AI attacking walls

[melia004]

I know this must have been suggested a million times already, but please, fix the AI so that troops do not get stuck attacking a wall while an enemy stands several spaces away. I have previously suggested making the jump spell (but not the fourth spell space) available to TH8s, as a way to counter this issue, but since that apparently isn't going to happen, please TRY and fix it so that my Barb King doesn't stand and attack a wall while the defending King is spaces away, killing my last few wizards.

[crushedhab]

I know this must have been suggested a million times already, but please, fix the AI so that troops do not get stuck attacking a wall while an enemy stands several spaces away. I have previously suggested making the jump spell (but not the fourth spell space) available to TH8s, as a way to counter this issue, but since that apparently isn't going to happen, please TRY and fix it so that my Barb King doesn't stand and attack a wall while the defending King is spaces away, killing my last few wizards.

Thats the whole point of the game to confuse the attacker. your inability to attack effectively is not a flaw in the AI

[melia004]

Your inability to understand my point is why I hate posting on these forums. Every time I have ever posted anything, there are always "Well you just suck" responses, which is your opinion. My frustration comes from being unable to plan an attack because I cannot tell when a unit is going to suddenly stop and attack a wall for half an hour. If you have suggestions that actually are helpful, by all means, please share them.

[bonesaw13]

Thats the whole point of the game to confuse the attacker. your inability to attack effectively is not a flaw in the AI

What? The point of the game is to confuse the attacker? I don't know if you fully understand the point of the game.

To the OP - I think we all have had problems with the AI. It actually got improved probably 6-8 months ago. Now the troops work together a little better. I would still like to see further improvements. Such as real time recognition. If troops are attacking a wall they should constantly be on the look out for other troops attacking near by walls as well and then work together. As it stands now troops will only start working together once the building they are going for is taken out, or a new pathway is opened up. But if the troops work together from the get go they could break down walls faster and work together more efficiently.

Not really sure how to correct this other than Supercell writing the code more effectively.

[Bwit51]

Fix this? What will the queen attack then? Lol

[edomingox]

I've seen this happen a lot. This is not happening from initial deployment of troops. This can occur well within 1/2 way or 3/4 of the way through an attack. Units will attack a wall to get at a target. But once the target is destroyed, the unit does not re-target until the wall is destroyed. Sometimes it does re-target. But I've seen it happen more times than not that it continues to beat down the same wall. Much like that real time movie they created that makes fun of the game that way, which was cool.

I think they should use triggers to know when to re-target or re-path the unit for whenever the target is destroyed or a wall is destroyed or the the jump spell is used. It may take more processing power and therefore slow the game down, but it should at least be tested if it's not already being done so. I'm sure it's evaluated when the target is destroyed or when a jump spell is used, so the big thing is evaluating whenever a wall is destroyed.

The barbs attacking the same wall even after the target is destroyed could be the result of the re-pathing still saying that it's best to go through THIS wall. But some kind of calculation should be made to determine if it's faster to go through the wall or go around it.

[DenyChristi]

Yeah, that is wasting my elixir

[WarrenJames]

Thats the whole point of the game to confuse the attacker. your inability to attack effectively is not a flaw in the AI

Obvious troll is obvious. Just ignore, or if you feel like doing something back, report the post to the mods.

Some trolls are cute at thier attempt at trying to anger others. For example, I've got a fanboy following that I've long since stopped reporting, since they come back like a weed after being banned. But they are my favorite forum weed.

---

Anyway, back to the topic...

I've addressed this several other times before on the forum, and the logical short answer is that Supercell has yet to find an AI ruleset that solves the issue, while not being too complex, as well as not causing other issues as well.

