Witch Doctor Fix - Cast your vote!

I think it should follow the pirate closest to it until that one dies. Then pick the nearest pirate again.

 

Does your incredibly experienced AI guy play the game on 500+ levels for test? )
Tell him that current WD AI brings him to death on forefront. Previous logic had a good manner of skipping units forward and walk behind. Now its gone.

 

Yeah considering the mass volume of negative comments, and my guild is near 100% in agreement, we must be playing a different game than the developers. The previous ai was far better than this. For what sounds like the overwhelming majority of us, the wd has been rendered largely useless. So please keep listening to the masses and the multiple threads on this topic. I have no doubt it's complicated, but the previous ai worked pretty dang well. Yes the lone gunner issue was frustrating, but that only rarely occurred for me. This change to fix that issue was over handed and upended the game for most of us. Straining for a gnat and swallowing a camel.

At the very least, give us an opt out, such as . . .

Or, let us toggle between old ai and new.

 

@Lynsey [Midoki], you mentioned in another thread that the WD AI programmed to protect the highest collective HP but how it is possible that instead of protecting 1jugg and 13gunners, it ran over to that 1jugg that is taking fire from a GP? Surely that 1jugg HP is lesser compared to the 1jugg and 13gunners which I deployed together with the WD.

 

Fair enough, I doubt anyone really think you are trying to break things but it is rather frustrating that some things remain broken for a very long time (e.g. replays) while changes that were not really needed, certainly not urgent, come first. It is then hard to figure out what is going on when you can't rely on the replays to tell you anyway.

I appreciate it is complicated. Much more so than most think, the above doesn't get close depending how complicated you make it. E.g. does WD work on remaining hit points, may decide to move even if nobody died, where should he stand, should ranged troops be preferred as, if he had any sense, could protect them while staying out of range himself (would be nice...). Does strategy change depending on whether he can resurrect. Others, should Brutes walk past cannons to attack towers (do troops change ai as they gain abilities), hope other troops can take care of it etc, etc, etc.....

The thing is, in game theory, what you do depends on opponents strategy, or what they are going to do. We had previously been told this was entirely deterministic, i.e. no decisions to be made, no random numbers - if this -> do that (this also raises the point again, why can't replays be fixed?). Even if defender behaviour changes this is the result of the attackers behaviour (e.g. activating Brute to force things to fire at him, other than that I don't see defences switching targets but even if they did that would be deterministic as well). If the WD is trying to second guess the attackers strategy (i.e. ability activation) then may be applicable, otherwise I don't really see how it applies. Minor events in one place (e.g. single gunner drops so the juggs miles away are now more important) is more akin to Mandelbrot and butterfly wing flapping than it is to Nash. In any case, other than guessing attackers moves a game should be entirely deterministic and optimal behaviour given current deployment is possible given a sensible utility function (this obviously involves winning, max loot, max stars, you don't get troops back casualties don't matter so would be in minimum time so can start recruiting again). This may not be possible in real time and player not behaving optimally but as someone suggested, try simulation. Given you likely have millions of actual attacks to backtest, could come up with something better than current behaviour.

For what it's worth I do think this is complicated, and I do think you get more flack than deserved but not fixing critical issues and an apparently repeated insufficiency of testing doesn't help you (simulation should do assuming you can replicate results...).

Lastly, I don't want players picking behaviour, this is thin end of the wedge and would lead to similar requests for other troops and almost certainly more problems. Obviously some others disagree.

 

+1 to the above, the multiple threads is crazy to keep track of.

Thanks Chris for responding. I'd like to add that our Guild is having the same horrible game play that others have mentioned here with the new witch doctor AI. With the old AI we all knew how he worked and only had a few instances of him wondering off or getting stuck. In my opinion he was never broken or needed any tweaks, the thought never crossed my mind that the WD needed to be fixed. He was dependable, he was predictable, he did his job and I used him as the corner stone to all my attacks.

By completely changing his AI, yes completely changing him, he might as well be a new character. I am at a complete loss and need to relearn how to play the game. That could just be me, however after reading through the threads and chatting with my guild mates, I am not alone in this.

This new AI change was hugely heavy handed and poorly thought out. Please fix this ASAP as currently the game play isn't much fun.

