Resdesigning RCI

[NichG]

I would actually redesign the buildings entirely to simply adjust the zero-point. So, for example, a building that would normally give +0.1/month to an RCI could instead just adjust the zero point by +10. That way you avoid the 'running off to +infinity or -infinity' effect that you have right now.

I don't think having a periodic cycle will help make things more clear to the player - its just another source of noise.

Having a kind of webpage-based model for these kinds of things might make it easier to actually understand how things would play out rather than the abstract discussion. I'm tempted to go and make a toy model just to showcase the various ideas and upload it somewhere.

Edit: And here it is, as a JSFiddle. http://jsfiddle.net/QtQ37/

Might be a bit buggy, but I think it should give a general idea of how the various trends would work.

[PokerChen]

Couple of quick observations.

My impression of an Earth equivalent of RCI values (capped between -100 and +100):

Economics:
-100: Pre-coinage barter economy. Almost everyone on subsistence.
-80: Great depression era in Germany Reich. Bad enough to burn money for warmth, but not absolute devastation.
-50: Crises of 1999 and 2007. Destruction caused by war.
-20: Africa average. Plagued by colonial history, but has positive outlook.
0: Global average.
20: OECD average.
50: 70s US boom. Golden age.
100: Unknown.

Environment:
-100: Extinction event from planetary catastrophe (KT-event).
-80: Ice-age level pressures, destruction of many biomes. Nuclear winter.
-40: Industrial revolution spread across the globe. Golden age where man "conquers" nature.
-20: Present day earth. Climate change cause regular disturbances, but damages limited to regional scales and rapidly reparable (in principle, given a space-faring civilisation).
0: Where we would like to be in 100 years.
+20: When greens would like to be in 100 years. Active habitat preservation and climate regulation.
+40:
+80: Early Jurassic.
+100: Cambrian explosion-level of biodioversity.

Medical:
-100: Pandemic in process of wiping out civilisation.
-80: Black death.
-40: Widespread epidemic.
-20: Global average before modern medicine.
0: Global Average with modern medicine. Medical research normal, but delivery is always hampered by ground reality.
20: OECD Average.
50: Pessimistic Futurist projection scenario, where we find cures for all common diseases but is unable to prevent death.
80: Average Futurist projection scenario, where we find cures for all common diseases, and is able to delay death.
100: Immortality discovered.

Public Order:
-100: Complete collapse of society.
-60: War-torn country beset by looting. Government has lost control and is ineffective.
-40: Status-quo of countries in civil war. Disruption of almost all civil services. Limits of unsuccesful martial law.
-20: Crime affects several sectors. Bribery is the norm, but society functions.
0: OECD average. Corruption/crime evident, but its influence somewhat mitigated.
20: Top of OECD. Corruption/crime evident, but is negligible. Also, a "successful" and "responsible" mafia-based public order who are able to stifle competition.
40: Limits of successful Martial Law.
50: Ideal democracy and other forms of government that still permit individual rights and freedoms.
100: Complete theoretical Utopia. (Andors start here)

 = = =
In the above scheme, buildings can have both a positive effect or a zero-adjust figure. However, it's formulated such that a well-managed planet should have a resting point of +20 maximum with most of the buildings in place, and a poorly-managed one with no building at about -20. Player actions and incitements can bring this to +-50, but only game-changing events are able to bring it to +-100 (with or without player help). On a normal game, you should only see a +-80 once in a century? Not sure on game versus reality at this point.

[ptarth]

Drive by comment
https://www.google.com/search?q=world+gdp+century&espv=2&source=lnms&tbm=isch&sa=X&ei=lu2iU-eyCs-HogTAgILYCA&ved=0CAYQ_AUoAQ&biw=1415&bih=756#imgdii=_

[PokerChen]

 Now there's one problem with a scale fixing zero to "average" and capping at 100. How would you account for the fact that medical advances absolutely improves over time, and that people do get richer on  average? The average has to scale up as you proceed forward by century, which contradicts the ability of e.g. universities, which simply lifts the average education up.

 Game vs reality.

 At least the technological singularity is loosely imitated by the time travel tech.

[GC13]

Now there's one problem with a scale fixing zero to "average" and capping at 100. How would you account for the fact that medical advances absolutely improves over time, and that people do get richer on  average? The average has to scale up as you proceed forward by century, which contradicts the ability of e.g. universities, which simply lifts the average education up.

This is how.

[PokerChen]

Now there's one problem with a scale fixing zero to "average" and capping at 100. How would you account for the fact that medical advances absolutely improves over time, and that people do get richer on  average? The average has to scale up as you proceed forward by century, which contradicts the ability of e.g. universities, which simply lifts the average education up.

