Proposal for a Complete Infrastructure Model

[axi]

***bump!***

Since we're OO-Designing this model right now, I thought I should bring up this thread, so that we can discuss the model itself somewhat further.

[Mark_Everson]

Axi:

I have finally had the time to give your spreadsheet a serious look. I have also reread much of this thread, and will comment with the new ideas I have based on the spreadsheet and my reading after the hiatus. (I won't mention any of my old criticisms and questions about your approach unless I have something new to say) But could you please respond at some point to my comments that were made on Sept. 24? I never heard from you on any of those issues, and some of them are fairly important.

In looking at the spreadsheet, I want to congratulate you on what a great job you did organizing it! it was for the most part easy to follow. However, I still don't know what RU means in the government area, can you explain it to me? But on to some more important issues.

1. I think the whole idea (and you can blame me for this) of having the people invest x% in a certain infra area regardless of how much the current stock in that area is, has serious problems. We have discussed this before in general, and you mention a way out in terms of emergency spending on infrastructure. I have a new suggestion. What I think would work best is that we simply define an "Ideal Value" (IV) for each infrastructure type. So, for example, let's say that for health-care and modification of the preference machine would tell us that the people want, in a certain province, 100 units of health-care. (This would be expressed as a certain percentage of spending on health-care x the gross provincial product or some such) Anyway, the important factor here is that what the people want goes into determining the ideal value. Then, when investments are made the appropriate weighting factor for each type of infrastructure would be proportional to something like (ideal value - actual value)/price. Of course this expression isn't complete, since it should include the difference between persistent and consumed quantities, decay rates and such.

This way of handling it would automatically include several features that the current model needs tweaks to fix. First, there was no longer be a need for "emergency" modifiers for the case where all the housing had just been destroyed. Second, when some areas had reached their ideal levels, but others had not, the people would switch off or strongly curtail investment in the areas where they were happy, and redouble their efforts in areas that were not sufficient yet. Obviously what I have is not a complete determination for how to change the existing preferences and investment machines to this new way of doing things. I wanted to see first if there are any problems with this approach I hadn't foreseen etc.

2. For something like the tech level A' for infrastructure I have come up with a fairly simple modification of what I said in the post of Sept. 26 that seems reasonable to me. Specifically, for each type of infrastructure we could use a formula very similar to the application effectiveness tech formula in the tech model. All that I think we would need is a way to give a certain level of infrastructure a calculated tech level based on a weighted average of critical technologies for that type of infrastructure, and take the ratio of that to an "average" tech level for the civilization. So if you are behind in key technologies for this certain type of infrastructure, A would be below 1. The weighted average would be done similarly to the relative tech level for applications. Specifically...

[Paraphrasing Richard's thing on the web page]
T1 through Tx are the levels of the technologies that are important for the application. R1 through Rx are the tech level modifiers for the application. The R values would probably usually be zero here. h1 through hx are constants that determine the influence of each required technology on the growth of the application.

RTL = (h1*(T1-R1)+...+hx*(Tx-Rx))/(h1+...hx)

So we have A' = RTL/OTL where OTL is the average or overall tech level of the civ.

Note that this is not a very large penalty factor for infrastructure compared to what would exist in the real world if a critical few technologies in the area were way behind. However, I don't think we need to belabor this point to much, since A for infrastructure isn't critical IMO to making the overall system work correctly. Or alternatively we could use something with a power law or exponent on the right side of the equation

That is all I have on big issues right now. There are a few minor things...

Kapital for each economic sector is needed, not just a single number as you have a written up.

You emphasize that at higher population density some types of networks are cheaper. Actually, I think the most important population density of facts are in things like water infrastructure in the ancient world. In this case, people spread sparsely on the landscape don't need any water infrastructure at all, whereas big cities need aqueducts etc. My only point here is that frequently population density has serious costs associated with it, in addition to the opportunities that you cite.


[This message has been edited by Mark_Everson (edited October 22, 2000).]

[axi]

Mark: I did indeed owe you a response on your two latest posts. Please have in mind that after 2 months I'm a little rusty. So, on with it, easy things first:

1.

quote: However, I still don't know what RU means in the government area, can you explain it to me?

RU means RULER, silly!

2.

quote: Kapital for each economic sector is needed, not just a single number as you have a written up.

