Then it is clear to me AI is failing
Regarding allied troops, in the way troops are deployed, a “realistic good” AI would never engage. It is not there to be good, is there to waste turns of the enemy, to give you a chance to recruit more soldiers and for flavor. And it does all that fairly well.
Enemy AI makes many mistakes, yes. And from time to time makes really stupid ones (as in any game I know). But that is not often, in general is challenging, and is not easy to cheese. So again, I think it does its job more than good.
Well, because making an AI is hard. And because making an AI that tricks veteran players is specially hard, unless the rules of the game are extremely simplistic.
I dont think anybody is saying PP AI is excelent. But It is not below average at all. Even compared with the best games of the genre, PP is not suppar in AI.
The thing is that holding PP to a standard superior than for any other game of the same genre, is not a realistic expectation.
I suppose other games don’t have myrmidon firefighters 
Nobody said the AI did not commit mistakes. And there may even be some bugged behaviors in the game.
Showing a mistake in AI behavior just shows a mistake in the AI behavior. As all AIs do mistakes, one example is not really useful to evaluate the general quality of the AI.
Important questions are:
How many mistakes does it make? how often are they made? how easy is for a player to realize and take advantage of specific AI patterns to remove the challenge of the game? how often the AI gets stuck, breaks or is unable to handle maps or mechanics in the game?
And in the end the only real questions are:
is it fun? is it challenging? is it immersion breaking? how it does compare to other games?
With exception of utterly broken AIs, an example would the Civilization6 AI, where the AI does not include behaviors to handle many mechanics of the game, and I’m not saying handle well, but use. For the most part, most of this is subjective. And the only objective appreciation one can make about the AI is, how does it compare with other similar games? And PP does well in this comparison.
Does the AI have bugs that need correction? certainly, could benefit of improvements? of course. Is it bad? I dont think it is.
1 Like
I would like taking this convo with you as part of CC in Discord, showing to you why it is not about fixing specifics regarding AI, but a major issue preventing improvements
right now I could enumarate the long list of “corrections” that will never happen in this game,
if you have played a couple of campaigns you have watched plenty, from human fixation with healing to triton thug using melee attack 0 damage instead of SG point blank, enemies ignoring fire, and OW, the Pure soliders using jumps, demolition, disruptors instead of fighting…
Actually, I’m not talking about Civilians. In my experience, this is a fairly common behaviour of human soldiers ranged against me. They seek out cover and hunker down behind it - only to discover that they are facing the wrong way and my guys are behind them! 
Actually it’s not - it’s enemy human AI. Humans in general are obsessed with healing - to the extent that I’ve seen one regularly run out itno the open to heal his mildly scratched comrade, only to then be healed himself by that self-same comrade. This does 2 things:
- It puts both humans out in the open where they are sitting ducks, rather than hunkering down behind cover to heal themselves.
and - It takes 2 humans completely out of the combat for a full turn - in a game where you often only get one turn in which to strike, otherwise you’re toast.
While I get that the AI decision tree is complicated by PP’s realistic ballistics, it really shouldn’t be that hard to add a line of code that says: “If you have less than half HP heal yourself, otherwise look around: If a friend within 2APs has less than 1/4HP go and heal them, otherwise kill them! Kill them all!” Or if that’s really too difficult, then remove the line of code that makes them heal completely - because their obsession with healing simply makes them easy targets, rather than enabling them to survive the battle any longer than they do already.
I was only addresing the state and quality of the AI in the game. What you are talking about is a completely separated issue that Im really not that interested in.
First because I know less about the situation. Second and most important, because I dont think it is an useful debate anymore.
I don’t assume something will never happen because I don’t know.
The game is still in active development. The devs seem fairly engaged with the concerns of the community. Beyond that, discussing what I feel will be done or why is something that usually ends in a crossfire of assertions and accusations of bad faith and a lot bitterness. So I’m just not interested in that specific point.
This is a list of stuff. Each one of this is a separated specific issue:
-
“enemies ignoring fire” I dont think is a bug, it was very easy to abuse the previous fire system. But not anymore. I have currently no issue with this.
-
“human fixation with healing”: Maybe, but is also not that problematic. Some humans heal too soon, or when is not really convenient. But nothing that cannot be improved.
-
“and OW”; Im also not sure this is a bug at all. The probblems I have with OW are not related to enemy behavior. If OW could be used as a tool to prevent enemies to move, It would be much worse. Overall I dont have a problem with this.
-
“triton thug using melee attack 0 damage instead of SG point blank” This Im not actually sure what you mean here. I dont think this has been an issue for me. Thought I guess the reason may be that Triton thugs almost never get to melee me. And when they do is at a moment of the game they deal damage to me and I have no problem with that.
Oh, Yea I guess that can be annoying, I have to say I dont remembering that happen to me, at least not often.
You are right, I think we are playing two completely different games
Is because we are talking about our perception of the game. How much we think a behavior of the AI disrupts our experience with the game.
Overall, it affects you much more than it affects me. And I kind of only agree on the “human fixation with healing” but to me is not even that problematic. Meaning it does not break my immersion with the game, even if it is obvious that they heal too soon, and in moments when it is not the best decission. (Also happens with allied human units which i find more annoying).
