The effects of troop AI changes

It is not just the main village, us BHB enthusiast have struggled with troop AI specifically the battle machine.

This seems like is has been the clash of clans version of Westworld or Terminator.

Quote:

---

Originally Posted by vikingGoalie

See my caveat, if they are not idiots it's a trivial change.

Supercell has coded this game in C++. That is an object oriented language. If they are following standard design principles and practices, most likely the troops classes are either implementing a common interface or inheriting from a super class basic behaviours.

Among which would be AI, next targeting. I'd actually be shocked if they didn't actually have an inherited function that returns the next target.
Of course there are nuances here where perhaps ground based, air based and so forth inherit differently. But my point is that in that language the way large systems (or any system for that matter) are typically designed/implemented you are not scatter shotting the same code all over the place.

You are in fact trying your level headed best to eliminate duplicate code as it makes maintenance a real chore. So while I can't *KNOW* without actually looking at the code base. Or putting on my hat of omnipotence. I can make a reasoned deduction that they are not complete idiots and have followed standard design patterns. If that is the case then ya, it really is not that big of a code change.

Especially if my theory is right about how the logic works on the next target selection.

---

You're making a ton of assumptions here. This game may have started off as a pet project by a couple devs who had no clue what they were doing. They may have written unmaintainable code while still managing to produce a usable end product.

That code could still be the base for the current game today, maybe parts or even large parts of it have been optimized but we really have no idea.

It's really difficult to take a project with bad code and start applying best practice to it without having to rewrite the whole thing. You may be able
to create new features in a fairly clean way, but old features will probably remain difficult or nearly impossible to change without rewriting large portions.

There is absolutely no way to tell and to assume that they have clean maintainable code utilizing popular or proven design patterns is again a huge leap.

So in theory it could be easy... but when does theory ever translate to practice in the real world???

Many many factors dictate an application's release cycle and rarely do we see "is the code clean and maintainable?" on the top of that list. Product owners are far more concerned about going to market than about having beautiful code.

In a perfect world your logic applies. In the real world *shrug*

so I did a war attack with my TH9 for the first time since the update... damn the AI is significantly different! I've been doing gobolalo for a looooong time on that account and have become fairly proficient with it - this war however? two one-star attacks. I thought I'd be able to compensate for the changes, but no such luck.

I do QW valk hits in low titans with my TH11, been doing that for months and months, definitely noticing the difference post-update. Really hard to control the queen's direction, and harder to get the valks into the core than before.

c'mon Darian... throw us a bone

Quote:

---

Originally Posted by NightsWatchMB

One thing that can be said with a fair degree of certainty is that there is an old version of the code laying around that wasn't prone to these ridiculous flaws in pathing that the new one is and that it would be a relatively simple process to reinstate it whilst they work out the bugs with the 'improved AI'

---

Correct. However for all we know they jumbled the new ai changes and a bunch of other changes into one big crappy release and cherry picking the AI changes out to revert them may be a huge bug filled task and leaving this horrible ai is the lesser of two evils for them.

Or the horrible ai is intentional. My guess is the later.

I think it's time to start stalking Darian... every time he posts we need to put a bunch of replies pointing to this thread. if we do that enough we might get an answer, or at least a comment

Quote:

---

Originally Posted by N1NJ4smurf

I think it's time to start stalking Darian... every time he posts we need to put a bunch of replies pointing to this thread. if we do that enough we might get an answer, or at least a comment

---

*I really like that girl, but she doesn't pay attention to me... maybe if I stalk her she will notice me and go out with me*

Does stalking someone typically end in a positive way? Or does it make things worse?

Quote:

---

Originally Posted by ClasherClash443

*I really like that girl, but she doesn't pay attention to me... maybe if I stalk her she will notice me and go out with me*

Does stalking someone typically end in a positive way? Or does it make things worse?

---

Married 15 yrs using the above mentioned method!

Quote:

---

Originally Posted by ClasherClash443

*I really like that girl, but she doesn't pay attention to me... maybe if I stalk her she will notice me and go out with me*

Does stalking someone typically end in a positive way? Or does it make things worse?

---

I don't want to go out with Darian, he's too hairy for me :D

I don't need him to like me, I just want him to respond to the damn thread. If SC's official line is "quit stalking me, I hate you and we're keeping the AI" then at least we know...

Quote:

---

Originally Posted by TheOldGods

Married 15 yrs using the above mentioned method!

---

stalking ftw? :saywhat:

Quote:

---

Originally Posted by vikingGoalie

See my caveat, if they are not idiots it's a trivial change.

Supercell has coded this game in C++. That is an object oriented language. If they are following standard design principles and practices, most likely the troops classes are either implementing a common interface or inheriting from a super class basic behaviours.

---

I'd be surprised if the different troops are classes at all. I'd have expected troop definitions to exist only as objects of a troop type class - with the objects populated from data file(s). If there are multiple troop AIs then the definition would include the selection of which AI class to use, and any parameters.

Which all goes back to what ClasherClash said... you're making an awful lot of assumptions.

Having said that, there's something in the pathing which is baffling. I had one today... I ran through the BhB tutorial on a gem farming account so I could get it to BH3 and clear obstacles for gems. So this is attacking an early BH2. I dropped an archer as indicated, and it targeted for some reason the arch tower:
http://twoseven.co.uk/coc/t2/archai.jpg

In case it was randomness, I dropped another in the same spot, and then another. And all 3 did exactly the same thing. It's almost the most distant target available whether measured as the crow flies, or measuring the actual walking distance around the wall.