I think this religious law is like an "invisible wall" - I don't really like this one. But the "okay guys, playtime is over now" may be needed :/

About means against revolts, what about calming revolts by paying with influence ?

Ok, now for the very very nerdy part (sorry. Warning, half of this will be giberish for 95% of people, and I have strong chances of being wrong here too) :

I'm supposed to know the ingeneering tools used to balance such a revolt - feedback automatic system but I was maybe the worse student in my school about this at the time :/

This is like tuning a PID controler (https://en.wikipedia.org/wiki/PID_controller). See Ziegler-Nichols's gif and imagine the red threshold is the revolts starts. When the curve is growing, revolts are growing, and when you go down of it = people are calming.

Currently, the system anarchy - strange governements is divergent so when it passes the threeshold, on top of the curve happens anarchy and the curve is getting down. It's working like a Lokta-Voltera differential system (fox vs rabbits).

Less fox = more rabbit then more rabbit = more fox. And then less rabbit. Then less fox, then more rabbits again.

This is because the only attenuation system you have is "religious law. I kill all the rabbits : stop NOW".

There is harmonics that resonate in your system because of the new governement and crowd control means, so the curve is getting up again and does not stop : oscillations are self-induced. It's like passing the threeshold then sin(x) to the infinity. There is no attenuation.

An army crossing a bridge with lilting steps or what happenned in AZF in Toulouse 15 years ago or what happened in the 2000's for cruise controls etc etc etc are all big failure of this math (there is a big list of industry failures on this. Lastly, Samsung batteries charge / discharge cycles, then grenade, can be seen modelized like that too.).

There, it's like passing the threeshold and then x.sin(x) to the infinity (oscillations are growing until the thing explode).

So you have to be very carefull with the attenuation you give the system, because it's entering the system as a command and this may explode depending of the system's equation (in your case, the revolts would be spreading on more planets, I guess)

In discrete systems, you have solutions that does not need lots of computation (you don't want a big computer for something like an Aibo, because energy-quickness-weight-etc but still you don't want him to break his leg when he makes one step. So, you'll want him slow enough so he computes a good correction with his small CPU).

If you've got an ingeneer in Amplitude that knows this (search for "fresh" electronic background), maybe ask him how to modelisate your system ?

(I'd recon this is the "big bertha to kill a fly" solution - but this would help to tune the crowd control laws instead of :

a - duplicating an hard "invisible wall" religious law for any political party, or

b -tuning laws (saying things like "-20% to disapproval at the expense of X industry points or Y food"). Then finding out that a "-20%" doesnt solve many problems, using days of test / failure with different cases of revolts, and having a recurring string of players saying the revolt system is divergent in their save. (there, you may still have the religious law as the hidden rabbit in the hat, but somehow you also want this not to be the universal solution)

Oh, those gains only serves for the science planning ?

If this is it, I have no more excuses ! :)

-edit based on your edit-

Yeah, all laws would come with an approval set to 0, then after an election all laws belonging to the party in power set this at, say, +5 (to be balanced), and all the laws for the last party gets -5.

The one in between would get something between +4 to -4 depending of the spread.

So if the gain attribute helps the AI to unlock stuff based on its needs, then I need to tell it laws that got a +5 are approval laws and the one that get -5 are disapproval laws.

If this is only telling the AI "hey, just use science toward this stuff because it's approval / disaproval", then I shouldn't tell the AI and it should be great.

But even if it can parse the xml, the thing is maybe the AI doesn't know it have to find approval/disapproval there without telling it in the gain attribute : the approval would be a new value it doesn't know, and the application of it would be decoupled in SimulationDescriptors[ColonizedStarSystem].xml or in Empire xml : I think AI have really no clues.

Without telling the AI at least "warning approval stuff" or "warning, disapproval stuff", it would make random choices and maybe break the approval in its camp.

Now, hope is not dead, I would have to create an xsd and use a new command. I have strictly no clue how do to this, and if the parser can use non-ASSET xsds, but there must be a way to change an attribute ;)

Or I can do this the ugly way, duplicate the laws  and then putting one gain attribute law with approval, and then the other one law with disapproval gain and then the whole computer becomes a heater... lol.

@Wilbefast

How about a short term law that deals with anarchy but could cost in life and/or infrastructure? 

A martial law?