But I guess that that is just because different people focusses in different things and plays in different ways. And we have probably also a different perception of what the difficulty of the game should be.
The AI of every game has weird quirks, but you appear to only focus on those. The AI does a lot of intelligent things as well, which you do not appear to value. For example, I like that the different panda species behave differently. Tritons and Sirens are more likely to make smart use of cover, whereas Arthrons are not big on subtlety.
1 Like
Sure thing, but the “weird quirks” are killing the fun. A Triton leaving cover just because using last AP or Siren using MC on a soldier without hiding after that action, just standing in the open, surrounded by PX operatives, that is not smart enough in my book
Like I said, I think the AI issues are a side issue. Phoenix Point has problems that are both much bigger and much easier to fix than occasional stupid behaviour by the AI.
The Siren doesn’t always have the AP left to hide. I find that they hide pretty consistently after MCing when they are able, and there is a clear place to hide. Same holds true for Tritons — I don’t see the reemergence from cover very often. In general, they play pretty smart. I’m sorry you find the glitches so deal-breaking. I think that the AI is, in general, pretty good.
1 Like
We haven’t counted that, but “occasional” is really light word. 
When I play I almost always expect enemy to do something stupid in any given turn. I’m not sure if it is occasional in the standard meaning of that word.
Just because you don’t pay enough attention to enemy AI it doesn’t mean that it doesn’t behave really stupid “occasionaly”.
Well some of us know that some things won’t happen. 
Yes sometimes it has good RNG of actions happening one after another. Shame it happens not so often.
I’m with you on the Sirens and totally disagree on the Tritons. I find that Tritons use their last fraction of an AP to step out of really good cover every single time!
It’s another one of those issues where I honestly don’t understand why an ‘IF Cover = TRUE & AP > 1 THEN Stop!’ command can’t be implemented.
1 Like
Same reason that Food Production Facilities take 53 days to break even, I suspect. The people who play the game and get annoyed with this stuff, and the people with the knowledge and position to change the code, aren’t the same people.
Same reason why you can’t make a levitating device out of your living room carpet by uttering incantations: it’s not the way things work IRL 
You are imagining the AI decision tree in terms that are familiar to you (the IF… THEN structure), where IF can happen at any time, because that’s the way we make decisions.
2 Likes
I’m sure you’re right 
You can tell I’m not a programmer. But it really shouldn’t be that difficult to add something to the decision tree that says: ‘You’re in cover and you don’t have enough MP to do anything useful, so stop now.’ Cos right now it’s following a decision tree that says: ‘You have half an AP left, so do something with it,’ and you can see it cycling through all of its possible options until it comes to the only one it can do - which is step out of cover.
I know some coding basics (I’m of the generation when this was starting to be taught in school, so as a reference think D in Computer Science As 
).
Anyway, about that Triton (and this also illustrates why AI in something like Firaxis Xcom looks much smarter):
The Triton does not “think” like the player, where you have an AP budget that you can spend continuously, but in terms of spending big chunks of the budget (if not all of it) in a combination of actions. So it will have something like “shoot twice”, or “move and shoot and move”, or “move and shoot”, etc. So when the Triton makes that final move out of cover, it’s not thinking. Most likely the idea of this action is that the Triton shoots and moves aways from the line of fire of the player (the assumption being that if the Triton can shoot at the player, the player can shoot at the Triton), and sometimes it works, but we don’t pay attention to those instances (“confirmation bias”, I think it’s called).
Why do it like this instead of something more imitative of what the player does, something more granular? My guess is that the architecture in place makes it prohibitive in terms of processing power.
For example, what is “cover” in PP? A reasonable evaluation of whether a position constitutes cover in PP is really difficult, even for the player. The AI would have to do a lot of calculations to determine what tiles are “in cover” and which ain’t, and this would have to be evaluated at the time of taking the decision to shoot and move, or move and shoot (you can’t make the AI “bear in mind” that staying in cover is preferable while it’s in the process of carrying an action).
And as I said before, I think you can also see how the comparison to Firaxis Xcom or similar games is not really fair, because there the player and the AI are limited to 2 discrete actions, positioning is very straightforward, the game has plenty of room to “cheat” with RNG (adjust chances to hit backstage depending on previous results), etc.
1 Like
Tritons stopping cold and reversing direction when they come under fire after stepping out of cover is not RNG — it’s good AI. That never happened in X-COM. Triton snipers hanging back where you can’t see them and always working to get angles on you is smart behavior. Moving so that they effectively use cover to intercept overwatch fire is smart behavior. Taking the longer but safer route to flank you is smart behavior.
Yes, the AI does do dumb things, especially for non-panda troops, but I really think y’all discount all of the good things the AI does because you’re laser-focused on the goofy stuff. Like Voland said, a granular game like this makes coding an AI much more difficult. I mean, the enemy turn takes long enough already
Just to show I’m not an uncritical fanboy: the Triton behavior I find most ridiculous is when they use their first action to spawn mist, but then leave it far behind. I mean, why bother?
1 Like