I shouldn't even be complaining about this, but

[igniteice]

In the picture below, the enemy has just used Zuul’Goth. It may not look like it, but indeed, Annihiliation was cast. Of course, my hero is Stealthed. So is Mother of Darkness. That leaves King Bloodwood and Enraged Kurandara.

It should be pretty obvious which target Zuul’Goth should target for an ability that reads “Kill an Enemy.” But apparently, the programmers didn’t bother to script it correctly, because the enemy AI targeted my Enraged Kurandara, which didn’t go over so well…

Why on earth would the AI try to kill a target that is immune? But it gets better… because the AI did it again moments later.

Here it’s less obvious that the enemy has just cast Zuul’Goth, but that’s because it matched several of the skulls that Zuul’Goth created. Unfortunately, the enemy’s front troop is also entangled… so… not only did it not kill anyone, it also did no damage with the skulls.

And I want to be perfectly clear here with my intentions with this post:

This is unacceptable levels of stupidity with AI programming. The developers and programmers who they listen to want us to set up defense teams for PvP and defense teams for Guild Wars, and this is the kind of AI they have playing our teams?

“I’ll never be able to kill that target, but that won’t stop me from trying, because I haven’t been programmed to do anything different.”

[Graeme]

Personally I would suggest that the devs spend exactly 0 hours trying to improve the AI’s decisions. Especially because GW is a thing.

They have far more important things to spend their time on like making new and interesting content that doesn’t just recycle old code… fixing bugs that have been around for months/years.

Oh wait…

/sarcasm.

[Shimrra]

If only this surprised me…

[ChunkyMono]

Back when Maw/Mercy/Infernal King was a thing, Maw would always seem to target my immune hero. That is if the ai even decided to cast Mercy instead of matching skulls with an entangled first troop. The ai has at least got a little smarter with that.

[igniteice]

Has it? Let’s see which of my troops the enemy Amarok targets in this clip below…

[Sytro]

Guess there might be a flaw in formula, which is used to target the troop →
From my observation, Zuul usually targets troops that are potentialy “highest damage dealers”.
As depending from current attack of my hero - it targets either hero or my Zuul. Other option that happens often - targets units with full mana → potential cast next turn.
That could be the case with most “kill on cast” effect (like devour).
In both situation it ignores unit traits.

[Fleg]

If the AI always made the best move then we would lose way more often. No thanks.

[FREDDY]

The AI is asinine thankfully; its programmers the same., unthankfully.

[SarcasticGimp]

Just shaking my head over here. There is no winning when it comes to the AI…1) Complaints when it goes right (goblin loops) or 2) Complaints when it goes wrong (targets EK).

I just take the good (thanking Karma) with the bad (cursing til the wife scolds me)

[igniteice]

@SarcasticGimp

1. Complaints when it goes right (goblin loops)

Looping is not a sign of good programming, at all. Looping is just the AI getting automatic matches and recasting an ability. In fact, it does that even when it would benefit it to STOP doing that. For instance, I’ve seen the AI repeatedly cast Thrall’s ability to destroy gems, even when Zuul’Goth is full on mana, and then instead of casting Zuul’Goth, it will just keeping casting Thrall until it doesn’t get an extra turn.

Or when Tinker Town Delve came out and the AI would repeatedly cast Detect-o-bot until it died. It would just keep looping itself over and over. The only thing that made it stop was the built in 15% destruction.

So no, looping is not a sign that the AI is doing something right. The opposite becomes very apparent during looping, when you notice that the AI doesn’t stop to look at possible moves on the board or abilities it could cast.

[SarcasticGimp]

@igniteice

Looping is not a sign of good programming

I do whole heartedly agree with this. This said, someone who’s running a Goblin defense team this week is going to think the AI was ‘right’ and did 100% what they expected when they check their GW defense score and sees 1/(insert # of losses). That 1 time the AI was ‘right’. It’s all the perspective, my friend.

[igniteice]

Here’s a common scenario. The AI just used Divine Ishbaala’s ability, “Divine Dance,” to transform red gems to skulls and green gems to yellow. Here’s the board after its cast:

1. It didn’t get any yellow matches. Not even a 3-match, because there were next to zero green gems to transform in the first place. But that didn’t stop it.
2. It didn’t get any skull matches. Not one.
3. It actually set up a 4-match of yellow for me.

Some players often respond to my criticism of the AI with, “Why would you want it to be more difficult?” And my answer is: I’m not asking for the game to be impossible, I’m asking for it to be enjoyable with a challenge. Here’s the final board before I kill the last enemy with a 115 damage Ultimate Oblivion cast.