 

+1 (I voted for none of the above - but I almost agree with Sparkle)

I believe WD should protect in the follow priority:

1) Nearest (i.e. within protection radius of where he is dropped - only look progressively further if there are none)
if you dropped him with a lone gunmen you must have had a reason

2) Largest Group/Largest total damage (prevents running off with a single wayward troop leaving main force undefended)
priority 1 prevents him running half way across the island if there is a closer group

I think choosing between largest group & largest total damage is extremely difficult because:

• 1 HG is more useful than a group of anything (unless their collective damage is greater)
  
• BUT Gunners are more useful than Bombers
• Also total damage would consider differences in troop levels (if GS donations were stronger/weaker)

 

Thanks for the thanks. I too noticed the many WD AI threads from 9.0.1 bug to General discussion etc, and posted links in each to this thread, hoping to get some unified comment space. It seems there has been much positive conversation in less space now, that was part of the intention.

I must agree with several posters and say it is NOT as simple as any one of the options, and there are some subtleties in implementation.
Problem 1: with any Option if the WD hangs in the center of the group his health is not visible (tough to know if I should wait to revive or will he die) and he is vulnerable to Splash damage in the middle. In that way, the last WD AI was better as he hung back.

Option 2 by damage alone could present a problem of favoring bombs over guns as noted in my parenthesis, that was why I mentioned the bombs @Barbed. In doing so, the point is there needs to be additional programming than simply total damage.

Option 3 does not seem to be working as intended in that he is not protecting larger groups, which is what Lynsey indicated was one of the goals (to get him from chasing 1 pirate to an out island). I did read the Nash equilibrium article and understand triaging (learned triaging as an EMT in high school) and the concept of gain/cost of movement to a new group. I also understand that it must be programmed to the lowest common denominator of processing power. However, at the risk of sounding insulting, the current triaging is not working well. Some of the replays I've seen in guild are insane, and the WD is running around like a chicken with his head cut off protecting no troops as he keeps changing which moving group he wants to move to, or just charging into the heat of Jugg battle to certain death.

Finally, I agree with @thomssi that I also don't want players demanding personal specifics. The intention was to give Midoki a mass opinion of what the humans generally would like to see the WD do in a constructive way.
Personally I would favor largest group, WD hanging back out of center of that group, guns/HGs getting priority, and stick with current group until nearly decimated. Glad I don't have to program that.
I look forward to the next implementation.

 

Completely agree. I read the article too. A lot of comments are looking like they didn't care to...

 

I attempted to - but failed . Thanks to @Becker Redbeard & @thomssi for translating it to idiot-proof

 

Forty votes! Thanks for the input Forum community.
And thanks to this post, it'll be the most recent in category again.
If you haven't voiced your opinion, please do so: vote and contribute something positive!

 

I would 100% like the old WD Al. Now I have to find bases with cannons and defences clustered together so that I can deploy Brutes and Juggs at 1 end and the other guys at the other. That way, they won't overlap.

 

I must say, watching replays of attacks to my island with the new WD AI are pretty amusing. The WD's are dropping like flies. Some attacks use upwards of 4 WD's and all 4 met their demise pretty quick leaving a bunch of gunners and other small crew to get mowed down by my towers.

The more I think about it, I just can't wrap my brain around how this ever got passed testing. Something sooooo messed up like this had to be noticed by at least one tester

 

Well, defence has been easy but attack is HARD...

 

Based on past experiences with troops and upgrades. The Devs seem to know things we just haven't figured out yet. In their cubicles, pirate AI might work excellent as they know how to utilize the troops and their deployment far better than us. So might be an issue should they dumb down the AI to accommodate the majority of players (I fall in this category) or keep it as is for the those who 'get it' and hope others eventually catch on.

 

Lol. We are sharing replay after replay in our guild where the wd is running around like a chicken with its head cut off. The new ai has completely confused him. Doesn't take a trained eye to see that something is amiss.

 

Lol... I'm with you Daddy. My runs back and forth in one spot. He leaves a group to chase after one. Then the one he chases after dies, so he runs back to the group. Well, the group dies without his protection, so he goes running off in a third direction.

 

I posted this earlier. Makes me laugh. It seems to change targets as health points change during battle. So if a stronger set loses health, the wd switches to the next healthier set. And so forth.

 

@ Daddy. That was a good thing in your case. Kept him from running ahead of the troops and getting killed off right away. Which mine does almost every battle. Granted, there are some battles when he seems to work perfect. But those are few and far between, and I have not figured out why.

 

Just 80%ed a previously simple 100% of a ship wall base. Best I can tell, you have to let your front line troops die, like skellys, then send in the wd to resurrect the corpses as opposed to letting the wd protect the troops from dying. (Or use multiple wds now instead of the 1 needed under prior ai). Mind boggling. (Yes I hear Skye and raptelan screaming "No one is listening to us!" Lol-we still love you guys. ).