Resdesigning RCI

[Teal_Blue]

I don't know about fastest. there are things I don't do because of the effects. Case in point, giving spacefaring tech to everyone creates problems with the relationships with the races. So I get the first one, usually acutians, then I leave the others alone, just doing research and dispatches.

Now I am not certain this is the best approach. But I mention this because maskitymask said they do an action because it's faster.

I'm not certain faster is always the best. In fact I think it sometimes makes things harder.

anyway that's my rather newbie opinion.

Take care,
-Teal

[MaskityMask]

What activity are you referring to? .-.

Also, giving everyone spacefaring is safe on normal, until there are like 5 or so races. After that at least one race has to caught you in drones to reach dropzone

[PokerChen]

This is a weird thought, but what if instead of working according to a linear scale, each of the RCI elements worked sort of like a model of multicomponent failure? Lets take one RCI parameter just to work out the example.
A planet's Economy rating has 200 cells associated with it...

It's crazy, but I like it. Unfortunately, it also takes a lot of work to implement.

[NichG]

This is a weird thought, but what if instead of working according to a linear scale, each of the RCI elements worked sort of like a model of multicomponent failure? Lets take one RCI parameter just to work out the example.
A planet's Economy rating has 200 cells associated with it...

It's crazy, but I like it. Unfortunately, it also takes a lot of work to implement.

There's a quick and easy approximate version that you could use I think. Basically, lets say the intended applied change at neutral RCI is 'delta' and the current RCI value is 'R'. Then, the average new value of R should approximately be:

delta>0 : R' ~= R + delta*(100-R)/100
delta<0 : R' ~= R + delta*(100+R)/100

This is approximate because you're flipping a bunch of sectors all simultaneously, whereas you should really do it one at a time (e.g. repeat this process for each integer unit of delta). However since delta is usually less than 1 in TLF, that should be okay as long as you clamp R to the -100 to 100 range after the process.

This doesn't let you do the stuff like having specific special sectors and the like though.

[ptarth]

http://en.wikipedia.org/wiki/Sigmoid_function

[NichG]

http://en.wikipedia.org/wiki/Sigmoid_function

Its actually more complex than just running the RCI values through a mathematical function. If you just passed the current values through a sigmoid, you'd still have the problem that when the (internal) numbers are in the -10000 range, it takes an unreasonable amount of positive influence to get back to zero. With the particular method I described, its always easier to get back towards zero than it is to get further away from it, so in the presence of a completely random driving force the value will tend to be pinned near zero.

[PokerChen]

It's physically more like a hysteresis effect than a sigmoidal - still a mathematical function, but one with if statements.

[ptarth]

• The exact function isn't particularly relevant at this juncture, because there are more important features to consider first.
• However...
• Centering a Zero - preferable, but not necessary.
• Bounds are arbitrary, -100 to 100 is convenient, but not necessary. As long as we can make meaningful changes, the scale isn't important.
• Asymmetrical changes are... troublesome, especially to casual players who don't do lots of math in their spare time.
• Do we really want 5, 1 unit increases (+1 x5) to be different than 1, 5 unit increase (+5 x1)?
• Dear player, sorry about spending 10,000 credits to get that +20 RCI gain. If you hadn't made a side trip to increase it by +1 (it was cheap), you'd have a much larger gain.
• Do we really want a -5 unit increase followed by a -5 decrease to have different ending and starting points?
• Dear Player, because your increase made the planetary RCI score higher, the Burlusts found it easier to break their planet and now it is lower than you started with. On the "bright" side, if you had waited for them to act first, it would have been the other way.
• Do we really want the balance of RCI to return to zero?
• Dear Player, Thank for building the Ultimate Wonder (tm). Its effect will gradually diminish to 0 over the next 10 years. Depreciation hurts! Actually its because we can't afford cleaning robots, eventually well just have to bomb it from orbit and forget it was ever built.
• Do we really want softcap deterioration?
• Dear Player, Your contribution of 10,00 Doctors is wonderful. Thank you. This will increase our medical abilities by 150% of baseline in year one, decrease to 125% in year 2, and so forth, eventually stabilizing at 105% of baseline efficiency. The problem is that Doctors are so delicious, and well, we eat them. It's only the fittest that survive to contribute in the long term.

A little tongue in cheek, but I hope you can see why conservative action is called for before making massive mechanic formulation changes without a reasonable amount of thought (which includes math and simulation) plus was hilariously funny.