Does this mean that we are definitely going to compute investment in kapital through the infrastructure model? Nevertheless, kapital per sector is also computed in your econ model. If we do sector kapital from here, we will have the advantage of political and social control over it, but we will lose the automatism that productivity of kapital per sector provides us with. I propose that for sector kapital, the base budget (col B of prefmach) should not be a game constant. Only the sum of all three base values should be constant and they should be estimated the way you do in your spreadsheet (in % of course). Then the various prefmach modifiers should come atop of that to give the final (obviously not economically optimum) solution. What do you say?

3.

quote: There is much more micromanagement than this possible... You aren't counting subsidies and penalties for infra types, which is one of the player's main levers of control here.

Read what you' ve quoted: "On micromanagment: Ruler has 3-5 cv's to input, per place of model implementation. One for Economic Planning, one for each of the 1-3 levels of govt, and one for subsidies and tariffs..."

In chapter 10 of the model text, you will find this:

quote: As you have noticed, price subsidies are not yet modelled in the spreadsheet. It wouldn't be hard to do so in the investment machine, as long as a way for the bill to be settled were devised. But, specially since the computations are made for all agents together, it sounds silly to adress a preference issue outside the preferences machine. On the other hand, here we have no standalone preferences but preferences linked with prices and actual money transfer. Things could be made easier here if the public and private sector were separated in the computations. So I would prefer to wait and see the final implemetation that we will choose and then apply subsidies.

Subsidies and tariffs are not modelled yet, but they are counted in.

More micromanagement could only come out of the type of player interface we decide to wear on top of this model (since this is only the machine, the "M" component in F_Smith's MVC approach). The "C" (controller) component could be a very bureaucratic one, taking exact production orders, civ-like or not: F.e:

civ2: "Build a temple"
Clash: "Make me 40 new religious infra units in prov A, however long it takes" which will be translated, depending on the funding and/or happiness limitations you have installed (f.e. no riots, from your own funds, no subsidies), to a certain set of cv's ( the EP one and 1-3 govt build ones) which will end up producing the 40 units in some number of turns you can't control. Of course you could tell it to build more types of units simultaneously (paralell build) or one after the other (building queue).

civ2: "Rush-buy a temple"
Clash: "Make me 40 new religious infra units in prov A, next turn, with my money" This will add the needed amount of ruler's money in the provincial budget and set the cv's as high as it can, to make the money you've put inside come out of the invmach as rel infra. This of course will have other repercussions on the provinces infrastructure market, but that's how things are in reality. This will also demand some calculations to determine the cv's. This can be extended in a longer turn span and also include parallel build and building queues too.

Of course such orders, by focusing on the output instead of the budget are somewhat changing the type of the control system in practice. One might not want to play "civvish" Clash and just give the abstract political directives that the cv's offer. Civ2 control freaks on the other hand could opt to wear the micromanagement "glove" that such a controller is and demand results from their subjects, instead of giving directives. But then they would probably have all infraclasses units translate into "city improvements" when they are gathered in a nice round figure. Duh!

4.

quote: When the player places a fortress on his map, equal amounts of money will be collected from all the provinces and be given to the province where the fortress will be and the CG cv will change, in order to provide for the production of the appropriate mil.infra units.

Yes this rather confusing. You see, building a fortress on or passing a central highway through a tile is an action from which not only the locals will benefit. It is also an outside order from the player, saying that "you won't build small forts or roads all over the place" (generic transp and mil infra units) "but concentrate them where I tell you to." This can obviously happen with the ruler's money and it will happen in the way I describe (without the italicised part) But how about the civs where the ruler can't (won't) build naught by himself but wait for the people to mobilise. Will they perform such concentrations on their own? I was thinking that they would maybe take the initiative to build such a thing and just let the ruler decide where. It is a minor thing actually and will be probably practiced via the AI (since it should be able to place fortresses and transp. networks for the enemy, it should be able to do that for your own people).

I owe you a couple more answers, the most important ones. I am currently examining them. I will be back to complete this post.
------------------
"In a time of universal deceit, telling the truth is a revolutionary act."
George Orwell
[This message has been edited by axi (edited December 21, 2000).]

[Mark_Everson]

Hi Axi:

Thanks for the response. I'll reply by the numbers...