This is how.

I should've clarified the problem as being conceptual, rather than mechanical.

 Conceptually, as soon as you stop funding a hospital, it stops working, whereas TLF Hospital comes with a permanent effect without needing to spend more after its construction. RCI buildings lifting anything above zero don't represent themselves so much as "a continued effort to dedicate attention to area X more than the average". On Earth, OECDs aren't above the global average because of the institutions they build, but more because of their dedication to certain policies (encouraging an environment in which such building are effective). E.g., Turkiye's mentality of "if you build it, they will come" hasn't had anywhere near the same impact as  China's "we fund you lots of money and build all this stuff".
 Effectively, many RCI-like buildings on Earth increase the efficiency of a race's spending on improving RCI, but has little effect on its own.

 It's just me being nitpicky. This also happens in the current RCI implementation. I would peg the starting equilibrium at -20 and force all races to spend a default, small percentage of its budget on maintaining 0. This also reflects the reality of a war economy slowly stripping away the societal fabric other than manufacturing - if a race chooses to spend 80% building ships and 20% expanding its war machine, it's going to lose out on other metrics. It's either this, or put in negative modifiers to certain racial actions.

[ptarth]

It's easier than that. Assume that it costs $20 to fund all government programs, regardless of number of efficiency. All Races then automatically spend that $20 on it, no more and no less. We then just drop the $20 cost from everyone, because it is constant across all races.

In other words, we assume that all governments are funding things at 100%. Recurring changes in RCI reflect efficiency changes caused by good and bad governmental policies. It is a little simplified, but we also don't get to manipulate the raw manufacturing or science capacities of Races.

We can then consider buildings to be enhancements of RCI efficiency. An example of medical RCI to illustrate the points.  At level 0 we have basic hospitals, they provide 0 +0 per day. At level 1 we have retrohospitals, they provide +10 and +.1 per day. At level 3 we have nano-hospitals, they provide +10 and + .1 per day. (It's culmulative).

The one-time bonus corresponds to the general efficiency brought about by an increase in practical applied technology. The recurring bonus reflecting ongoing improvements in medical technology by researchers in the institutions and by increasing the amount of medical trained personnel in the population.

As an aside I think we need to toss the current population system. We don't need kids or old age, they only make things awkward.
Give each race a birthrate, which is a function of racial fertility, planetary habitability, medical technology, maturation rate, lifespan, and so forth. Give each race a deathrate, which is a function of racial lifespan, medical technology, cultural values promoting risk-taking, personal violence levels, amount of orbital bombardments, and so forth. Subtract the deathrate from birthrate and that's the population growth per day. Population is this more akin to the number of fully functional adults of a race (children and old people represent fractions of an adult). This system makes balancing it easier, it also then allows for more straightforward applications of RCI and the eventual addition of population based production. The recent work that the devs put into the races won't be wasted, it will be part of the birthrate and deathrate coefficients.

As to continuous positive growth of RCI, this is one of the reasons I'm in favor of a sigmoidal (or logged) function for RCI. We create a system where getting "better" or "worse" is culmulatively more diffcult while not forcing arbitrary caps. It is also more reasonable. A planet can always "get better" and constraining them within an easily reachable boundary is a bit unsatisfying. If the boundary is unreachable, it is the same as being infinite. So my answer to the conundrum is, I don't see a problem with things always improving (with reasonable policies and technological innovation). That being said, I don't see a problem with allowing for massive crashes given reasonable justification either. If the Acutians bring themselves to the edge of ecological collapse, so be it. That should be an option, and there should be reasonable consequences. I also think that the Hydra should have access to anti-RCI dispatch missions where via cash or long term dispatch the Hydra poisons, slows, or cripples RCIs. Furthermore, I really think this should not be publicized to the other Races. That the Hydra has a publicist who feels it necessary to apprise every race of every action is unsatisfying. If I secretly arrange the Thoraxian Queen to accidentally be impaled by a small moon colliding with the surface of the planet, then I don't think I should then send out letters telling everyone about it. The same is true for all of those rescued pilots who somehow found themselves kidnapped from my ship and sold into slavery by the evil Peltians.

Note: NichG agree with this and is rebutting zharmad's previous post.

[NichG]

Realism doesn't always make for good game design. Before worrying about things like planetary budgets and government spending allocations, it makes sense to ask 'what does this do for the player's experience?' Also 'can we abstract this in a way that makes it clearer?' If there isn't something particular you're trying to create for the player's experience with the sorts of budgetary allocation things you're talking about, I'd say its fine to model things as abstract 'buildings' rather than get too into things.

Edit:

For birthrate and death rate I'd suggest going to a logistic growth model where the death rate scales nonlinearly with current population. That way you avoid the entire planetary population switching very suddenly from 'at the population cap' to 'extinct'. (They may already be using this, I don't know).

E.g. if you just have:

- dP/dt = (growth-death)*P ; if P>cap P=cap

Then P is basically pinned to P=cap or P=0, because the system is always exponentially approaching one of those two limits.

In a logistic model, you have something more like:

- dP/dt = growth*P - death*P*P

This means that the steady-state population at infinite time is P_cap = growth/death, and the system exponentially decays towards that particular population cap. This way, if you have something like a plague, the population doesn't shoot towards zero but instead the cap drops to something much lower but non-zero.

(One slight wrinkle is that the death rate is measured in different units than the growth rate in this model, so you want to introduce a normalization of the population variable P so that the average population cap for most worlds is P_cap ~= 1)

[NichG]

I don't know how we manage to fail to communicate so much, NichG. That's exactly what I was trying to suggest. (Followed by some caveats about other things.)

Sorry, I think I was unclear - I was responding to zharmad's suggestion there. I agree with your post.

[ptarth]

(Removed my old post to try to keep things clean.)

I'm not sure I follow, it is late though.

I don't see the problem with the population rate of change.

Examples that illustrate my confusion.
Skylxians
Birthrate 2%
Deathrate 1%
Normal Growth 1%
Growth after a 200% deathrate increasing plague  (2% - 1%*2) = 0%
Growth after a 50% drop in fertility (2% * .5 - 1%) = 0%
Growth after a 100% increase in lifespan (2% * 1%*.5) = 1.5%
Growth after a 500% deathrate increasing plague (2% - 1%*5)=-3%
Growth after orbital bombardment kills 10,000 per day (2% - 1% - 10,000/current_population) = 1% - 10,000

That all seems to work. Your concern I believe was nested in the 500% deathrate case, population will be  current_population*0.97^t after t turns. That would be 23 turns to 50% population, from a disease that kills up to 5% of the population per turn (congratulations, no antibody immunity here). Alternatively, you could break deathrate down into its various components, something like:

accident rate: x
old age rate: y
disease rate: z

deathrate = x + y + z, in which case this disease would be: x + y + z*5, etc.

At this time however, I believe the simplified version keeps everything abstract and pretty reasonable.

[NichG]

(Removed my old post to try to keep things clean.)

I'm not sure I follow, it is late though.

I don't see the problem with the population rate of change.

Examples that illustrate my confusion.
Skylxians
Birthrate 2%
Deathrate 1%
Normal Growth 1%
Growth after a 200% deathrate increasing plague  (2% - 1%*2) = 0%
Growth after a 50% drop in fertility (2% * .5 - 1%) = 0%
Growth after a 100% increase in lifespan (2% * 1%*.5) = 1.5%
Growth after a 500% deathrate increasing plague (2% - 1%*5)=-3%
Growth after orbital bombardment kills 10,000 per day (2% - 1% - 10,000/current_population) = 1% - 10,000

That all seems to work. Your concern I believe was nested in the 500% deathrate case, population will be  current_population*0.97^t after t turns. That would be 23 turns to 50% population, from a disease that kills up to 5% of the population per turn (congratulations, no antibody immunity here). Alternatively, you could break deathrate down into its various components, something like:

The problem is what happens if you let that trend go for a very long time. If the planet has some 'population capacity' C, where the game manually shuts off population growth, then a fully linear model will always have P=C or P=0 at long times. The difference in which you get can be the matter of a fraction of a percent. Furthermore, the population will never decrease at all unless its destiny is to decrease all the way to zero.

So in your examples, the population will always increase towards C unless you're in the 500% death rate increase case, in which case it will decrease towards zero. So if the player leaves things in that state for a long time, it always runs to one extreme or the other. That makes it very subject to huge swings (its pretty clear that diseases were doing this all the time in some of the earlier versions of TLF, until they tweaked the values a bit).

On the other hand, if you use a logistic model then there's some effective equilibrium population 'E', for a given growth and death rate. If you're below E, the population will grow towards it. If you're above E, the population will decrease towards it. If the death rate goes up by a bit, then E drops by a bit. But you can end up having a steady-state population which is anything between 0 and whatever the effective maximum growth rate can give you, and as you improve the conditions on the planet then E smoothly increases in response.