• Many of the RCI changes that have been made have focused on scaling, not on making it part of an underlying system that is interesting to interact with.
• The RCI mechanic is still too undefined.
• As I consider it more, the less certain I am that it is as central to the game as player's perceptions suggest.
• Previous patch history suggests that it did not originally play a central role, however after feedback, the devs have been increasing its effect.
• It annoys players because it is one of the few pieces of information we do have access to, but we don't have much control over it.
• At it stands, in my understanding it matters as thus:
• 0 to +infinity : minor bonuses, increase in relationships with our races, not very important.
• -X to 0 : minor penalties, not important
• -Y : Cheap federation entry, shortens game length, not that important.
• -infinity to -X : Allows for instant planet death, inconvenient if you wanted the race to live.


• I've been consider that tLF is not a 4x, it is an adventure game.
• The type of game wherein the shopkeeper sells you a $10,000 (and woe be it if you only have $9,999) sword to save his village from absolute destruction.
• The type of game wherein if you feed the eagle with the pie instead of the side of beef you lose, but it won't tell you for another few hours.
• The type of game wherein you can buy an infinite number of arrows from a single keeper, but he only has enough money to buy 5 back from you.
• That is to say, a series of contrived events that aren't very deep. In tLF we have a system that has more simulated randomness, but still very similar systems.
• We are playing the role of: Anti-suicide line hotline attendant for the solar system. We are trying to keep X races alive for the long term, but getting them in a Federation. While doing so, we have to prevent them from committing suicide, over and over again.
• Many of my ideas have required revamping the entire underlying system in tLF and replacing it with one that is more 4x-based. However, that's a ton of work. Especially since their is an expansion alpha coming out soon.
• I've been refraining from making many guided comments, because I'm a single perspective. However, looking at the comments, most of them focus on:
• Expressions of dissatisfaction with RCI in general.
• Complaints of the trivial impact of the player on RCI and the massive swings due to random effects.
• Confusion about other game mechanics.
• Discussion about our favorite function (go sigmoids!)
• Very little about what and how RCI should effect anything.
• To make a car analogy (since everything boils down to car analogies), we are trying to drive to the store, but keep arguing over what to listen to on the radio.

So, riddle me the following:
Assuming a -100 to 100 bound RCI with 0 being neutral.

• On the Economic RCI, what does a +50 mean?
• On the Economic RCI, what does a +10 increase mean?
• On the Economic RCI, what does a -50 mean?
• On the Economic RCI, what does a -10 decrease mean?
• If you have a Stock Exchange, how should that manifest in Economic RCI?
• Is that a continuous bonus, a one time bonus, or both?
• If your Stock Exchange is destroyed, how should that manifest in your Economic RCI?
• Is that a continuous penalty, a one time penalty, or both?

• Replace Economic and Stock Exchange with Medical and Hospital and answer them.
• Replace Economic and Stock Exchange with Environmental and Recycling Plant and answer them.
• Replace Economic and Stock Exchange with Public Order and Police Station and answer them.
• Where would Earth be at on the RCI scales?

[NichG]

• The exact function isn't particularly relevant at this juncture, because there are more important features to consider first.
• However...
• Centering a Zero - preferable, but not necessary.
• Bounds are arbitrary, -100 to 100 is convenient, but not necessary. As long as we can make meaningful changes, the scale isn't important.
• Asymmetrical changes are... troublesome, especially to casual players who don't do lots of math in their spare time.
• Do we really want 5, 1 unit increases (+1 x5) to be different than 1, 5 unit increase (+5 x1)?
• Dear player, sorry about spending 10,000 credits to get that +20 RCI gain. If you hadn't made a side trip to increase it by +1 (it was cheap), you'd have a much larger gain.
• Do we really want a -5 unit increase followed by a -5 decrease to have different ending and starting points?
• Dear Player, because your increase made the planetary RCI score higher, the Burlusts found it easier to break their planet and now it is lower than you started with. On the "bright" side, if you had waited for them to act first, it would have been the other way.
• Do we really want the balance of RCI to return to zero?
• Dear Player, Thank for building the Ultimate Wonder (tm). Its effect will gradually diminish to 0 over the next 10 years. Depreciation hurts! Actually its because we can't afford cleaning robots, eventually well just have to bomb it from orbit and forget it was ever built.
• Do we really want softcap deterioration?
• Dear Player, Your contribution of 10,00 Doctors is wonderful. Thank you. This will increase our medical abilities by 150% of baseline in year one, decrease to 125% in year 2, and so forth, eventually stabilizing at 105% of baseline efficiency. The problem is that Doctors are so delicious, and well, we eat them. It's only the fittest that survive to contribute in the long term.

