Fractal Softworks Forum

Please login or register.

Login with username, password and session length
Pages: 1 2 [3]

Author Topic: Stalling AI and CR is an unfun mechanic.  (Read 10788 times)

TaLaR

  • Admiral
  • *****
  • Posts: 2794
    • View Profile
Re: Stalling AI and CR is an unfun mechanic.
« Reply #30 on: October 24, 2018, 09:43:17 PM »

First I pitted three long range equipped Eagles, vs the Sim Aurora. Obviously not a fair fight. The Aurora enters their range, realizes it is outmatched, and starts to pull back. At this point the AI seems to get locked into a loop of indecisiveness - it never pulls back enough to vent and try again (it's definitely fast enough to, with the boost), nor does it try to outmaneuver them and take one out. It just sits there at the limits of their range, presumably thinking, "duhhhhhhhhhhhhhh" until it eventually is worn down and destroyed, having not fired a single shot.

This is a bug and should be fixed in next release - http://fractalsoftworks.com/forum/index.php?topic=13387.0
Logged

Thaago

  • Global Moderator
  • Admiral
  • *****
  • Posts: 7214
  • Harpoon Affectionado
    • View Profile
Re: Stalling AI and CR is an unfun mechanic.
« Reply #31 on: October 25, 2018, 12:43:38 PM »

In mid game fights I put hardened subsystems on frigates, but I usually start getting rid of frigates from my fleets at about the same time (as they start blowing up, though I'll keep a few really fast things around for special duty). Destroyers are usually fine without.

The game heavily rewards the player aggressively killing enemy ships; either hunting stragglers or breaking through deathballs.
Logged

Blaine

  • Lieutenant
  • **
  • Posts: 72
    • View Profile
Re: Stalling AI and CR is an unfun mechanic.
« Reply #32 on: October 26, 2018, 11:15:08 AM »

I've been following Starsector since before the name change, bought and played it extensively in 2014, played extensively in 2016; now, with the new update coming, the game has my attention again.

I've finally registered for the sole purpose of weighing in on this discussion, though I may participate more in the future.

Like Schwartz says, AI ignores or does not care about CR.  It will happily kite and deathball indefinitely if allowed.  If they hit zero CR, you have to kill them before your fleet decays to zero CR next.  Few times, I saw AI vs. AI fleets run out of CR then drift helplessly through space like Trumbles in Oolite.  Neither side could do anything.

This is the real key here: Of course the AI doesn't care about CR. The AI is there, essentially, for the player to fight and destroy (or not, but that is the primary interaction with AI ships). The AI doesn't care if it has fun playing the game, it doesn't care about the long-term welfare of its fleets, it doesn't have to buy ships, and it doesn't pay for supplies, repairs, or fuel. It can therefore afford to (and very likely will) burn itself out like a candle blazing at both ends in a single engagement. Sure, that single AI fleet may be crippled, but there will always be more AI fleets spawning in as the game proceeds.

The player's fleet isn't disposable. You must survive, you must maintain fleet functionality over an extended number of engagements, and unlike AI, every loss costs you dearly.

So in effect, although it may seem on the surface that every fleet has to play by the rules of CR, in practice they're a completely different set of rules for AI than they are for players.

I favor relaxed CR restrictions (or a rework) for players for this reason; and frankly, the notion that a combat ship can only operate for a few minutes before it seizes up is pretty silly and artificial. CR as it currently exists may help to prevent "cheese" tactics, but an artificial limitation isn't a proper fix for those cheese tactics, in my view.

Finally, I take some issue with the notion that high-skill soloing in a powerful, suped-up single ship is in any way actually cheesy. If anything, I prefer that possibility to the opposite extreme of "Bigger is always better."
Logged

Alex

  • Administrator
  • Admiral
  • *****
  • Posts: 24118
    • View Profile
Re: Stalling AI and CR is an unfun mechanic.
« Reply #33 on: October 26, 2018, 11:37:48 AM »

(Hi, and welcome to the forum!)


Finally, I take some issue with the notion that high-skill soloing in a powerful, suped-up single ship is in any way actually cheesy. If anything, I prefer that possibility to the opposite extreme of "Bigger is always better."

Just had to chime in here, because, well, even though I see where you're coming from here - what you're saying makes sense! - I think this is also almost exactly backwards when we look at the details :)

Infinite kiting is more time-consuming than skill-demanding. I mean, it certainly requires some baseline level of skill, but beyond that it just takes a lot of time and makes everything else not cost effective or riskier.

Peak times, on the other hand, encourage faster kills - which rewards skill *more* than "infinite kiting", since doing the damage with time constraints in play is a lot harder, and the faster you get the kills, the more you're rewarded (by having more time to handle other targets), which rewards skill still further.

It's not going to the other extreme of "more is always better". Rather, it's moving away from the extreme of "one ship is always better" towards the middle, and putting a greater emphasis on player skill in determining exactly how much is optimal.


That said, concerns re: the AI stalling unnecessarily are definitely valid, and I've already made a number of improvements on that front for 0.9a.
Logged

Blaine

  • Lieutenant
  • **
  • Posts: 72
    • View Profile
Re: Stalling AI and CR is an unfun mechanic.
« Reply #34 on: October 26, 2018, 12:28:49 PM »

(Hi, and welcome to the forum!)

Thank you!

Peak times, on the other hand, encourage faster kills - which rewards skill *more* than "infinite kiting", since doing the damage with time constraints in play is a lot harder, and the faster you get the kills, the more you're rewarded (by having more time to handle other targets), which rewards skill still further.

I'm very glad you've brought this up, because there's an important distinction to be made here. In my reading, I've noticed that you (and other board veterans) very often use terms like "reward more" when describing this mechanic—but in practice, what actually occurs is that the player is "punished less," i.e., they lose fewer supplies and less CR at the end of the engagement.