1. Hmmm, perhaps it is not needed to have an abbreviation that adds confusion and only saves typing three characters

2. I am inclined to handle capital for the economic sectors as an infrastructure class. I certainly think it is better to unify the capital in each sector with infrastructure just to keep things as simple as possible. And as you point out we get all the good effects of social mods etc. in the bargain. I don't think there should be any big problems, just perhaps some tweaks of the system as you suggest might be necessary. Perhaps it would be simplest just to add a set of economic modifiers to the preferences machine such that if food is scarce (or food prices are high) it would increase investment in agriculture etc. We could use the utility function in a similar fashion to the way it is used in the current Econ model as an input for the modifiers, but probably using prices is simpler. In this way we wouldn't need to link the three base values as you suggest, which seems to me to be overly complicating things. I'm sure we can work together to produce a good result on this.

3. Oops, I just skipped over that part about subsidies . But on a related note I just noticed something. Do you really think that the ruler is going to add all these correlation values keeping everything consistent? (I have to ask this because I don't Really understand everything in the preferences machine.) It seems to me that there is a very real chance that the player will get incredibly frustrated because of unforeseen interactions between the different levels of modifiers. Then again, if it can work it will give a very strong and versatile system to the game that will be able to reduce micromanagement drastically. Personally, my thought is that only the preferences machine for the central government should be able to be modified directly by the player. That is because the player does not directly control local and provincial governments. I think the correlation values for local and provincial government should be determined by the culture itself. The only way for the player to influence these numbers would be to try and change the culture. (Or the player can get his hands on these resources by simply taking control of them with much higher central government power.)
As you know, in addition to the central government values, the player can favor or penalize various infrastructure classes through subsidies/taxes on particular infrastructure classes. Additionally, for resources that the player controls, like spending tax money, or when you have a command economy, the player should be able to just say what should be built specifically. I think these three alternatives (central government cvs, subsidies/tariffs, and spending money from the treasury on specific items) are sufficient levers for the player to control directly. What you think?
So I believe we should essentially have separate private and public sectors. In the private sector the main influence of the player is through subsidies and taxes on various infrastructure classes (provided the player's government allows this much messing around with the economy ) . In the public sector it would work with the "public" preferences machine, which is not in the direct control of the player for local and provincial governments, only for the central one. With central government funds, the player has a choice. Either use the central government preference selections and shovel money into it, in which case the decisions are made automatically depending on the preferences, or buy something particular. The case of buying something particular is, as I see it, somewhat like your example "Build a temple". The player would express a desire, and local tax dollars could then be used to build whatever infrastructure is needed. And as you did in your "Rush-buy a temple" example the player can always add funds from the central treasury to hurry things along. But as I said above, I think this is best done without messing around with the cvs explicitly. Changing provincial cvs every time you want to build something seems to me to require much more effort than should be necessary. Do you agree?
We will also need a set of commands for something like "build forts along this boundary, one in every square, over the next five years". But that is probably better handled at a later date as a general interface issue.

4. I had misunderstood what you are saying on this one. As you say, this isn't something we need to solve immediately. But I have some comments anyway! Again, I think that if it is not the central government that is paying for the fort, then it doesn't get to say where it goes. So, for instance, we can have cases like feudalism. The local and provincial share of overall taxes will be much higher than the central government's share. Each Duke in his own area will want his own castles, roads, and whatever. Historically, the feudal system was inefficient, but it was also very robust. By having lots of little castles all over the place, as one example, it means any conquest of the civ will of necessity be a very slow grinding process. A more centralized civilization might place all its fortresses on the border, but the problem there is that once the border fortresses fall that civilization can be annihilated in quick order given the right circumstances. To sum up, I think that the player should only get to place things that he or she pays for. Otherwise we should use the same procedures and AI that is used for non-player civs. Does this sound reasonable to you?

[Mark_Everson]

Axi:

You'll be pleased to know that I am now taking a stab at a simple implementation of the infrastructure code. I am going to try using infra classes to hold kapital for the economic sectors as we've discussed. There will be no preferences yet, or actually just very simple ones, but maybe soon things can get more detailed. Since the govt code is far along and the econ will be soon, perhaps we can get going on coding the joint between them within a month, maybe less.