A little tongue in cheek, but I hope you can see why conservative action is called for before making massive mechanic formulation changes without a reasonable amount of thought (which includes math and simulation) plus was hilariously funny.

This particular model doesn't actually do a lot of these things, and the others can be adjusted pretty easily. It is in fact like hysteresis - actually, in physics one would use the word 'spin' in place of 'sector' and basically you end up with a model of magnetism called the Ising model, so the mathematical properties of the thing are very well known. The version I proposed is what would be called the 'infinite temperature, infinite field strength' limit, where there's a first order phase transition in terms of applied biases (causing the aforementioned hysteresis), but in the absence of a bias it zeroes itself. Another possibility would be to do the same thing at a finite temperature just below the critical point, which would have a sort of bistable behavior where the planet 'locks in' to positive or negative RCI indices, and crossing zero is the most important event. The update function is a little more nontrivial in that case but I think I can probably work it out or at least an approximation with the same behavior.

Anyhow, it certainly doesn't do the overshooting phenomenon where by pumping the RCI up you make it easier to make it highly negative at any of these parameter ranges.

I've been consider that tLF is not a 4x, it is an adventure game.

Yes, I think this is a very important realization.

The type of game wherein if you feed the eagle with the pie instead of the side of beef you lose, but it won't tell you for another few hours.

This however I don't think is true. At least in my experience, its not too hard to forge a win even if you're pushing Peltian armada production against the Trifecta of Superiority. It may just take a very long time and a lot of tedious fighting off entire armadas.

Many of my ideas have required revamping the entire underlying system in tLF and replacing it with one that is more 4x-based. However, that's a ton of work. Especially since their is an expansion alpha coming out soon.

This I think plays against the strength of the game. Trying to make it into a 4x will be disappointing. Amping up the adventure game aspects so its more clear that that's what it is might be more useful.

[NichG]

Small update, the formulas with lock-in behavior should be:

delta>0: R' = R + delta*exp(a*R/100) (100-R)/200
delta<0: R' = R + delta*exp(-a*R/100) (100+R)/200

The parameter 'a' controls how strongly the system tends to lock itself into either a +RCI or -RCI phase.

[ptarth]

The list of properties was created to summarize important characteristics of all previously suggested formulas.
Your specific formula does have several of those properties.

By definition your formula does change the size of a gain or loss proportional to the current position. This is one of the features that was behind its suggestion in the first place.
It cannot return to its original starting point by taking the same sized "step".

> r.fun(0,20,1) = 10
> r.fun(10,-20,1) =  0.0467884

As R approaches the limits step size decreases.
> r.fun(0,20,1) = 10
> r.fun(10,20,1) = 19.94654

Likewise, a single big step and multiple little steps do not add up to the same sized step. It accelerates when moving towards 0 and decelerates when moving towards the boundaries. Except that this only occurs with multiple steps in the same direction.

It does have symmetrical step sizes so a step moves the same amount as long as the starting point is the same. (Given a fixed R, d or -d result in the same |R'-R|.)

It has a hard cap of 100.

Assuming R = 0, a = 2, delta = a series of random steps of -10 or +10, a simulation demonstrates that it takes ~175 steps for 99%+ of the simulations to reach the +90 or -90 marks, and then hover about that range until the maximum step number is reached. Reducing delta to -1 or 1 results in 82% of the population reaching the stability points of +90/-90 after 1000 trials. Random steps therefore push the function towards the boundaries, and not towards 0.

Again though, the actual formula is NOT important. If we want to dwell upon the formula of changing RCI and not what it does, then characterizing the properties of a desired formula is the first step. Once a set of desired properties is reached, then people can argue about the math.

[NichG]

The list of properties was created to summarize important characteristics of all previously suggested formulas.
Your specific formula does have several of those properties.

By definition your formula does change the size of a gain or loss proportional to the current position. This is one of the features that was behind its suggestion in the first place.
It cannot return to its original starting point by taking the same sized "step".

> r.fun(0,20,1) = 10
> r.fun(10,-20,1) =  0.0467884

As R approaches the limits step size decreases.
> r.fun(0,20,1) = 10
> r.fun(10,20,1) = 19.94654

These are intentional properties and ostensibly the main  reason why you'd use this system instead of a linear scale.

Likewise, a single big step and multiple little steps do not add up to the same sized step. It accelerates when moving towards 0 and decelerates when moving towards the boundaries. Except that this only occurs with multiple steps in the same direction.

This however is because of approximating the 'sector' system with a simultaneous formula. If you want to add 20, the right way to do it is to add 1 twenty times. Its basically approximating an integral with finite step sizes. Its not too computationally expensive to make that adjustment and improve the approximation.

The actual integral would be something like the solution of:

dR/dF = exp(a*R/100)(100-R)/200 for increasing R
dR/dF = exp(-a*R/100)(100+R)/200 for decreasing R

(F is the total forcing needed to achieve a given change of R)

The resulting integral doesn't have a closed-form solution afaik, but it can be approximated via 'stepping' in F.

It does have symmetrical step sizes so a step moves the same amount as long as the starting point is the same. (Given a fixed R, d or -d result in the same |R'-R|.)

It has a hard cap of 100.

Yes, the hard cap is an intentional aspect.

Assuming R = 0, a = 2, delta = a series of random steps of -10 or +10, a simulation demonstrates that it takes ~175 steps for 99%+ of the simulations to reach the +90 or -90 marks, and then hover about that range until the maximum step number is reached. Reducing delta to -1 or 1 results in 82% of the population reaching the stability points of +90/-90 after 1000 trials. Random steps therefore push the function towards the boundaries, and not towards 0.

Yes, because you're using non-zero 'a'. That's what that parameter does in that variation of the equations. In the physical system, having 'a' sufficiently large causes a transition from the unmagnetized state (average of <R> = 0) to a magnetized state (average of <R> is either +X or -X depending on which direction it randomly went first, where X scales with a).

If you go to a=0, it should push towards zero because of entropy effects. There's some critical 'a' where it switches over, where the system is maximally sensitive to perturbations (the 'critical point'). I think a_critical=1 in this case, incidentally.

Edit: Anyhow, as to the 'why' for these properties.

- A hard cap means that you have concrete endpoints to understand the scale. With a soft/non-cap, you have the problem that if you think -100 is really bad, you also have to conceive of -1000, -10000, etc. Which make -100 look not so bad, and so on. That makes it hard to develop an intuition about the numbers. With a hard cap you know -100 is the worst possible and +100 is the best possible.

- With a hard cap, the problem is that if its too easy to push things either intentionally or via random events then you get to the cap really quickly, which can make it feel like there isn't true variation. So having the system become less responsive to pushes as you approach the hard cap makes it so that you don't just drop and lock into -100 in the first few days of a plague - things can always get worse, but they get worse in a diminishing returns fashion. Diminishing returns also encourages diversity of approach.

- The existing system has the problem that it can get stuck out in the -10000 boonies and take forever to 'fix' back to zero, even if the player puts a lot of effort in. This is a consequence of an open, linear scale. Even if you transform that scale with a sigmoid, you have the issue. Essentially this comes down to a conflict between the idea of 'extreme events should make this aspect of the planet suck' and 'linear adjustments per time that make the planet suck in the short term will make it unrecoverable in the long-term'. So by making it easier to return towards zero (or towards some stable point) you resolve this issue.

- One feature of the a<a_critical model is that the resting point can be driven by persistent bias in the random walk. That is to say, if buildings are always adding a little bit or subtracting a little bit, it will pin the RCI values at some non-zero equilibrium average. Whereas in the open system, if the net change per time is positive then it will wander off towards +infinity and if the net change per time is negative then it will wander off towards -infinity.

- An added bonus: the 'sectors' imagery has a direct connection to the implied underlying structure of planets having regions/etc. That lets you do things like associate buildings with particular sectors and create higher-level organization and ties between the RCI concept and the other aspects of the simulation, in a way that's physically concrete (with a fair amount of dev work though).

[ptarth]

• We could argue all solar month about the math and what properties and formula are appropriate, and still not reach a conclusion.
• It isn't going to address the issue of what RCI does and how it is modified, it will only address how it changes.
• The major problem with the current system isn't the boundaries, it is that the uncontrollable changes are 2-5 orders of magnitude (depending on the source) more powerful than anything the player can do. Simply putting all effects into a more reasonable range will fix (enough of) the system for the majority of players.
• Thus, I propose a cease-fire.
• I further suggest a productive course would be to characterize the properties of RCI of interest (a discussion for which I created a reasonable question starting point above).

[NichG]

Alright, that's fair. I would say that RCI needs the following features:

- It must be able to respond quickly to bad situations, but have a permanent response which is less severe than its immediate response. E.g. if hurricanes start devastating the environment, the Environmental RCI should be able to capture that negative quickly in response to the onset of the catastrophe, but at the same time, ending the hurricanes should end a large chunk of the negative impact.

-- This may mean that in effect, one should decouple disasters from RCI influences entirely have have them perform explicit adjustments to planetary properties.

- RCI shouldn't drift in too much of an open-ended fashion, but should instead indicate a (relatively static) degree of advancement of that feature of the planet. That is to say, if the player does things like construct buildings, do dispatches, etc to adjust the RCI then it should relatively speaking stay where the player has put it, and fluctuations due to trends/disasters should be on top of that action. This keeps the player's manipulations relevant.

- Each of the RCI categories should have qualitatively different impacts for 'the red zone', 'the yellow zone', and 'the green zone' at the very least. The red zone is 'highly negative', the yellow zone is 'mild values', and the green zone is 'highly positive'. What these zones would be exactly depends on the math. An additional division could separate 'mildly negative' and 'mildly positive'.

-- Qualitative differences mean:

--- Different quests are offered

--- Species behavior changes in concrete ways that are explicitly identified for the player - for example, they become willing to start wars to capture better planets because they're dying on this one, or they buckle down and produce no ships but triple the rate at which they repair their economy for a solar year.

---- In order to achieve this, perhaps on planets where the player has a Diplomat or informant they can read off a 'current goal' for that species. Behind the simulation, the species is selecting actions in pursuit of that goal. When RCI tanks or spikes, it changes what goals are able to be chosen. Goals could be things like 'resolve internal instability', 'conquest', 'research', 'gather resources', 'construct infrastructure', etc. Part of doing this would be changing the way that the behind-the-scenes planetary choices are made so that a given planet can only focus at one particular thing at a time, rather than doing all of them all the time at slightly different rates. E.g. the planet needs to be in 'train troops' mode to increase its ground troops count, it doesn't happen passively.

--- Special one-off actions occur or become available. Playing on the 'adventure game' angle, having things like the Evuck planet destruction event (less severe) that are triggered by low or high RCI gives more experienced players particular cards to play in order to achieve strategic goals. Having things like 'will join the Federation if their Medicine is at -1000' is a good start actually, because it means that you can associate a strategic goal of getting them to join with the strategic choice of how to manipulate RCI while maintaining sufficient Influence to be able to push the option). Having a lot of the various inter-race options have requirements of 'RCI must be >X' would also help pull this off.

--- Disasters gain a grace period in which they can be stopped by raising particular RCI values by a deadline. For example, if Global Flooding is picked to occur, it spawns a notice. The player can prevent it by raising the planet's Environmental RCI above 0 within 4 solar months.

--- Some concept of 'building up reserves', 'self-development', etc should be introduced, aside from things like armada strength which is highly variable. Essentially, introduce a medium-term resource variable that is very visible to the player and whose dynamics are strongly tied to RCI. Have this resource variable be necessary or beneficial when taking short-timescale actions. For example, if the planets themselves also had 'credits' and could use them to purchase certain actions much like the player does - spend 5000 credits to delay an enemy invasion for a solar month, pay 10000 credits to generate 5 armadas instantly, etc. If the player were made aware of the credit supply and the actions the planet could take, it'd give them an extra idea of how to think about abstract strategic goals like 'make the Peltians strong'. Credit production and use would be tied to RCI - Economic RCI controls production rate, Public Order RCI controls whether it can be spent on troops, etc.

[ptarth]

• Effects should be both one-time effects and recurring effects.
• Buildings should provide a one time effect and then a smaller daily effect. Loss of the building will only remove a value equal to the one time increase.
• Some events should follow a similar pattern: e.g., planetary wide hurricane, -Y (can't do things when the hurricanes are blowing) & -1 per month for X months (The hurricanes continuously damage the infrastructure). After X months the event ends and the planet regains Y. The -X (from -1 per X) stays (permanent damage).
• Some events should just just have a monthly effect, e.g., Doctor shortage -50% of the gains from all medical facilities.
• Options should be available for the player to neutralize an event or terminate it earlier. Any recurring damage (e.g., destroyed infrastructure) remains.
• RCIs should have a natural cycle of RCI drift with period K and amplitude J. This reproduces the idea of periodic trends.
• Replace the single drifting RCI with a planetary focus, a particular RCI bonus that is recurring (e.g., the government is cracking down on poor education this year).
• Allow multiple events to occur simultaneously, only list the most devastating or most serious event on solar map.
• An RCI 0 point refers either to an arbitrary "average" or to a Racial and/or planetary "average".