Kweel_Nakashyn wrote:

Yeah, I totally agree with you. I just hope "cognitive overload" would be less important with direct levers than with things that you have to planify a few turns before they happen (like buildings etc) and stay permanent.

True - humans have a natural tenancy toward post hoc ergo propter hoc, so a "tight" causal link (ie. immediate response to input) will tend to be more satisfying than a loose one. When you think about it though strategy games are fundamentally about loose causal links: about doing something now so that, in a few hours, will win you the game. If you subscribe to Raph Koster's theory that fun is an evolutionary response to learning then you could argue that a strategy game is fun because it teaches long-term planning.

A better precise thing would be to do this for Democracies per planet, for Federations per systems but I think it would not work, I'd use the whole senate for this.

Democracies per planet? Interesting... again though the human mind has only such much processing power: if you add too much complexity people aren't going to be able to keep up - unless of course you remove complexity elsewhere to make up for it. We could make the political simulation in ES2 much deeper if we took a machete to the military side of things. Ultimately it's a matter of what you want players to focus on, which in turn is a matter of the overall experience you want to create.

I had some technical troubles on this that made me hesitate to try which was "how can I tell the AI about the approval/disapproval on each law, because these indicators are in xml attributes (attributes, not values), and I don't know how to use a prop modifier to change attributes".

I can't do this until I know how to do, or I'll break the AI I think.find

There are "gains" in the XML for certain specific elements, notably gameplay unlocks - edit: ah... but the tricky part here is that the approval gain depends on the support for the party the law comes from  Yes, that would probably require some specific code.

Ultimatly, that's somewhere near I want "Happy Dictator" to go : adding non-permanent and direct levers on the approval system & trying differents levers for each regime. "Happy Dictator" would be the one for Dictatorships. I may have approval / disapproval tags on laws for the Empire, and others are still WIP ;)

It's possible we'll make propaganda provide approval too  This would enable dictators to play for time a bit more effectively.

I told you that was (very) nerdy :)

Take a system where you have an input, your system, then the output. The output is cyclic.

The output gives you :

- a - the result of the system

- b - because you want it corrected (because you don't want your cycle to oscillate forever, for exemple, and you want a needed result, and quick), you're comparing the result to your input to enter the system again.

This is giving this sort of shematics (ignore the PID part):

The PID is a method to control the output cycle so instead of this (ignore the negative part) :

... which is meaning that the more time your system oscillate, the more time it will explode in the future (aka Nokia Grenade, aka the army on the bridge, or something everybody experienced with an amplified guitar, the larsen effect)

The more time in this system, the more your amplitude is growing (which may be a compagny goal :) ). But the more boom risks if the system can't handle the amplitude diverging to the infinity (which may be... err).

Or this:

Which is AZF version (disregarding the aging of the machines... then when aging had an influence on the chemical reaction, coupled with half of machine working this day, switch from this to the previous image and then boom).

This is your anarchy - dictatorship -anarchy endless cycles.

But instead, it tries to do this :

You have this last cyclic behaviour if you correct the input depending of where you are in the output. Red, green and purple plots are typically the result of a PID correction.

(next post because I lack of images in this one)

PID is the cheapest way to adjust the command you give to electric engines to make robots move smoothly, very quickly, or forever. But you can use the math there in discreete algorithms.

Each corrector have this effect on your system:

Precision is the factor to make the exact result you want with a command.

Stability is nerfing the cyclic behaviour so it ends at the wanted result.

Quickness is the speed the system to acheive a result.

You can't have everything : any correction cost something.

P correction boost precision and quickness but nerf the stability.

I correction boost precision but nerf quickness and stability.

D correction boost stability and quickness but nerf precision.

Once again:

* Blue plot is the perfect theoric result. There is way to do this in discrete systems but it implies some problems (it's 80/90 technology known as "réponse pile" in french, there is huge problems with it. You draw all your points on the blue line but nobody know exactly how the real physic world behaves between them : it can be a sinusoid, with an amplitude that can make a system "explode").

* Purple have lots of precision problems (huge difference with the blue line), and stability problems (it's oscillating forever). If your system can't handle the 1.6 result then it'll go boom. But it's the quicknest to go at the blue threshold.

