I'm little sick of AI cheats

Phoenix Point Game Feedback
121

Nobody said balance was easy. The thing is not about just making enemies weaker. Is about making WP balanced and not solve every problem by just make enemies stronger

If the only solution found was to make all pandorans inmune to MC by just giving them WP, then either it is not true that they tested it to death or they just lack imagination to make it right. Because in my opinion the solution they got is obviously wrong.

Yes, I reject the claim that the only possible balance is an unbalanced one.

Also It is incorrect to say, “well you cant have Pandorans not be much stronger than humans, it has been tested”. Again, this involves the behavior of the enemy AI, how much are they affected by losing an ally, how many WP they use and how, and other things. It is obvious for example, that if Pandorans did never use recover, they would be panicking regardless their WP, and if they used recover more often they would not panic even with less WP than now. Since they barely use the ability as it is.

In the end they created the mechanics, if they are not balanced there is a problem, and the solution is not to leave them more or less functional and ready to go.

I don’t think this is the only case where this happens, and too often the response to a possible exploit was to simply nerf the player, make the enemy more powerful, or remove the usefulness of an ability. Certainly, if something is mostly useless, probably is not OP. But aint fun either.

This has a lot to do, in my opinion, with aproaching difficulty with almost no change in the enemies your soldiers or the mechanics, for a very easy and very difficult game difficulty at the same time.

122

To clarify, panicking triggers at 0 WP, cost the whole turn but also automatically recovers WP, so the next turn after panicking he is automatically back in business.
Recover is the same but controlled by the character before it gets down to 0 WP but cost also a full turn.

It is basically the same with the difference that the first one triggers automatically and the latter by AI / player.

A panic loop is abusing the mechanic that every kill also decreases WP of all his buddies → they panic and are out of business for this turn → the player can kill the next bunch easy because the panicked ones didn’t act → next WP for all enemies decreases and so the auto-recover-by-panic WP are again away → rinse and repeat.

Now change panic with earlier recover and you get the same result … only faster because they get faster in the recover-do_nothing-stage …

They use it when they are close to 0, Sirens AFAIK even earlier.

123

The cost of panic is much higher than losing a turn. Is being exposed to be killed in a bad position. And maybe even more importantly, to lose a critical chance.

For example, crabs should never really need to use it in the midle of an engagement. As they life expectance is usually very short. For a crab a sensible strategy would be, recover to fill your WP before engage if they are below 10, and go for it.

For sirens, they have extremely high WP late game, so they will probably only panic when a player is trying to remove MC. An aropiate behabior for a Siren would maybe be, if your WP are below 10 or are infected with virus and predict to be below 5 next turn, seek cover and then recover unless inminent treat. If you have no enemies in sight and WP below 20, recover.

A competent AI would not panick by a small decrease in WP in a turn that could be easily predicted, but only by more relevant actions.

Edit: Sidetrackinging more the topic, seem to not have the mood to study.

1 Like
124

Ok, then lets play the game, you say below 10, from what maximum?

Lets assume 16 because you want to have it lower, I think it was around 16 before for midgame Arthrons.

Now I start the turn and kill 2 Tritons or 3 Arthrons and this way all other Arthrons are already down to this recover threshold of 10 WP and for this out of business for their turn. If I kill another foe then I will maybe also get all Tritons out of business for their turn. So with killing 3 or 4 enemies I’m there where I want to have them, most of them do nothing in their turn except recover and in my next turn I go on to repeat the same. Rinse and repeat until all are dead, the enemies didn’t do anything except permanently recovering.

This was the most abused tactic in the early builds except it needed more kills to get them down to panic.

Edit:
Of course this is simplified and it was also not always possible right in the first turn, but that doesn’t matter much because at a certain point you simply had all of them in a panic loop.
As you will also have all of them in a recover loop and it also doesn’t matter where they are, in the open or in cover, elsewhere, they just can’t do anything else then recovering the whole time as long as you keep the killing go on.

1 Like
125

Of course u want them to have a random component in each threshold and then maybe use a % of their max WP, as to not make all behaviors identical. So maybe not ten but 40% of max WP.

Also u want crabs to be crabs, so agresive behavior, if they are engaging they will not stop. So most of the time you will see crabs coming at you at full WP, and filling their WP before going for you.

Of course you want to be even more sophisticated, they need to evolve right? So early enemies (the ones with 9 WP now) you want to just panic and only include advanced behaviors latter. Pretty much no point in taking care of 10 WP.

That in theory if they use (as I suspect) simple behabioral trees, to model behabior. With branches of te tree assigned to specific limbs that may evolve.

And if course there is no point in theoretically discussing the AI, u need to see it in action.

126

I meant tested as in tested by the players, since release until July/August last year, IIRC.

Of course there were other changes as well.

But look, tbh, enemy morale panic lock (or whatever we call it) was a BIG issue, MC by Player is just not. Sure, if you try out a Priest and see that you can’t MC enemies right away, you might be disappointed. But if you are playing oriented towards taking advantage of MC you get around this; have someone disable the head first, use viral, kill off a few buddies, etc. Try the following: disable the head of a Siren with slasher arms; she will immediately lose 30 WP. Now MC her for between measly 4 - 8 WP, and tell me that experience is somehow lacking :slightly_smiling_face:

Is it possible to come up with a better system and implement it? Sure. Is it worth it? To enhance MC by Player and other uncertain returns, probably not.

I like the idea of different WP for enemies depending on different difficulty level, but I wouldn’t assume that even that is easy to implement. However, if we are getting to AI routines and behaviors that’s just massive amounts of resources to implement, and to operate because AI is a big drain on performance.

3 Likes
127

Not difficult to implement difficult to balance sure. But That is the nice things of having difficulty levels. U dont needto balance every one of them. U only need to get a reasonable progression of difficulty and really balance only the “normal” difficulty. As long as easier ones get progresively more easy and more difficult ones more difficult it will work.

Also no, at the level of AI we are talking about, it is neglectable. If you wanted to get mad, you could model evolutionary AI with genetic programing using battle metrics, and would be still nothing compared to the power of computation the graphics need.

1 Like
128

What I mean is rn probably there is no code to allow adjusting stats of enemies depending on difficulty. And if I had to guess the devs don’t know yet if they actually want/need it, because as new features are added, bugs are fixed, other features are changed it might all resolve itself in a different manner, or require a different solution.

I have no idea about AI in general, but in PP performance was the bottleneck to get it to the point where it is now. Idk how that compares to other cases or industry standards or whatever, but for PP AI performance is a huge issue.

129

Regardless where you put the threshold to recover, you will get them there as you have mostly only one real goal in the missions against Pandorans: kill’em all.

Killing them decreases WP for all other Pandorans on the map and it doesn’t matter at which specific threshold it happens for each of them, because it happens for sure that they have to recover. Maybe not all at the same moment but it will happen.

In PP there is no time or room for something like that. If they prepare (filling up their WP) for a turn and then want to engage the turn after almost all in their vicinity is over and out.

PP’s maps are small and not much room to rest an prepare for an engagement.

Even when this would be possible, as long as the player keeps killing the fools they will always stay in their prepare state until the players units are there where they hide … and still prepare.

130

I was actually uncertain if it was to all of the Pandorans in the Map, all the Pandorans in Visual Range or all the Pandorans within a certain range. All would make sense,

131

Do you have experience programming games?
In my expierience (not a lot, but a bit) AI is one of the hardest part of developing. The graphics are not really “programmed” these days, the game engine does that for the most part. You have to have good graphic artists who create beautiful objects and someone who has the logic of the engine under control. As far as I have noticed with my few attempts on the Unitiy Engine, the most difficult part is to implement the behaviour the opponents so that they look “clever” and also so that it doesn’t take too much computational power.

All of the map …

132

The (fig) original design goal was that the difficulty should come from the strategic/tactical layer, not enemy stats, so the enemies are always the “same”, they “just” get better versions faster (or counter the player in more “clever” ways, when mutating is according to what the player did), the player has less room for economic “errors”, etc.

Which is still true I believe? (Can’t remember right now if there is anything breaking it…)

2 Likes
133

Im actually certain thet that there is already some routine to recover WP if the enemy is iddleç, or symply a threshold reached

This for sure there is. You dont code constant values into routines, much less these kind of constant values. Pretty much Im certain the code is already there to read a table of values, you just need a midle men with a modifier depending on the difficulty, or just a larger parameter table.

The actual hard thing about difficulty is to make smarter enemies at harder settings. Which is done usually by cuting the search space in finite state problems like chess or disabling behabioral branches at easyer difficulties in actor-like AIs.

Probably, but if you code the AI, you just want them to minimise the turns lost in batle due to it, every other thing does not matter in this issue. AI is usually an optimization problem where you want a suboptimal solution. Most enemies are inactive till they locate you, or you locate them, them come for you and die 3 turns latter. Healing or recovering, offscreen is pretty much obvious behavior.

134

Maybe, but we don’t know it for sure. Except that they not code values into routines … i hope at least :wink:

There is also the evolution process that changes the enemies dynamically in different directions. At the end you have seen them all but how they evolve is different from run to run. I assume there is a bit more in the background as only some tables and reading values from them.

But basically I can agree, it SHOULD not be very difficult to implement something that modifies the values dependant on the difficulty.

I think it is a threshold, but pretty low for Arthrons and Tritons and somewhat higher for the higher Pandorans. I don’t think that they recover when they are not alarmed (idle) and not under their threshold. The only full WP enemies that I’m aware of in followed up turns are the ones that come as reinforcements from off the map.

But again, it doesn’t change much where they are and what they do as long as you go on killing them and this is simply what you (should) do in the most part of the missions.

1 Like
135

What I mean is that all the evolutions have predefined fixed stats. So once the type is selected the definition of such type is a constant table read from somewhere.

I actually also miss a bit of variety in the mission goals, it is made worse due to how much you do certain things in a campaign.

1 Like
136

Probably yes, at least all of them are also listed in the wiki, so it seems that it is working this way. But we still don’t know for sure what ever they do when they instantiate the enemies and if it would be easy to change this part or not.

And yes to more variation for mission goals, for sure :wink:

137

Mmm true, but lets assume its not stored anywhere and the stats of a evolution are obtained from some complex way.

Imagine once the process is done, you get that HP should be then 100.

Then you need to add, lets say if difficulty is leggend, then HP = 1.15 * HP, if difficulty is easy, HP = O.85 * HP.

Regardless, I can’t think on a scenario where that sort of change would be hard to pull.

1 Like
138

That, alongside level of enemy progression via mutations, would be best difference between levels. While there are currently several difficulty differences, they seem not to differ levels enough. Problem is mostly rookie seems to difficult for newbies, and ensuring enemies would less use MC, higher powers, or panic easier would do additional magic.

Since our squad has pool of max 20 WP, why not limiting enemy to e.g.
Rookie 10WP max
Veteran 15 WP max
Hero 20 WP max
Legend 30WP max or unlimited

Or, other way around - on rookie, letting our WP, strength … scales be unlimited (as much as gained). Strange thing is what to do with all SP left per soldier, there are some missions that add SP that can be used for all - but why not ability to transfer SP to another soldier (badly needed) at cost (e.g. 50% transferred)

1 Like
139

I think it would be better to transfer overgrown SP into the general Phoenix SP pool, even 10% would be better than just lost because no longer usable:

grafik
grafik536×536 93.6 KB

And this is only one of 5 that are full maxed out with much overhead.
I mean I don’t need more Phoenix SP in this run (as you can see I have enough) because I have only a minimal roster of soldiers (9) and they are all pretty maxed out, but on a “normal” campaign these SP are simply dead SP.

1 Like
140

One sudden question…

Dos this mean that non alerted Pandorans, receive a WP penalty when a fellow Pandoran dies outside their perception range and remain unalerted?.

Location: border patrol on pandoran lair, 15:30 pm.
Crabman 1: oh no… I suddenly… Its like this sudden unexpected depresing…
Crabman 2: Im feeling it too, like Im losing my will to live, but there is no reason for this… Is like this anxiety, like Im going to have a panic attack!.
Crabman 1: Im sure is nothing, lets finish this patrol, and buy some Nachos is gona be alright!
Crabman 2: Ok, but this time dont let them run away.

Does WP actually work like the force?