Here’s what the AI is really good at: getting dreadfully easy kills in situations where it doesn’t have to be programmed to do much problem solving at all. For instance, my one troop that died this game was The Gray King. He died because the AI used Rope Dart, stripping him of 78 Armor, then dealing about half his health in damage, and because Rope Dart grants an extra turn, a single skull match was all it took to kill him.

That kind of programming requires nothing except: target enemy, match skull. That’s why Life and Death works so well. The AI doesn’t need to know the state of the board that its leaving after a cast, it just fires and forgets. That’s not bad programming, that’s a lack of it.

[ChunkyMono]

I just had a match, sorry no screen grab, where my Divine Ish was ready to cast, but I swear I did not see a match, let alone a free turn. Enemy Turnip had a board full of Green and Red, so I figured what the hell, at least I’ll deny them those colors. Bam, free turn.

Maybe your ai was just hoping for the best. Maybe it calculated three moves in advance, and determined it would still lose, and decided to throw the match.

It’s a free to play game, not Grandmaster’s Chess. I don’t expect greatness, or even basic competence. How about the game doesn’t crash or take 10 minutes to decide the “perfect move”.

[mld-81]

Exactly this. Some people should be careful what they wish for. If the coding was changed, we’d see an increase in moaners about how the AI is winning more.

[xolid99]

I agree with your posts regarding the AI making poor choices.

It does make some really bad ones, like ZG on Enraged Kurandara, but I believe that it is because it is (as others have mentioned), the perceived biggest threat.

A good example of the AI making odd choices is watching the behaviour of Spirit Fox.

Once you pass the turn it’s going to cast. I’ve recently been farming Maugrim Woods for souls on explore 9, and faced it quite a lot.

I’ve been running Gobtruffle, Arboreal Crystal or Mountain Crusher and 2 x Pharos Ra.

If I don’t match much mana, it will always without exception target my first Pharos Ra. Anytime I’m above 12 mana with it, it’s getting hit. The only time it has hit anything else is when Gobtruffle is at 15+ mana or Hero is full.

The problem with ZG/ER and Bloodwood imho is they are all Mythic troops with a strong threat level. Put a Roc in slot 4, high mana cost and I’d be surprised to see it targetted that often as opposed to ZG, but it’s the AI so who knows?

The AI however imho still works okay, despite the obvious issues we are aware of.

I find in Delves that Ancient Golem is ridiculous, in that it will stay on the same troop til it kills it often, even though it has random targetting. Same with Frost Lizard, my experience of that in Delve (because it’s one of the few times you encounter it at endgame) is that it will tend to stay on the same troop with it’s casts once it gets going.

I’m currently finding the Eldrazhor delve frustrating. It’s a little bit of a luck based Delve with some choices at least in when to cast your troops. Like Velenne when as many targets are webbed, or Rogue when you need to reduce attack on top troop, or try and subdue others.

Whatever decisions the AI makes are generally never fatalistic for it in Delve, with such limited margin for error for players.

We have Eldrazhor with no skull protection whatsoever. So any skull hits hurt. With the blocks in the board, the board shuffles a lot. I’ve counted the results of shuffles for the past 10 delves and the AI is 25 out of 28 in matching skulls from shuffles. It’s basically skulling me to defeat. Yet when I had all attack stripped at the boss vs the last troop, I could have won ten times over with skulls…

With the way this game works, sky skulls are all the AI needs to win in delve. I started counting the hits from sky skulls.

I just tried it this morning. Got to Goblins T5 in the middle and I’m all intact. I lost every single troop to sky skull damage on top of 1 Grapplepot cast. Mostly the reasons were Goblin King and Fizzbang changing the board state and sky skulls falling, but if I moved, I also brought sky skulls, nearly every time to align for AI. Even when I’m 3/4 of the board away and moving stuff, lines would fall.

Did the AI make any boneheaded decisions today and for the past 10 delves? I counted 1 miss skull trap they didn’t take. I was watching how they worked a lot more closely than normal, and this perceived threat or best choice argument seems to factor into their decision making. An example was a skull trap, 1 for me, 1 for them. When I and they had little mana, they skipped the skulls. When I had the same experience when I had to take skulls but still leave one 3 match and I was nearly full, they took the skull match as that nearly killed my troop. They also by taking that skull match, pulled a vertical sky skull match to kill my troop.

Did the AI know to do that, or was it just luck? Why didn’t it take the earlier skull match, and why does it cast when it could do more damage with skulls.

We’ll never know, but for delves at least, the AI already takes advantage of changing board state due to the disparity in level so much. It doesn’t need much brain to win.

I’d sooner they leave it. Well I’d like to see some info on sky skulls and their drop rate ratio, it’s the biggest leveller in delves, but that’s as likely as snow in Sahara.