Riots Model

[LDiCesare]

This thread follows
the Social Model v3.0 thread.
I propose a new riot model, which I attach hereunder (it's actually html, but I have to zip it in order to attach it). It's a bit big, so I'll put some extracts in the next post and comment further there.

[LDiCesare]

The main changes I suggest are the following:

-Mergeing Nationalistic Rebellion and Self Determination Feelings into a single Separatism Feeling.
The equation I propose is:
((1-DN)*1/(1 + exp(8-13*NAT))*(1-PAL) + (PI*PAL*politicalpower))*sqrt(1-CR)
where:
DN is a dummy variable taking the value 1 if the govt nationality and the EG's nationality are the same and 0 otherwise.
NAT is EG's Nationalism.
politicalpower is the political power share of the social class.
PAL is Province Autonomy.
PI is Province Isolation.
CR is Civil Rights level.

-Having separate PAFs for Revolutionary Feelins for each regime instead of lumping them all together in one event. This is very important in my opinion.

-Making economics a multiplier to other effects rather than its own effect added to the other PAF's.

-Changing Poor Welfare Feeling into:
PWF=(1 - AEL)*(1 - SP)
where:
AEL is Administration Efectiveness Level.
SP is Social Policy level.
This might need work if SP is 1 and AEL is 0 though.

One PAF has only one effect, and the more sever the PAF, the more severe the effect.

I also simplified several equations which I felt were too complex and made some changes to the civil war and internal conflict parts.

[Mark_Everson]

Hi Laurent:

I can't say that i've read the whole thing yet, but I looked at the intro and conclusions and skimmed the rest. Your comments and criticisms seemed reasonable to me. Thanks for all the work that you've put into this! I want to play it!

Returning to the conversation in the other thread about leaving the player the opportunity to use the old complex model. You said the AI wouldn't be able to support it. I propose that if the player wants to play the old more complex version that the AIs should still use the new version.

[LDiCesare]

I propose that if the player wants to play the old more complex version that the AIs should still use the new version.

I don't like very much keeping both codes, but activating different codes for the player and the ai is definitely something I loathe.
Who will be using the old model? I can keep it for testing purposes, but long term it doesn't make much sense to keep both codes.
One problem with the old model is it really requires you to look at spreadsheets in order to be able to correct riots. I don't think spreadsheets are necessarily bad, it's good to be able to get as much info as possible, but it shouldn't be mandatory. There's also the problem of coding all the user interface needed, which would take time and I'd rather skip it for the moment (my not liking ui coding and not seeing it as mandatory to be able to play).
I must also check what exactly is wrong with the ai. I think the problem comes from the thresholds used in the Revolutionary Feeling. The equation currently used is a (disguised) step function. You jump from a PAF of about 0.007 to 0.99 in less than 10% change, with 1% change making for a step of 0.6. This means the side effects on other PAF's of changing regime policies will hide any good effect from changing these policies in terms of revolutionary feeling.
Separating the PAFs by regime should help as we will be sure to head in the right direction when tweaking values and we may be able to optimize one of them rather than all them (3 or more) at once.

[Mark_Everson]

Originally posted by LDiCesare

I don't like very much keeping both codes, but activating different codes for the player and the ai is definitely something I loathe.

Ok, it's not that big a deal. Spreadsheets are definitely to be avoided.

On the step function thing, I think it's very important to get rid of that sort of behavior. Step functions lead to the need for micromanagement, and as you noted also make it hard on the AI. Not to mention being really annoying for the players.

[LDiCesare]

I committed my changes to the repository. I altered the model again a little bit to remove the step/threshold effects.
The most up to date version of the model with the equations is attached.

[LDiCesare]

In order to prevent some riots, one needs to change the province autonomy in some provinces.
Therefore, I need a user interface to allow such changes to be done. Since it's a province only thing, it raises user interface issues about how to show province specific information. Currently, info is shown by domain (there's one econ gui, one social gui), each of which cn be applied to the whole civ or just a province (or even a square). I can do something similar for province autonomy, but I don't really think it'd be appropriate. I think a list of provinces with 'PAL' level adjustable, and later whatever variable we may want to put there for the player to tweak, would be more appropriate. Opinions?

[Mark_Everson]

Your preferred solution sounds better to me.

[LDiCesare]

One more thing.
I don't like the Replace Ruler Feeling. It's basically a mix of Bad Policies Feeling and Revolutionary (Regime Change) Feeling, except it focuses on the ruler's profile rather than the government's. It's also more pronounced in the capital province.
I don't like tweaking the same parameters for too many reasons and would like to scrap this one altogether, but that may be oversimplifying. This needs testing.

[Lord God Jinnai]

I still have yet to figure out what real signifigance RRF plays since if your wanting to riot for other reasons, your wanting to replace the ruler anyway.