[Mark_Everson]

Status report:

The econ coding is going pretty well. I have crude infrastructure classes running. They handle kapital in production sectors. Also I've put in a "consumer stuff" class to give the people something to invest in that improves their physical well-being. Once all the bugs are kicked out, I could add as many infra classes as we want. First I will flesh out the PublicSector class so the govt can give orders to the economy. Then I will work on a simple alpha interface. Following that I'll be ready for infra again.
(I'm also cross-posting similar to this in the Econ OO thread).

Axi:

If you have time, we should get this discussion to a point where I can code it... It may be as soon as two weeks from now that I'll be ready! So I would like to hear from you on those outstanding issues we still have hanging.

-Mark

[axi]

I'm still reading the forum, but the truth is that I'm in deep **** with my faculty obligations. I'try though... with no guaranteed results.

------------------
"In a time of universal deceit, telling the truth is a revolutionary act."
George Orwell

[Mark_Everson]

Hi Axi:

Sorry to hear you're in deep doo doo . Well, do what you can... But its a win/win for me: If we discuss things a while they'll turn out better -- If we don't, I'll do what I feel like .

[Mark_Everson]

I am now coding building of military units. The approach I'm taking is that when there are orders to build a military unit a temporary infra class is created to build that unit. The cost of that infra class will be the raw materials (food, manufactured goods, and services) amounts needed to complete the unit.

So ancient warriors with clubs might cost (0, 3, 2) in the same order as above. The services need is mostly initial training. Additionally the people would also be needed. A legion might cost (0, 70, 50), and a modern armored division (0, 90,000, 120,000).

When the cost is reached the temp infra class disappears, and out pops a shiny new unit. In times of emergency it would be possible to just pop out the unit with whatever sub-normal strength is available at that moment. I think this system is very flexible, and will give us most of the benefits we are looking for without too much overhead. You can simultaneously build as many different things in the same place as you like. Each square will keep track of the units it has built so we can model happiness effects, and when a square thinks it has built enough of its own units, etc.

Comments?

[axi]

On your mil infra post:

1. IIRC, we are not speaking of units here, like in civ2, but rather for TF elements, right? A roman legion f.e. is 8 infantry elements, 2 cavalry elements, 1 catapult element and 1 engineer element. Different elements, coming from the same (or even differrent provinces; should we allow this?) can form a TF which fights as one.

2. Given 1: Each element has a different cost, both in quantity and in quality. So, we can't have a single "mil units" infraclass, as I had proposed, but several different ones. These are:
a) added and removed from the infraclass list as military tech is changing, but exist for every province at a given time, even if they are not building the corresponding elements (in which case the are zero)? or
b) instantly created once the production order for the cooresponding elements is deposited? In this case, how can the provinces build TF elements on their own, via the preferences machine?

Did I get this straight?

3. Given 2a: One infraclass for each element type is IMHO too much infraclasses. Shouldn't we instead have one permanent infraclass of each generic type of TF element that we can have, so that similar elements will have the same resource analogy? F.e. a Greek style Phalanx and a Roman style Cohort cost 100 and 150 infantry units respectively, each infantry unit being 1F-3P-2S, while Light cavalry and Heavy cavalry cost 100 and 150 cavalry units, each cavalry unit being 2F-4P-3S. What is good with that is that, when a province that is raided by pirates decides to build ships, the choice of what ships will be built is left to you. In general, it makes for better organisation and it might be useful for the military model (appointing abilities according to unit type, but I'm not familiar with the mil model anyway). This is an intermediate solution between my proposal and yours. I'm not insisting on anything here, I'm just proposing an alternative, just in case you might be interested.

------------------
"In a time of universal deceit, telling the truth is a revolutionary act."
George Orwell

[Lord God Jinnai]

quote: Now, if you started building a cohort and it suddenly happens 1- you discover a new tech making this obsolete . . . you can keep the higher level class and switch from cohort to phalanx if you like. Does this match what you meant by having the choice left to you?

As far as that is concerned, you must first realize that gaining a tech and implimenting its use are 2 differnt things.

If it were to happen and say all other RQ are met (social acceptance, ness resources, etc). That should IMO be allowed, but at a cost at say 25% more to build that unit because you must retool it and retrain them. Even if you are simply going from bronze axes to iron axes, that needs reteaching of some sort (iron is heavier and more powerful, so the axes are used diff) and retooling of the blades.