In dance, these are acting like 70's popping breakdance (electric boogaloo style. Modern break & poping does not do those exact robot moves anymore, they are quicker, they are more like green curve. There was a change of technology...). :)

* Red have quickness problems, is very stable (no oscillations) and have precision problem in the begining phase. Other system depending on this may have trouble when staying too long under the blue line. Plus in electronics this one is slow, so you have heat / electric consuption problem (+ heavy batteries...) problems. You don't want this in autonomous systems. But you want this in systems that will be go boom if the result crosses the blue line.

These are like my grandmother crossing a road in slowmo. :)

* Green is a good compromise. It's quite quick, don't oscillate forever and is also quick to be near the blue result you need.

These are like 3d sprites repositioning one from each other to play a 2-sprite animation, they have to "snap" together (they correct each one the opposite way to get to the right distance between them), then adjust, then play the animation (Total Wars, Assassin's creed, etc.). Not sure they use a complete model of PID there though.

A combinaison of the three makes you able to control totally the behaviour of your result, depending of your needs. In music, PIDs are the one thing that is doing most stuff in an enveloppe filter (aka "autowah", an automatic wah-wah pedal if you want).

Jimi Hendrix was a automatician :p

- You don't want your robot to run at 0.1 km/h ? Boost the P and D correction and nerf the I of the legs engines. But your system may be not really precise.

- You want your robot to climb stairs ? Boost P and I so it can adjust very well to the step. But this nerves stability so your command may be autocorrecting forever. That's why you see robots climbing very slow : they boost I and D for stability so the robot eventually climb the stair very safelly and not repeat first stair's step again and again (this may make the robot falling... in a stair. aka no way.). This cost precision and quickness, and you care precision.

In Europe most people know how to adjust a PID to their needs, but what knows robotic compagny in US/Japan or here is to adjust these PID depending of the system's behaviour (in robotics, on top of this, they are good at data fusionning from captors, includes a little of prediction to fill gaps and have better ready-made AI algo and low energy computing power than when I learnt this but whatever).

OK. That's for theory. There is also other correction methods than PID.

For gamedesign, this applies anytime when the result of a part of your gamesystem influence the next state, with a cyclic behaviour by design that you want to correct (the cyclic behaviour is fondamental or else you don't care all this stuff).

You have a anarchy - dictatorship cycle here that may need this math. Alternativly, we can use the religious law as an (ugly ?) "invisible wall".

The uglyness of the "invisible wall" there is a valid question. I think it's a good solution, if the player still have control to get that religious law (in the current build, they have to wait for the religious dictatorship to "pop" either as an anarchy result or an election, then be aware of that law and choose it : they have little control over it).

The problem today is cycles of anarchy happens to fast to get an election I think (among other things like the lack of ways to have "overhappyness" in your systems. You can have "over production", over dust etc, but no overhappyness, because happyness doesn't snowball).

If you tune "correction" (aka what you present as negative / positive feedback), you can have result influencing the next state like the red curve, the purple curve or the green one. With PID math, you can control this very smoothly depending of what rewards you gives to players with buildings or laws (but there is stabilization to have in your game balance before, because you can't finelly "tune" these things if most of the things are not near final balance) :)

That's why I told this may be the "big bertha to kill a fly", as this may be a difficult way to modelize then balance the system, with a lot of die/retry, if you allready have an "invisble wall" solution - I think you should just give control to the player there and this will do the job I guess ?

-edit- there, I told you this was nerdy !! :)

Let's get back to the gameplay... I forgot this:

wilbefast wrote:

Kweel_Nakashyn wrote:

what about calming revolts by paying with influence ?

This would make sense I suppose - after all you can capture others' system with influence so why can't you keep your own? There is one thing that concerns me though: we had, at a time, too few things to do with influence; I think now we may be getting to the point where we have too many

I feel that strategy games are most clear when a resource is the reification of some means of action: the FIDS are defined more by what they do than what they are. If influence does too many disparate things it'll be harder to grasp, conceptually.

Influence & riots snowballs, when happiness doesn't.

On one hand, you can solve a snowballing issue (riots) with a non-snowballing tool (happiness), but in the end the issue may grow too strong.
Or alternativly use a snowballing tool to solve a snowballing issue. Imho, this is a good way, because you don't have to define the limits of the snowballing crisis vs the non-snowballing tool.
You also don't have to define acceleration of the snowball, like colony disapproval % bonus (which are sorts of "bonus on malus" ?), and this helps to keep things elegant (this is not a reification vs elegance debate, but this debate exists :) ).

Laws are free for dictatorships. Diplomacy aside, they have almost no influence need (this is all the influence reification that is, unless you made recent changes to the game - I will happily wait for tomorow for this :) ) ?

So this would just boost the inf need, since they are (in v0.2.5 S1) the one that have the most regime instabilities (but I guess there was a lot of change allready there)

==> any regime but dictatorships would have inf use for diplo + laws

==> dictatorship have inf use for diplo + crowd control.

==> Cravers with vanilla dictatorship would have inf use only for crowd control (and slaver's unhapiness). In the left quadrant, they have to choose between happyness / food / ships / influence, and they may forget food there. They would... Well... Crave for food ?

==> any regime have influence spheres around systems. With this, Cravers & Horatio having that inf need again would have less problems against the UE ?

wilbefast wrote:

I'm not sure divergence is necessarily a bad thing in the absolute though - convergence means setting everything up, flipping the switch and waiting for things to, well, converge. Divergence means needing to keep your eye on things. It's a question of how much planning you want the player to be doing, and how much juggling.

Snowballing divergence is okay if you give the players the correct snowballing tools to correct the situation and converge again ?
Or else you have some nerd posts on the forum about obscure automatics :) :) :)

wilbefast wrote:

Approval affects all the FIDS, it's just that industry and food and tied to system approval while dust and science are tied to empire approval.

Not true.  On a system with 8% ownership and mutinous approval rating, industry output continues at 100%.

I wrote up a related idea (https://www.games2gether.com/endless-space-2/ideas/1413-system-ownership-should-penalize-industry-dust-science) before I noticed this post which was bumped.

-HP

It's not (only) a problem of spread.

The main problem is the flow of anarchy -> new gov -> anarchy -> new gov -> anarchy etc.

Spread prevents you to go in anarchy the first time, but once you're there, you'll be here forever :

- You're in a democracy with 100 elected rep (50 industrialist, 25 science, 20 ecologists, 5 religious) -> 50 * 5 disapproval = -250

- Anarchy because of "whatever problem"

-> you're in a dictatorship with 100 elected rep (because last time I checked (in 0.23), reps in senate did not update), forced to use science, with 75 * 15 disapproval = -1125

-> Anarchy

-> you're in a dictatorship with 100 elected rep, forced to use industry, with 50 * 15 disapproval = -750

-> Anarchy

-> you're in a dictatorship with 100 elected rep, forced to use science, with 75 * 15 disapproval = -1125

-> Anarchy...

Until there is elections in a non-anarchic period. Spread may help to not enter anarchy in the begining, but once you're in anarchy, you will yo-yo at least until the elections.

When elections occurs, you get a normal number of rep for a dictatorship (here that would be 34, with something like 17 * 15 = -255 approval or something, and does not solve your "whatever problem" that lead to anarchy with -250 in the begining, so... Anarchy again).

Unless you use Religious Dictatorship + the content law (or my mod removing all the laws, so you have a "light" dictatorship), there is no way out of that anarchy pit, because dictatorship doesn't nerf disapproval when (I think) a dictator should have direct means of controls on it.

One solution among others:

I think a dictator can't tell people to be happy but at least he should mute the unhappy guys.

There could be 5 dictatorship-only laws like "+2.5 approval by representative not in senate, -X% to  in systems" (and one for , one for , one for , one for ), where X is a number like 10, 15 or 20 or I don't know,  depending on how you want to balance this.

Having all 5 of them running (if dictatorship gave you 1+5 law slots) gives you -2.5 per rep not in senate, which is bearable even with post democratic rep. So you can stabilize until the next dictatorship elections & get elections.

From there you could fix the problems then go back to your old governement 20 turns after, which should be a valueable goal.

You could even go to democracy -> rebellion -> anarchy -> dictatorship, fix problems -> switch gov -> anarchy -> democracy -> election in one, or more probably two election cycles with this. It would cost you some FIDSI though, but that would work : 20-40 turns unstability is a huge punition enough I think.

Ya right now the game does favor democratic forms of government with a greater ability for representation and more law slots.

On tanget, I would like to see a change where dictatorships should be able to pass the most laws because its a dictatorship and you should be able do w/e you want. I mean did Stalin ever give a damn what his people thought? And did he ever have a limit to how many laws he could pass? No, he was like I'm Supreme Commander: and you get a purge, and you get a purge, and everyone is getting purged! Like Oprah gone wrong.
I digress...