I’m really enjoying the panda AI these days. The other humans are still suicidal morons, but the pandas actually do some smart things now, and their initial deployments have gotten much better. I played a nigjt ambush today that had a great mix of enemy in smart initial positions. Good fun!
They also seem to be more keen on taking the player on, instead of just concentrating on their main haven objective.
Their target priority has certainly changed for the better for straight up engagement missions, however I wouldn’t say the AI itself is altogether ‘better’.
If anything their change in focus on your own soldiers make them even more easy to cheese in bottlenecks where the Pandorans will just blindly charge in to get slaughtered where as before such tactics wouldn’t work in certain mission scenarios as the Pandorans would concentrate on eliminating objective targets.
Started a new game last night which is the first time I’ve fired up the game in some months, and the standard early game tactic for dealing with Pandorans now is to pretty much just turtle up in one spot, set up smart overwatch placements and just watch the Pandorans burn all their action points mindlessly rushing in to get gunned down.
Those that get through the overwatch burn all their action points just to charge in that turn only to place themselves directly next to your units for easy clean-up your next turn.
Later as evolutions roll out the tactic pretty much sticks, except you’re now wanting more enclosed areas to cheese the Pandoran AI’s desire to just charge at you if they don’t have immediate LOS.
Are we even playing the same game? I find that the pandas avoid overwatch if they know that it’s there and they have a safer path to close on you. They move in cover down your flanks when able. Tritons especially make good use of cover, and the alpha snipers on night missions are really a pain to deal with now, as they intelligently park in themselves at long range to snipe from out of LOS.
The mix of troop types in each mission is also much improved.
Seems we are playing entirely different games.
Pandorans for me will just blindly charge into overwatch, the only time they don’t is when there are only a single one left who then sits back kind of just camping behind one particular set of cover.
Non-sniper variant Tritons in my engagements will typically charge forward to try get shots off even if it means rushing through 6 units of overwatch coverage to do so, pain chameleon forcibly overrides this by making them do a dash in a seemingly random direction including on several occasions parking itself right next to my heavy with a hell cannon.
Most of the ‘tactics’ I see from the Pandorans aren’t the AI itself but rather forced triggered abilities like pain chameleon or selective ability use like Sirens mind control. Even Arthrons with shields will often charge right at my turtling formation of overwatch and get slaughtered, any left frequently then charge forward still exhausting any chance to attack that turn and then position themselves right next to my units and deploy their shield facing a seemingly daft orientation that still exposes themselves to heavy damage dealers when if they had stopped the charge half that distance and deployed their shield facing at my squad they would have had full shield coverage from my entire squad and at least made me consider having to push some units to the side to deal with that enemy that next turn could easily finish their push and immediately engage, and in having to push to the side to deal with it potentially expose my units to other enemy fire solutions.
But if I simply stay at a distance and overwatch majority of Pandorans that don’t have extreme range capability will reliably just charge forward to their death all for the sake of getting a good firing solution on one particular target, which seems to part of the shortcomings of the AI decision making. It seems to be singular target orientated for decision making. It picks a target to attack or position itself relative to and disregards any other factors.
I can see it perhaps working better against certain playstyles that favour cheese tactics like dash spam and close quarter weapons to try clear as much as possible within a single unit turn as it’s often the case such approaches lead to somewhat isolated squad units and requires pushing into the enemy which makes their decision making pretty straight forward… target the suicidal guys charging at you from existing cover or with little movement adjustment needed… but if you hold back and let the AI do its thing to engage you on your terms from a overwatch position… well, the Pandorans seem to just want to die and end their miserable lives, they don’t really grasp the concept of defensive positioning, their thinking seems to go "Can I attack or position myself defensively to engage ‘Object A’? If not just charge to a tile location that will get me close to ‘Object A’. ".
Heck, one mission I had my squad inside the same one room of a building the entire mission just overwatching the doors and killing Pandorans as they mindlessly charged through them turn after turn, eventually there were just two rifle Tritons left who then decided to just camp the other side of some windows because according to the AI decision making there were no enemy to act upon. Gave them 3 turns of just overwatching and doing nothing else, and the two Tritons just walked between 4 titles back and forth on the other side of the glass, not overwatching for any potential push I might do or anything until I finally just shoot through the glass and took them down… something that was easily available to them if they didn’t want to just rush through the doors to immediate death like the 12 other Pandorans did.
Edit: I will say that the variations in Pandorans is much improved now and certain variants like the Sniper Tritons make for a nice change from the suicidal closer range zerg rush Pandoran behaviour, but they’re the exception rather than the rule.
TBH, I’m not sure if there has been any change to AI at all.
A couple of months ago I was investigating the claim that AI ‘cheats’ because it knows where exactly OW cones are and avoids them. So I experimented with saving the game and setting different OWs. The result of the experiment: not only is the AI oblivious to OW cones, but there is a random component to its decision making. Under exactly the same circumstances the AI will do very different things, and of course some of them are ‘smarter’ than others. So sometimes the AI avoids the OW, and of course it looks like it knew where it was and avoided it on purpose, but others it will charge blindly into it. The real explanation though is RNG.
I honestly see an improvement with the AI since 1.7, but perceptions can be misleading. Either way, I really like my perceived improvements. I’d be curious to hear what Snapshot have to say about the matter.
That’s not my experience at all. What difficulty do you play on?
The only pandas that rush in are the pure melee ones, and even they are making better choices now. The ranged combat pandas stay at range and try to get angles. They make better use of cover now.
Now the pandas have got “a little” better. At least I don’t see an Arthron or Tritron walking around a cover object in confusion and finally closing their movement completely open to be shot.
The Human AI is really bad. Here is an example from the special mission “The Pirate King”. My soldiers are all further down (south). The AI had a turn to position itself. Find the AI error in the picture
Yeah, humans are still suicidal morons.
Apparently the human AI is supposed to receive improvements with the upcoming (Polaris) update.
Source, please. There are any patch notes for Polaris?
Apparently, this update would be huge, i have that, but it doesn’t talk about AI.
And be sure to finish your actual campaigns, because the saves will not be compatibles with the new content!
That new armor surely looks a lot better! Eager to see the full changelog!