in the first clip, where the horde of WB's stayed on the strip. I tried to rationalize their behavior using the presented information on the AI with no lucks.

    Here's my take on the other clips:

    In the 2nd clip, after breaching the Archer Tower compartment, they targeted the strip due to the long distance (around the strip to the right) from their spawn point to the closest "enclosed" building, which was the cannon. After that compartment was breached, they targeted the layer in front of the AD. Since they were spread out, they all went to different individual walls, whichever one was closest to each of them.

    The WB's in the 3rd clip were all over the place so I can't really pin-point any out too well. However, it seemed the attacker deployed his WB's in a horizontal swipe so each WB had its own relative closest building and path length, therefore splits off in all different directions.

    The 4th one was definitely strange. The Archer Tower to the right was definitely closer to the spawn point than the DE Drill, but they ignored it completely. Do you think it has something to do with the Archer Tower being under attack when the WB's were deployed?

    Also, I am pretty frustrated with "hole-widening" bug as well. Have had to pass up one too many easy loots just because my WB's wouldn't open up the inner chamber.
    Last edited by Pingfao; March 26th, 2013 at 12:34 AM.

  5. #5
    Senior Member Monferno01's Avatar
    Join Date
    Mar 2013
    Location
    Behind you
    Posts
    428
    thnx pekkachu

  6. #6
    Forum Superstar
    Join Date
    Feb 2013
    Location
    Drinking coffee with P.E.K.K.A.chu
    Posts
    3,931
    Nice, and how do you think this design will do?



  7. #7
    Forum Superstar Junliang's Avatar
    Join Date
    Jan 2013
    Location
    Country: Singapore Clan: Oldies
    Posts
    3,622
    *come here chop front row seat first*

    Good luck Pekkachu !

    *Back to working on my special video*

  8. #8
    Forum Veteran Pingfao's Avatar
    Join Date
    Jan 2013
    Location
    Partying up in DTP
    Posts
    1,720
    Quote Originally Posted by pinser View Post
    Nice, and how do you think this design will do?

    Wow that is an amazing design. Where did you find it?!

    Haha let me know how it works out vs. the new AI.

    Quote Originally Posted by Junliang View Post
    *come here chop front row seat first*

    Good luck Pekkachu !

    *Back to working on my special video*
    Better have "SPECIAL THANKS TO PEKKACHU FOR BEING MY ***TH SUB" at the end.

  9. #9
    Forum Superstar RegaliaRemix's Avatar
    Join Date
    Feb 2013
    Location
    Raiding your base, taking your precious Dark Elixir
    Posts
    3,457
    They seriously need to change the "Wallbreakers will attack the nearest wall" on the loading screen message because right now, it's not entirely true..... <_<
    Lightning my DE? I'll revenge, take all your resources and leave at 39%
    Farming TH9 till I get both level 30 heroes and all lava core. TH10 SUCKS!!!

    Do your part in improving the loot: Tell all new users and Android users to max collectors first so we can farm them.
    Gamecenter: RegaliaRemix

  10. #10
    ★★★ Champion
    Join Date
    Aug 2012
    Posts
    10,204
    nice thread! some great tactics listed also. the video quality makes it hard to watch for me but i get the idea.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •