Poll

Is RCI working?

No, it needs major work.
10 (43.5%)
No, it needs some work.
7 (30.4%)
Mostly, just a little adjustment.
5 (21.7%)
Yes. It's fine.
1 (4.3%)

Total Members Voted: 23

Author Topic: Resdesigning RCI  (Read 7104 times)

Offline ptarth

  • Arcen Volunteer
  • Hero Member Mark III
  • *****
  • Posts: 1,163
  • I'm probably joking.
Re: Resdesigning RCI
« Reply #90 on: June 28, 2014, 05:13:22 AM »

Yes, this is basically what I want. The idea is that the player's ability to adjust the scenario should be roughly the same whether its game year 1 or game year 100. Otherwise what happens is that all the important game decisions are frozen in early on, and the late game is just playing out the 'destiny' that you set up during the short time-frame that the scenario was actually mutable, which is generally dull gameplay (lots of time but few actual decision branches, and if you screwed up early on you can in principle leave the game in an unwinnable state without being aware of it for a long time).

I'm comfortable with some degree of freezing-in like that, e.g. due to advances in a race's technology or the presence of additional buildings making it harder to shift things by a bit, but divergence towards infinity means that the game becomes arbitrarily unplayable the longer you wait. For example, if the first 20 years has RCI values that start around 0 so are nearly trivial to shift, but then the last 80 years has RCI values that hover around +50 or -50 and so require a (non-diverging) fixed amount of effort, then that's okay.

This divergence thing has shown up in the complaints other people have about RCI too. Its not just me being mathematically fixated. I've seen at least three posters on these various threads comment that in the late game the RCI values become really big or really negative and you basically can't shift them without sitting and doing dispatches for a year. This is actually something people are encountering and being turned off the game by.

  • I currently finish games in 10-20 years + whatever it takes to finalize things on x2. If an RCI only takes for a single year, the long term effect is... negligible. Oh now, fleet production is down to 10% of normal. That means instead of finishing in 3 years, it not takes 30 years. Oh wait, a year passed and RCI is back to normal. Total fleet construction time will be 3.9 years instead of 3 years.
  • Currently scaling is massively off for positive, negative, and player effects. I think you are getting conflating this with your divergence issues.
  • In my trend example above, 75% of the possible cases were in quite a reasonable zone. Even then, I think you are right and 0.10 per month is probably going to give a better fit (75% reduction of example).
  • I also think your divergence issue is examining only a component of the system in isolation and not taking into account the other parts.
    • Sources of variance in RCI
    • Player intervention
    • Trends (i.e., noise)
    • Positive Race Passives (buildings, technologies, Federation Pacts, etc)
    • Negative AND Positive Events
  • With sufficient forces pushing both ways that are roughly balanced, we should see a general equilibrium. We currently don't have that balance. Events are massively overwhelming everything else, and they are predominantly negative.

Quote
Take an exponential growth curve. If I want populations to be about 10 billion people by 100 years, and the game starts with populations of 500 million, then what growth rate should I choose on average? Well, I need the population to be multiplied by ~exp(3) over 100 years, so the time constant of growth should be 0.03. But now lets say the game goes 10% longer than I expected (110 years) - now I'm at 14 billion people, or 40% higher than my target. If game goes 30% longer than expected (130 years) I'm at 25 billion people, or 150% higher than my target. So thats pretty error-amplifying, and the same is true if I screw up the growth constant by similar percentages.

This is a bad example because population is limited to the population capacity of the planet. Trying to generalize to RCI or something else also fails because they are not exponential (rapidly accelerating) functions acting in isolation. However, we are both throwing up vague examples and words here when we need concrete models of what should be going on. It's too easy to misspeak and/or be misinterpreted.

Quote
Case 1, Case 2, Case 3

I need to think about this more. My kneejerk reaction is you keep bringing in baseline alterations and trends accelerating in particular directions and I keep fundamentally rejecting them. To frame the scenario in requiring them forces me to adhere to your view on how it works, which I still reject. I suppose then I would have to understand your system in entirety and then develop a NichG consistent argument and demonstration of failure. However, I believe your system is overly complicated, unintuitive, smells like dirty socks, and would require a lot of work to get to that point. In doing so I would also have to neglect my own model of the situation, which because this isn't my day job, ends up taking a ton of time. My goal today was to try to spend some time adding in the other effects to the base RCI trend model in the previous post, but I got nothing accomplished due to real obligations.

I'll hopefully have something more solid tomorrow. I've been basing it current in game buildings, technologies, etc, and it is obnoxious to dig this stuff out by hand through the game interface. Access to the source would make this so much easier. If you anyone knows or can get the Event data (likelihood, impact, duration, etc) that would be great.
Note: This post contains content that is meant to be whimsical. Any belittlement or trivialization of complex issues is only intended to lighten the mood and does not reflect upon the merit of those positions.

Offline NichG

  • Full Member
  • ***
  • Posts: 125
Re: Resdesigning RCI
« Reply #91 on: June 28, 2014, 09:15:40 AM »
    • I currently finish games in 10-20 years + whatever it takes to finalize things on x2. If an RCI only takes for a single year, the long term effect is... negligible. Oh now, fleet production is down to 10% of normal. That means instead of finishing in 3 years, it not takes 30 years. Oh wait, a year passed and RCI is back to normal. Total fleet construction time will be 3.9 years instead of 3 years.

    So make a disaster last for 3 years and then go away. Or even 6. That's pretty easy to adjust, I think.

    Anyhow, if you're finishing in 10-20 years and the devs are aiming at 100 years, thats exactly the kind of different that gets hugely hugely blown up by diverging mechanics. If they balance for 20 years, then 100 years will be unbalanced. If they balance for 100 years, then 20 will be unbalanced. If they mis-estimate how long the game takes, then the numbers can get wonky before the game ends - which is exactly what a lot of players are experiencing already.

    Quote
    • Currently scaling is massively off for positive, negative, and player effects. I think you are getting conflating this with your divergence issues.

    Well, what I'm saying is that divergences, aside from the problem they create for very long games, also make it much harder to get the scaling right because they're so much more uncontrolled. Essentially, the divergences make it hard to balance correctly, and I think we may be seeing the consequences of that underlying part of the engine in how the numerical ranges keep shifting so extremely from patch to patch.

    Essentially its like trying to build a house during an earthquake. You figure out how to space things, but then the ground keeps moving and invalidates your measurements. With something that is unstable, a change to one subsystem or one part of a subsystem can invalidate a lot of the careful balancing that had been done to the rest of the game. The end result is that balance develops more slowly and you have a lot more reversions where things that were okay become unbalanced because of a change somewhere else.

    Quote
    • I also think your divergence issue is examining only a component of the system in isolation and not taking into account the other parts.
      • Sources of variance in RCI
      • Trends (i.e., noise)
      • Negative AND Positive Events

    Random walk style variance (which is what the trends are) is a ~sqrt(t) divergence on its own, and it doesn't suppress a ~t divergence, which exists due to biasing factors like per-month RCI effects. I reordered your list to put all of these together because they're the same thing - noise is 'sources of variance' and sources of variance are noise. Random events, if they're unbiased around zero, are sources of noise as well (~sqrt(t) style). If they're biased - e.g. if there are more positive than negative events, or vice versa - then they're a source of ~t divergence.

    So all of these factors make the problem worse, not better (or rather, they're equivalent to the problem).

    Quote
    • Positive Race Passives (buildings, technologies, Federation Pacts, etc)
    [/lsit]

    This is a straight-out linear divergence.

    Quote
    • Player intervention

    This is the real wildcard, because it involves human behavior. I can't address this point easily with the same mathematics I can use for the purely simulated factors. The closest I can get is to try to analyze this as the dependent variable. That is to say, if the player has a specific goal (Peltian Medicine RCI > 50), how much effort is required to achieve that goal? And then assume that they will in fact put forth that effort.

    Using tools such as dispatches, the amount of effort scales the same way as the dominant divergence. If the player has divergent tools in their arsenal, then it can be different (and the mathematical analysis boils down to 'do the player's tools diverge faster than the ambient factors?'). However, giving the player divergent tools strongly emphasizes early game play over late game play (because what matters most is getting those tools into play early - the more time passes, the less net benefit you'll receive from them over the course of the game). I see that as a negative, because it can make the late game an irrelevant but still necessary task - e.g. you get the standard late game 4x ennui where you know you've won but you haven't conquered the 50 planets you need to in order to get the game to recognize your win.

    Quote
    • With sufficient forces pushing both ways that are roughly balanced, we should see a general equilibrium. We currently don't have that balance. Events are massively overwhelming everything else, and they are predominantly negative.

    That's not how the math works. If you remove all the bias that diverges as ~t (thats what the 'rough balance'  you're talking about does), you're still left with the sqrt(t) divergence from the random walk. And if you don't actually remove the bias perfectly, then if you wait long enough the ~t scaling will still end up dominating. E.g. if the random events are all perfectly balanced, but there are RCI improvement buildings.

    Quote
    This is a bad example because population is limited to the population capacity of the planet. Trying to generalize to RCI or something else also fails because they are not exponential (rapidly accelerating) functions acting in isolation. However, we are both throwing up vague examples and words here when we need concrete models of what should be going on. It's too easy to misspeak and/or be misinterpreted.

    Limiting the population to the population capacity is a nonlinearity of the sort that would actually improve RCI behavior and remove the divergences I'm complaining about. Simply capping RCI between -100 and +100 would actually remove a lot of problems. It would introduce some new problems (planetary RCIs being most often found at either +100 or -100), but ones which are fundamentally easier to solve.

    Quote
    I need to think about this more. My kneejerk reaction is you keep bringing in baseline alterations and trends accelerating in particular directions and I keep fundamentally rejecting them. To frame the scenario in requiring them forces me to adhere to your view on how it works, which I still reject. I suppose then I would have to understand your system in entirety and then develop a NichG consistent argument and demonstration of failure. However, I believe your system is overly complicated, unintuitive, smells like dirty socks, and would require a lot of work to get to that point. In doing so I would also have to neglect my own model of the situation, which because this isn't my day job, ends up taking a ton of time. My goal today was to try to spend some time adding in the other effects to the base RCI trend model in the previous post, but I got nothing accomplished due to real obligations.

    I'll hopefully have something more solid tomorrow. I've been basing it current in game buildings, technologies, etc, and it is obnoxious to dig this stuff out by hand through the game interface. Access to the source would make this so much easier. If you anyone knows or can get the Event data (likelihood, impact, duration, etc) that would be great.

    I kind of feel like telling you 'don't waste your time, which is unfortunate. If your results are going to depend sensitively on getting those exact numbers, I can tell you right now that they're not going to convince me of anything new because they don't address the fundamental underlying complaint that I have. The sorts of things you're talking about will change the structure of the jags in the noise, and may rescale the x or y axis of the plots, but none of those numbers will actually change the underlying instabilities that I'm talking about.

    As long as you don't address that behavior, its not going to be possible to show me something that satisfies my complaint. Let me put it another way. If you can't show me a model where I can start the game at the year 10 billion instead of the year 0, and still have the same gameplay experience, then we simply won't agree on it.

    Now, that leaves us with the question of what we should actually do to productively move forward. It may simply be that we have to accept that we won't be able to agree on how the RCI system in TLF should behave, or even what 'good gameplay' is. Its pretty clear that we fundamentally want the game to behave in different ways, and we're having quite a hard time explaining those ways to eachother. So we may have to make use of third parties in order to resolve the stalemate - that is to say, we design and put up polls that see what the broader TLF community actually wants, and then resolve going in to listen to that result.

    And of course, the devs are the final judges of what they want to do anyhow, so even if we're extremely happy with what we come up with, they may reject it for coding complexity or because it conflicts with the game that they want TLF to be (or the game that it should be given their knowledge of the broader player community)[/list]

    Offline Misery

    • Arcen Volunteer
    • Core Member Mark V
    • *****
    • Posts: 4,108
    Re: Resdesigning RCI
    « Reply #92 on: June 28, 2014, 08:55:59 PM »
    ....good grief, this topic has exploded.

    Just.... argh, the numbers.  They're like bees.   Big horrible swarms that I try to avoid at any cost.

    I'm not gonna bother pretending that I'm following what's going on here.

    Offline ptarth

    • Arcen Volunteer
    • Hero Member Mark III
    • *****
    • Posts: 1,163
    • I'm probably joking.
    Re: Resdesigning RCI
    « Reply #93 on: June 30, 2014, 06:00:37 PM »
    • I was able to spend some time modelling the problem.
    • re: wasting time
      • Quote
        I kind of feel like telling you 'don't waste your time, which is unfortunate.
      • I was trying to say: This is complicated. I have to think about it using numbers and simulations more rather than responding quickly.
      • In other words, it's not you, it's me.
    • I'm going to see if I can paraphrase the issue correctly.
    • The summary I came up with:
      • Ptarth - The problem is 50% A, 40% B, and 10% Other.
      • NichG - No. The problem is 50% C, 40% D, and 10% Other.
      • Where A = D and B = C.
        • Trust me, this kills in mathematical circles.
      • The speed and direction at which divergence occurs (and its magnitude) relative to the period of simulation is the issue at hand.
      • I am focusing on the speed (1) and direction (2) of changes relative to the player's ability to manipulate (3) RCI.
      • NichG is focusing on the eliminating the divergence (4), so that the speed and direction aren't that important.
      • I want to parse control of the various forces in the RCI System into adjustable components and keep each component reasonably face valid.
      • NichG wants to make a single RCI system without (as many) components.
      • At the end of the day, putting everything into a black box, there would be little, if any difference, between each of our visions.
    • So back to something useful.
      • RCI is a large mix of different ideas and abstractions. Because of this there are lots of conflicting ideas about it.
      • I believe that many forces should be divergent, because of my interest in a lower level of abstraction.
      • I believe that the player should have access to divergent mechanics and should be rewarded for thinking about.
      • I don't like the idea of an autocorrecting RCI because it implies that there are natural forces at work pushing RCI towards balance.
      • I do think we should have ways of correcting RCI, but it should require player or Racial AI actions, not be automatic.
      • Aftering thinking about this, I think that NichG's continued campaign for caps is possibly the way forward.
      • I don't think there should be a cap on negative RCI (or at least a very lenient one). There is always room to make a situation worse.
      • I do think there should be a cap on positive RCI. Making things better is sometimes hard or impossible.
        • My previous list of changes, still stands (for the most part).
        • Player and Racial AIs will have access to the previously described divergent positive and negative actions.
        • Add a positive RCI cap (race dependent). The cap increases by X when the race develops a technology or builds a building.
        • Growth over the cap is lost.
        • Loss of the building lowers the cap until it is rebuilt.
        • The Player (and perhaps races) also have access to a new line of buildings and technologies.
          • The I can't believe your Race can't take care of themselves better than this, so I'm going to have to do it for you technology tree.
          • A series of successive buildings that have a high positive impact, BUT only work when RCI is at a low level.
          • Each level requires an even lower level of RCI to function, but provides a correspondingly higher RCI increase.
          • This is effectively the same as a rescaling RCI monthly growth, but is more straight forward and requires player or Racial intervention.
            • The autorescaling happens regardless of choice and does not reward continued effort.
            • By giving RCI responsibilities to the player and Racial AI, the simulation is enriched.
            • If you know that the Acutians are going to eventually start poisoning their planet, you should preempt them by building Toxic-waste Clean up facilities. Then when they finally get around to it, your buildings will be able to quickly reverse the trend.
          • This rewards technology progressive races by allowing them fast access to more efficient RCI scores and the ability to recover from events as well.
          • Technologically inferior races will be stuck in the past with lower limits and less adaptability. Regardless of how much they wish, they haven't been able to "grasp" the new and improved ways.
        • This makes a adjustable hard cap on positive RCI, and a variable soft cap on negative RCI.
        • After X units of time, RCI will approach the positive hard cap, but be knocked down by events (which are primarily negative).

      So, how do you feel about this?
    Note: This post contains content that is meant to be whimsical. Any belittlement or trivialization of complex issues is only intended to lighten the mood and does not reflect upon the merit of those positions.

    Offline NichG

    • Full Member
    • ***
    • Posts: 125
    Re: Resdesigning RCI
    « Reply #94 on: July 02, 2014, 06:26:32 AM »
    I actually agree with parsing RCI into components, but I still disagree with the lack of some kind of cap. I think you can use parsing RCI into components to create a sort of soft cap that would satisfy my complaints. That is, you build RCI out of a number of components each which has a hard cap. The result is that as you 'finish off' components, there are fewer components left to optimize.

    Anyhow, we should probably spend some time trying out 1.027, since aspects of the situation may have changed.