So in the Skylaxian example, lets say the population scale is 1 billion people. In the initial case you gave, the equilibrium population would be E=growth/death = 2 billion people

Now, when you have a x2 death rate plague, E becomes 1 billion people. When you have a x5 death rate plague, E becomes 400 million people. When you have a 100% increase in lifespan, E becomes 4 billion people. So the equilibrium population adjusts smoothly with the things that perturb it - the result is that if you write in modifiers to growth and death rates in buildings and such, the long-term effects are much more predictable compared to the linear model.

In the linear model, for example, a building that increases the growth rate by 1% can either have no effect in the long term (because growth-death > -1%) or it can switch the fate of the planet from extinction to abundance (-1% < growth-death < 0%), or it can do nothing at all (growth-death < -1%) - so its very swingy.

[ptarth]

I'm still not sure what the problem is. If a population contracts a disease that kills more people than are born each year, don't we want the population to move towards 0? And eventually if no one does anything, shouldn't they all die out rather than stabilize at some fraction of the maximum population?

You were talking about setting up some scripts to produce graphs and this might be a good test case. The easy way would be Googledocs. Setup a sheet that runs for 100 solar months with two columns, one for the linear growthrate & deathrate, and one for equilibrium formulation you are suggesting. Then we just plot them against each other and share the link. We can then get some concrete examples.

The main  problem I can see is that if we model the Burlusts as having a high adult mortality rate due to duels parameter, then the deathrate would be high as well. Any multipliers then would quickly cause the balance to jerk out of control. That might be an argument for breaking down the causes of death parameter as stated in my previous post.

Example
Burlust
birthrate 10%
deathrate 8% (1% disease, 1% old age, 8% duels)
growth 2%
growth after a 500% disease (10% - 8%*5) = -30%
growth after a 500% disease using deathrate components (10% - 1%*5 - 1% - 8%) = -4%

[PokerChen]

I don't know how we manage to fail to communicate so much, NichG. That's exactly what I was trying to suggest. (Followed by some caveats about other things.)

Sorry, I think I was unclear - I was responding to zharmad's suggestion there. I agree with your post.

Did I happen to bring agreement between you two? My job is done then.

 = = =
 Anyway, gamey game is fine with me. I've tolerated laser projectiles since "forever", I can tolerate this.

 I suspect part of the reason RCI hasn't been tweaked, beside the work on expansions, is that it's so closely linked behind the scenes to other things raised in the thread. Although only the Boarines explicitly base their AIs around RCI, all of other race AIs would have RCI components that gets impacted.

 So zero-centering RCI with diminishing returns and no cap? Any other popular suggestions?

 

I'm still not sure what the problem is. If a population contracts a disease that kills more people than are born each year, don't we want the population to move towards 0? And eventually if no one does anything, shouldn't they all die out rather than stabilize at some fraction of the maximum population?

Well, for one thing very few diseases are actually capable of wiping the population. Someone's resistant, governments will quarantine, etc.
 A better reason is that the underlying simulation doesn't benefit from having many extreme effects on the system. If one disease is sufficient to kill a planet, then a second event on top of that may put it beyond player and AI control, and a third one will doom the race quickly no matter what you do. This is already the case with RCI.

[ptarth]

• As I understand it, current events are limited to 1 per race.
• As per disease lethality, my perspective is that the ability to effectively quarantine a disease is built into its effect and duration.
• RCI has already been tweaked a number of times, by huge amounts.

[NichG]

You were talking about setting up some scripts to produce graphs and this might be a good test case. The easy way would be Googledocs. Setup a sheet that runs for 100 solar months with two columns, one for the linear growthrate & deathrate, and one for equilibrium formulation you are suggesting. Then we just plot them against each other and share the link. We can then get some concrete examples.

I set it up on JSFiddle so you can mess with the equations. For some reason it sometimes gets very laggy when you change the starting population - not sure why.

http://jsfiddle.net/J7jws/1/

The various curves show the population at different lengths of time from the start. You can see how the linear model goes to a step function, whereas the logistic model remains a smooth curve.

Anyhow, this visualization does suggest a problem with the logistic model, which is that if the death rate gets very very low then the population can diverge. This suggests that there should be a basal 'minimum' death rate of something like half the growth rate, which is always present even when all other things are adjusted away. I've capped the population at 2*landAvailable in the model to deal with this effect.

Another model would be the 'mixed model' which is somewhere between the two, which basically just looks like:

dP/dt = (growth * (1 - P/landAvail) - death)*P

As long as 'death' is always kept lower than 'growth', this model has a smooth response at long times and will never grow beyond P=landAvail.