[Deprecated] Considering the removal of AI Progress.

[x4000]

Originally this was going to be a response to this thread: http://www.arcengames.com/forums/index.php/topic,19013.0.html

But I figured a bombshell like this deserved its own thread.

Want to hear something nuts?  I'm actually considering removing AIP in general.  I don't like that it's just one big bad number.  At one time I did, but I feel like that can lead to it feeling overly-gamey.  I think that it can also lead to players feeling overly restricted in what they can do in some cases.

I'll give you a comparable example: in GTA5, if you commit a crime even when there isn't a witness, you'll still get the police after you.  I HATE that.  I actually tended to play with a mod where the police were a) a lot more aggressive; b) more limited in LOS and so not on top of you immediately, giving you a chance to hide; c) only alerted at all if you did a crime in front of them, or someone calls them on the phone.

I have always thought of the AIP mechanic as being kinda-sorta like the stars in GTA, mostly going back to GTA 1 and 2 (5 did not exist when I created AIP, and I didn't like 3 or 4 or any of those).

For me, AIP has always represented the idea that "every action has a consequence."  But unfortunately as the game became more complex in expansions and so forth, it started to represent "every action has every consequence" instead.  The AI was gaining a lot of new abilities and verbs, and they all hung their shingle on AIP.  I don't like that.

If you'll notice in The Last Federation, we implemented something vaguely like AIP in the form of the attitudes and influence there, between the various races.  For one thing, I think that this is really nice that you can see why the number is what it is.  No matter what, I think that's worth doing here.  But even more than that, the fact that there was influence with each of 8 separate races, and often things would increase your influence with one at the expense of a rival worked out really well in practice as well as theory.

In AI War II, what I'd like to explore is a more nuanced system where the AI has various things that it is good at at any given time, or possibly those can be spun a few different ways.  At any rate, the AI dumps points into "logistics" for instance you do certain things, and so reinforcements start coming faster and maybe CPAs build quicker.  Who knows -- this isn't the real design yet.  But you could take OTHER actions that might harm logistics of the AI... yet cause them to pump a bunch of points into something else, like "aggression."  Obviously that isn't a skill, but rather a behavior style.  But you get the general idea -- with that, it might make them a lot more dangerous around your borders, for instance.

Everything was based around AIP before... from all the various forms of attacks, to tech advancement for the AI, to how fast they reinforce... etc.  Obviously there were some other factors in there like how long you'd been playing and how many planets you'd taken, but even those were both invisible and crude.  They were also UNIVERSAL, which I also don't like.   Why can't one pocket of the AI feel more aggressive than the rest of the galaxy based on how your past relations with it?  Etc.

On the one hand, this seems like it is getting a lot more complex, right?  AIP was good because it was simple.  I would counter that AIP was hard because it was simple: hard to balance, hard to extend, hard and intimidating to understand as a new player.  However, people more intuitively understand "actions have a consequence," since we're used to that in life in general.  I'd rather have a much more reactive AI that has many micro-adjustments to its behaviors in specific locations based on what you do, and that has larger macro-adjustments to what it does based on more major actions, AND that you can weaken in interesting ways by doing more than just "reducing AIP" via generic fashions.

We have done exactly this sort of situation before in The Last Federation, except it was modeled as many races and was both less and more complex than what I am thinking here.  The beauty of this sort of system applied here, designed in a middle-complexity fashion, is that it would provide more targets for players to do things against in interesting ways, a more alive-seeming AI, and more crazy stories about unusual edge cases (the most fun things).

[Mad Rubicant]

I'm... actually ok with this. Perhaps have an aggregate "AI Progress" from the various subscores that gives you an at-a-glance general idea of how dangerous the AI is (with the actual subscores on a tooltip over it).

[x4000]

Nice!  And yeah, I agree that at a glance numbers are helpful for a whoooole lot of reasons.

[Draco18s]

I am probably also ok with this.
One of the things I didn't like was that Exos were on a timer that you pretty much couldn't do anything about.  If you sat back and did nothing at all, even if the last exo died to your turrets and you lost nothing, you'd get another exo in exactly 2 hours 27 minutes and 18 seconds.*  I mean, yeah, you could poke the AI with a stick and accrue a little more AIP and that'd make it show up sooner, but that timer was still there in the background ticking away.

Whereas the shard collection was "ohshit, you got a shard, fuckduckfuck, send all the things!"

*Arbitrary numbers

[chemical_art]

I am fine with this because it allows a much more granular rather then all or nothing response from the AI. Different actions cause different reactions. This allows much more flexibility and future improvements on it as well.

It is imperative that this does not get complicated for its own sake however. What new values that arrive from this need to be easily understood at a glance.

[PokerChen]

Want to hear something nuts?  I'm actually considering removing AIP in general.

Okay. Not unexpected, in the context of you previously posting about wanting to make hacking less "number-y", when HaP is itself a mirror image of AIP. From the perspective of meta- game-design, I do feel that minimisation of spreadsheeting and unnecessary digits is a beneficial trend, in reaction to current gaming populations (clicker-heroes types excepted). XCom remakes, Civ 5s, Stellarises, all hunted for smallest meaningful numbers. It's also possible to represent threat without any numbers at all (visible to the player). You can conceivably abstract it into smiley faces (Space Pirate and Zombies), color palette shifts (red-blue please), and other graphical replacements.

Having a mind-map drawn would be handy, although forums isn't the easiest platform to share that kind of thing.

[kasnavada]

And... I'm the one breaking the mood.

This smells like SBR diplomacy all over again. I've got no issue removing AIP... but that kind of one of the central game mechanic of the game. Like you said, all of the AI's answers hung upon that. Like you said, it's simple to code, intuitive to understand, difficult to balance.

Yet. The real question ain't "remove or not", it's "is there a fun & intuitive & working concept that can replace it ?"
Problem is that it's probably not a good idea to discuss specifics with the forum.

[Tridus]

Want to hear something nuts?  I'm actually considering removing AIP in general.  I don't like that it's just one big bad number.  At one time I did, but I feel like that can lead to it feeling overly-gamey.

I want to say that I don't think this is a problem. It is gamey, but one of the key things about AI War is that we're not fighting an AI that's trying to act like a human (and failing). We're *deliberately* fighting an AI. It's okay if it's gamey, since we're trying to game the system. That's the whole point.

If you can legitimately come up with something that both works better and is just as simple to understand? Great! If you can't, but want to overcomplicate it because it's less gamey? Don't do it.

[tadrinth]

It sounds like you're more talking about breaking AIP up into multiple numbers than removing it. 

I think AIP is key to the 'guerilla warfare' feel of AI War, and I think that's one of the most unique things about the game. Definitely keep that as a strong gameplay option.

[Orelius]

I've always kind of disliked how AIP worked - too many resources at your disposal require you to take a permanent penalty (AIP) for a relatively short-term gain.  For instance, warheads, lots of capturable structures, etc.

I'm of the mind that permanent gains should be balanced by permanent penalties - for instance, capturing a planet is essentially a permanent gain if you can defend it, but if you can't in this moment, you can still do it later - you never 'lose' anything that cannot be replaced.  The opportunity to gain knowledge, resources, and more territory should obviously be balanced by increased AI aggression.  This is something I think should not change.  The larger your empire, the more the AI should throw at you.

Consequently, temporary gains should be offset by temporary losses.  It always sucks to have to feel that you need to 'resort' to a warhead, because while it may save your bacon for the next five minutes, it'll set you back for many many hours for the rest of the entire game.  As a result, it never ever feels 'good' to use warheads.  It's the same thing as with other disposable resources.  Often I feel like there's never a good time to use them, because of the permanent penalty that'll be imposed for a fairly ephemeral and minor gain. 

Similarly, capturables as a result never feel particularly 'good'.  Unless you manage to completely envelop your empire, all capturables you have are all always at risk, and they are all really difficult to defend because of the nature of the game.  Oftentimes you just lose things without noticing, encouraging savescumming because in most cases, all of your things are perfectly defendable, it just requires too much attention to deal with it.  In AI War 1, hacking was a decent way of handling this - you hack the fabs that you really need and can't afford to lose, and just use the others until they break.  However, for superweapons, there's never really a case where losing one is in any case justifiable, they should never die if you are playing right.  It's way too easy to accidentally lose a golem if you tab out of a fight for 15 seconds and didn't notice a mass driver or an AI artillery golem or something like that.

I think AI progress should still exist, just not really in the form we know now.  There should absolutely be a baseline AIP or AI aggression value based on the size of your empire - maybe make it dynamic so the AI lets up slightly if you have uncolonized worlds.  I'm just against the idea of paying permanent prices for temporary gains - especially since I like to play long and drawn out games.

Here's my idea:

-Keep AI Progress in some form, but make it constant and only based on permanent gains (destroying AI stations, colonizing, destroying critical AI structures and posts, perhaps even doing certain kinds of research)
-Make a 'temporary added AI progress' system for certain disposable and temporary resources like warheads.  Instead of making lightning warhead a '+1 AI progress when killed or destroyed', instead make it '+10 AI Progress for one hour when killed or destroyed', or something along those lines.  Then, with repeated use, this penalty gets harsher, perhaps adding more temporary progress or increasing the duration when.  This encourages use of these resources when you're in a pinch, but it doesn't make you wince because you'll have to pay for it the rest of the entire game.

[Vinco]

I really like where Orelius is going.  The ability to have temporary AI reactions is intriguing.  I'd also love the ability to localize these reactions somewhat.  Convince the AI to focus efforts on one front while I prepare some strikes elsewhere.

I feel that we need an overall escalation mechanic, but I'm certainly open to changes.  And making it more granular and more reactive to EXACTLY what the player is doing makes sense.

[Elestan]

The one comment I'll make is that it currently feels like the AI has a real blind spot to being neutered; I can go through the galaxy and basically wipe out every guard post and Sentinel, taking only enough systems to avoid Deepstriking, and the AI will barely notice.

I've wondered if it might be better to have Guard Posts trigger 1 AIP each on destruction, and reduce the AIP for Warp Gates to 3 and CSs to 5.  That would keep the AIP cost of the system roughly the same, but would make neutering less effective.

[kasnavada]

The one comment I'll make is that it currently feels like the AI has a real blind spot to being neutered; I can go through the galaxy and basically wipe out every guard post and Sentinel, taking only enough systems to avoid Deepstriking, and the AI will barely notice.

I've wondered if it might be better to have Guard Posts trigger 1 AIP each on destruction, and reduce the AIP for Warp Gates to 3 and CSs to 5.  That would keep the AIP cost of the system roughly the same, but would make neutering less effective.

And, following that idea, I felt that the AI floor concept, and basically "AI progress reduction" as a whole should have been destroyed. That was a huge blindspot in the AI too.

[Pumpkin]

I like the idea of replacing numbers with "organic" things (AIP, HaP, etc). But I want to make sure things get more visible for that, and not more cryptic and hidden in the engine (for instance, nothing tells players the distinction between threat and threatfleet).

In regard of "no magic move" (here), "agency against AI operations" (here) and the ideas around astro trains in the base game, I think there are many possibilities for a more organic and visible AI behavior/economy.

I won't develop my ideas here and now (I'll refine that on a text file first), but I'm generally okay with that design philosophy.

[Mánagarmr]

Splitting AIP up into multiple levels is pretty neat of an idea.