If it were easy to solve, or had more time to work on it, since they have to work on many other things we are totaly ignorant about, it would have been fixed by now. But it hasn't, and we are still talking about it. They have added in a retargeting check when a nearby wall gets destroyed sometime in the past, so that helps

So if they add in one check when a wall breaks, what is so hard about adding a periodic check every few seconds? For one, it isn't hard; and two, they didn't, despite what seems like an obvious fix. So obviously people at Supercell are stupid and inept became I see a fix that would solve thier problem for them. (*loud error buzzer* Wrong! Logic error!) This kind of thinking is common among humans, and can only sigh, then , and then explain why they are so horribly wrong. (note: I'm human as well, so that means I am prone to that kind of thinking too. It takes work to not fall into that kind of a trap in thinking)

Supercell staff are a rather bright group of people, given the huge balancing act they have to do with several different opposing game features. Plus, they have to have done something right, and kept doing it right, to be where they are today with thier successful, but still flawed (but improving) game. They have no incentive what so ever to keep any glitch in the game going longer than they can help it.

So given that bit of logic, there is more going on than what people would see on the surface. There is a good reason why they don't do constant retargeting checks, and only opted for the modest recheck when a wall breaks. Best guess I can make is that such constant checks will be too much server load than is worth. And if that is true, then that should give people an idea just how delicate AI programing is for as large a game CoC is.

I hope this helps explain the seeming illogic of the situation with some much needed logic. Knowing how the system works will help how to offer meaningful suggestions, than just rehash the same shot down ideas. Not that attempting to give them is bad in of itself, but starting at square one each and every time just does not help. I hope this brings people up to speed on things (and I damn well hope I'm correct on all this logical guessing ).

[edomingox]

How do they test for this? I would suggest they take an existing base that got attacked, use the data and show real time debugging information on what the specific unit is targeting and when it re-targets when the parameters are fullfilled (i.e. wall destroyed or target is destroyed).

It seems to me that whatever they did actually does work, just not every time. maybe there is a small bug in the formula that someone is overlooking.

making a "catch-all" check like a re-target check every 5 seconds is not good enough. Because during that 3 seconds it takes for the catch-all to trigger again, the barb is attacking the wall and is being noticed by the player, which in turn is reported.

I'm wondering if the calculation to determine if it's faster to go around the wall or through it is flawed by the amount of damage the unit needs to make to get through the wall? For example, it would be faster for 1 barb to go around a base rather than trying to go through a level 9 wall. Likewise, going through the same wall, it would be faster with 30 barbarians attacking it rather than to go around it.

[WarrenJames]

How do they test for this? I would suggest they take an existing base that got attacked, use the data and show real time debugging information on what the specific unit is targeting and when it re-targets when the parameters are fullfilled (i.e. wall destroyed or target is destroyed).

It seems to me that whatever they did actually does work, just not every time. maybe there is a small bug in the formula that someone is overlooking.

making a "catch-all" check like a re-target check every 5 seconds is not good enough. Because during that 3 seconds it takes for the catch-all to trigger again, the barb is attacking the wall and is being noticed by the player, which in turn is reported.

I'm wondering if the calculation to determine if it's faster to go around the wall or through it is flawed by the amount of damage the unit needs to make to get through the wall? For example, it would be faster for 1 barb to go around a base rather than trying to go through a level 9 wall. Likewise, going through the same wall, it would be faster with 30 barbarians attacking it rather than to go around it.

Whatever they do, I wouldn't be surprised if the Supercell programing team isn't already 10-20 steps ahead of us on this issue, given they have the source code, and can do stuff to it only game-code crackers wish they could do. Given the slow, deliberate, regular, batch upgrades they do, they are likely very thorough on fixing things, given the time allotted.

One thing I tend to notice is a 'thinking' pause as things retarget, so too many checks could stall things. Also it suggests a rather long game tick. Tick as in a clock tick. For each tick, a given command could be given to each unit, and each tick they follow that command, until something changes (dying included as 'something changes').

If ticks are long, then a check every few seconds could cause unneeded pauses in action, which would make such checks ineffective to performance. I honestly don't see evidence on checks being done when they attack walls, since they go about it for more than a few seconds at times while the rest of the troops go thier own way.

So the best fix would be some trigger condition that 'wakes them up' from targeting walls too long, but not cause problems elsewhere. Time based triggers are too arbitrary and could do more harm than good; so some event that gives a 'wakeup call' to the wall obsessed units would be ideal. Just... I don't know what would be a good one. And we can make a reasonably accurate guess Supercell has thought of ways too.