[LDiCesare]

I still have yet to figure out what real signifigance RRF plays since if your wanting to riot for other reasons, your wanting to replace the ruler anyway.

Yes, that's exactly my problem too.

This feeling may actually encourage the player to state he wants policies different from what he wants in order to avoid it:
e.g. I want slavery 0 and tax rate 20, but can move these values only in the ranges 15-25 and 5-15 respectively.
If I state I want slavery of 15 and tax of 20, I get the same actual results in the negotiation system, but the replace ruler feelings will be lower, so I'd rather be Machiavellian here.

I added a quick user interface for province autonomy in the social panel as it's something the ai can advise you to change in order to reduce riots, but it's not really good. A province panel will require a lot more work, but that's where we should go. Just not for the next demo yet.

[alms66]

I mentioned this to Laurent already, though I don't know if he already has plans to fix it - the current testbed produces no riots whatsoever. I've mainly played with the social test scenario (since that's what it's there for) and the ancient agean scenario (since I'm finally getting around to tweaking the numbers on the simple tech system). I just thought this little bit of info should be here as well, incase someone else wanted to suggest 'fixes'.

[LDiCesare]

I don't know why you don't get the riots. I uploaded a testbed which was quite buggy but riots happen in it. The current code has riots too, but I have problems committing my changes. Here's a picture of a lot of riots. It may not show very well (I had to resize the image to 800x600 down from initial resolution), but the Pella task force just revolted.
I currently have more problems with revolutions and independence happening everywhere, and some bugs with the newly created civs.

[alms66]

I didn't know about the repository problems, so it's probably just out-of-date files on my end.

[LDiCesare]

I'm finding that the model provides far too many riots.
The Social Test scenaro has been made in order to have many of them, but should be playable.
I'm removing the Replace Ruler feeling, which is a mix of bad policies feeling and revolutionary feeling to start with (or, more precisely, which is countered by the same actions).
I'm also going to tweak the Poor Welfare feeling again, and do some change to the rebels so they do something bad to you if you let them near a provincial capital (like taking power).
I may have to raise the threshold for riots to happen.
I can also increase the riots suppression effect of troops present in a province.
I may also use province autonomy to reduce the overall riots chances for all riots, not just the declare independance kind (though province autonomy can actually increase independance chances in certain cases), so an ethnics in a far away province can be coped with without sending tons of troops.

[Mark_Everson]

Sounds good Laurent. I'll have time to check it out over the Christmas break.

I wonder if we could use a global riot-severity setting for playtesting. That way one could reduce riots arbitrarily in a given scenario if riots were getting in the way of being able to achieve designer goals. It would also help to find the right overall balance between fun and riots if it were settable by individual playtesters.

[LDiCesare]

I wonder if we could use a global riot-severity setting for playtesting.

There is already a minimum riot probabiliyt variable, that can be set and tweaked. It is I believe different in various scenarios, for example I don't want riots in dawn. However the effect is the following:
Min 10 or 20% before a riot happens. If you have 10 probabilities with a base chance of 25% of riots, you'll have 10 times 15% or 10 times 5% of riots, which means respectively at least 3 riot per 2 turns and 1 every other turn. So it does change a lot, but when the riot probabilities grow high and when there are many ethnic groups you wlawys get many riots.

[LDiCesare]

I added an event when rebels conquer a capital (provincial or civilization): They disappear after doing some major riots and change the policies and make the m unchangeable for 3 turns (the murder the ruler event, if successful, does the same). It needs tweaking, particularly for province capitals, but makes rebels more of a nuisance while having them disappear after they've got what they want. In particular, they can hardly be used as a buffer against incoming persians for instance.

[Lord God Jinnai]

Originally posted by LDiCesare
I added an event when rebels conquer a capital (provincial or civilization): They disappear after doing some major riots and change the policies and make the m unchangeable for 3 turns (the murder the ruler event, if successful, does the same). It needs tweaking, particularly for province capitals, but makes rebels more of a nuisance while having them disappear after they've got what they want. In particular, they can hardly be used as a buffer against incoming persians for instance.

Is 3 turns really enough? I mean, if it comes from bad to worse, wouldn't it be better to let the rebels take the capital, reset policies for 3 turns then go back to normal? Ideally squashing rebels is better, but if you can't, 3 turns is a small price to pay to get rid of them.

[LDiCesare]

The three turns will need testing, but they can be annoying, particularly when you have several rebels in different provinces which want different things. Some rebels may set slavery to 100% for three turns, and this can lead to riots so other rebels want slavery of 0%. If you let rebels out of control, with this system you're very likely to be unable to stop them until you kill all of them, and then some still spawn. Of course, if you take care, you can avoid such a situation, but if it does happen, you will really dislike the rebels.