This may seem like splitting hairs or semantics, but it really isn't. In psychology, there's a distinction between positive reinforcement and negative reinforcement. Taking away more of a player's supplies and CR because they didn't finish an engagement as quickly as they could have isn't a "lessened reward," but is in fact an "increased punishment" and a textbook case of negative reinforcement. This works both ways: A player who finishes an engagement quickly is in fact working to minimize his punishment, not to maximize his reward. I know that "punishment" has certain negative connotations, but I'm (trying to) use it very neutrally here.

Psychology aside, this has practical implications as well, and it ties into my point about AI fleets being throwaway. During extended voyages, the player's goal is usually to lose as few supplies and as little CR as possible from one week to the next until next reaching a safe port. This makes sense, but as I've mentioned, the AI aren't playing by the same rules. Given all of the relevant factors, I believe that CR loss weighs a little too heavily against the player as currently implemented.

It's not going to the other extreme of "more is always better". Rather, it's moving away from the extreme of "one ship is always better" towards the middle, and putting a greater emphasis on player skill in determining exactly how much is optimal.

I don't necessarily disagree with you that there ought to be a balance or happy medium. In Escape Velocity Nova, for example, which I'm sure you're aware of and have probably played, a single Pirate Valkyrie Class IV with the right outfits could take on an entire small fleet of Federation carriers by kiting. Although fun in a storytelling RPG where story came first, I agree that it's a bit much for a crunchier game like Starsector.

Excellent work on this game, by the way. It looks beautiful and plays smoothly, and seems to have a lot more complexity under the hood compared to when I first played it and it was basically just a fleet vs. fleet brawl-a-thon.
« Last Edit: October 26, 2018, 12:30:34 PM by Blaine »
Logged

Alex

  • Administrator
  • Admiral
  • *****
  • Posts: 24118
    • View Profile
Re: Stalling AI and CR is an unfun mechanic.
« Reply #35 on: October 26, 2018, 01:17:55 PM »

I'm very glad you've brought this up, because there's an important distinction to be made here. In my reading, I've noticed that you (and other board veterans) very often use terms like "reward more" when describing this mechanic—but in practice, what actually occurs is that the player is "punished less," i.e., they lose fewer supplies and less CR at the end of the engagement.

This may seem like splitting hairs or semantics, but it really isn't. In psychology, there's a distinction between positive reinforcement and negative reinforcement. Taking away more of a player's supplies and CR because they didn't finish an engagement as quickly as they could have isn't a "lessened reward," but is in fact an "increased punishment" and a textbook case of negative reinforcement. This works both ways: A player who finishes an engagement quickly is in fact working to minimize his punishment, not to maximize his reward. I know that "punishment" has certain negative connotations, but I'm (trying to) use it very neutrally here.

Psychology aside, this has practical implications as well, and it ties into my point about AI fleets being throwaway. During extended voyages, the player's goal is usually to lose as few supplies and as little CR as possible from one week to the next until next reaching a safe port. This makes sense, but as I've mentioned, the AI aren't playing by the same rules. Given all of the relevant factors, I believe that CR loss weighs a little too heavily against the player as currently implemented.

So that's a very good point, and I definitely agree with you as far as the principle goes. It's also something I've been trying to be more aware of as a general design point.

But! I think there are actually two different things we're talking about here, as far as "reward" and "punishment". There's often not much direct reward from fighting, since - well, it *can* be profitable, but oftentimes you're fighting to get to something else profitable, such as salvage from a derelict, a bounty, or a trade fleet's cargo - so any sort of direct reward, there's just not much potential for.

So, basically, you can make the decision to spend less, if you're confident in your skill and your assessment of the engagement's difficulty. That's not quite a "reward", but it also doesn't feel quite like "lessened punishment", at least, not to the same degree that running out of CR in combat does. Where you do get into solid reward territory is when you're able to collect on more of the indirect consequences of fighting - such as bounties and salvage - more effectively, because you're making better use of the ships you've got. Whether it's because you're able to keep the overall CR higher, take on tougher enemies with a smaller force, or simply stay out on the fringes longer, you get to more *non-combat* rewards by doing better at combat.

Going back to in-combat, once you're out of peak time and CR starts ticking down, then I think we do get to the point of "lessened punishment", feel-wise. But that's already a "you were either desperate and didn't have more ships to use or you made a mistake and should have used more ships" state. So, yeah, you're avoiding punishment, but things have already gone wrong for you at that point, if that makes sense. It's not something you ever want to aim for, right? It's cheaper to deploy 3-4 frigates than to deploy one and have it tick down to 0. If you get to that point, you really *are* trying to cut your losses after things going bad. So if it feels like that, then that seems rather appropriate.


It's not going to the other extreme of "more is always better". Rather, it's moving away from the extreme of "one ship is always better" towards the middle, and putting a greater emphasis on player skill in determining exactly how much is optimal.

I don't necessarily disagree with you that there ought to be a balance or happy medium. In Escape Velocity Nova, for example, which I'm sure you're aware of and have probably played, a single Pirate Valkyrie Class IV with the right outfits could take on an entire small fleet of Federation carriers by kiting. Although fun in a storytelling RPG where story came first, I agree that it's a bit much for a crunchier game like Starsector.

(Am aware, of course, but haven't actually played it. Or may have played a demo at some point? Either way, not familiar enough to comment one way or another.)

Excellent work on this game, by the way. It looks beautiful and plays smoothly, and seems to have a lot more complexity under the hood compared to when I first played it and it was basically just a fleet vs. fleet brawl-a-thon.

Thank you! I'm glad you've been enjoying it.
Logged
Pages: 1 2 [3]