[LDiCesare]

If I understand Mark, there is an instance of some infrastructure class created that symbolizes the building of a unit. Other infrastructure tend to stay there when created, while this one will eventually morph into a unit.

I coded the military units as each unit (e.g. the unit number 42 in square x,y) having one archetype (e.g. phalanx). It seems to me the infraclass could refer to these archetypes. That way, you do not have any extra class. The idea behind the archetype is you could create them from a text file with figures about cost etc. so having all data about mil value and infra cost (build, upkeep) appeals to me.

Now, should they say I am worth 100 units of cavalry and refer to a new cavalry object, as axi proposes?

I would say yes if this cavalry object has uses beyond just the infra. It can be used for AI to understand what kinf of unit it is.

That way you could have a high level type of army which is referred to by several archetypes, which are referred to by several actual units. All data could basically be taken from this higher level class and passed down as needed.

axi, does that match your thoughts?
Mark, can that match your code?

Now, if you started building a cohort and it suddenly happens 1- you discover a new tech making this obsolete or 2- you are attacked and decide to rush by so you can make it a phalanx instead, you can keep the higher level class and switch from cohort to phalanx if you like. Does this match what you meant by having the choice left to you?

[Mark_Everson]

Hi Axi, good to hear from you!

Your 1 is a good point. I don't have the answers there yet. Right now I'm treating each unit as if all the elements are the same... I guess I will try soon to put in the possibility for the build-by-elements way to keep it flexible.

On 2, It will work as if 2a were true, but is technically 2b. People's preferences can create a needed infra class to build units (elements) any time they are called for. This is just like the class being created as needed when the govt (player) orders building of a unit. The AI will need to have some lookup function to figure out what Sort of military stuff the people should be building. I think this genral approach will work out fine...

I'm not sure we gain anything with your suggestion 3, but maybe I just don't understand it. The kind of build classes you are talking about might be a good simplification for the interface though.

None of these things are set in stone. Its just the way I'm putting it together at first. FE the current approach can handle either 2a or 2b equally well. I think what I have will support most of what we would want to do, in terms of detail, although I don't intend to go to the limit on much of it.

On your "mil units are infra" idea expressed much earlier, I frankly still don't like it in general... I prefer "mil units are built the same as infra" which is the approach I've taken in the code so far. But if you have a chance to write more on the "are infra" case we can see if the concept has appeal. Here's to hoping you get some more free time!

One other issue while I'm thinking about it. Some kinds of infra, as we're handling it, should really be what I think of as 2-stage. FE in the case of food storage the two stages are 1. build storage and 2. fill it with food. Because surplus food itself does no good unless you have an appropriate place to store it. Before I just combined some storage building in with food in the hope that it would average out, and not be too annoying . Does this two-stage infra strike you as a reasonable thing to include, or too detailed? It is especially important for building say battleships where you need 1. shipyard facilities before you can 2. build the ship. While we could finess the distinction in food, for military builds I think the 2-stage is fairly important. In summary, 2-stage is more complicated for the player and AI, but more realistic. Whatcha think?

[Mark_Everson]

Laurent, I'm sorry, I missed your post of January 23 because it was posted almost simultaneously with my text that follows it. I'll give my answer to what you said now...

Indeed, as you say, currently when I build a unit I make a dedicated infrastructure class with the same name as the unit archetype. Any archetypes can come from an initialization file, or unit workshop if we ever go that far. So if you check the Econ demo code I sent you, when the government builds a Legion there is in infrastructure class created called Military-Legion. It springs into existence when the order is received, and disappears when the legion is created (unless its a continuing order). Now if we go down to the element level for building military units then the cost of a Legion would just be the sum of the cost of all the elements.

This is where axi and I differ in opinions. I really don't see any point in keeping many infrastructure classes around that aren't doing anything and only connect to units that are somewhere out in the field potentially. Personally I think a square should know which units were created there, and just have pointers to them. If you see a big advantage to doing it Axi's way, then we should discuss it at more length. Unfortunately Axi has been tied up recently and hasn't had time to discuss the idea much. But IMO keeping military infra around when nothing is being built will just complicate the infrastructure class with no real game benefit. In terms of using the infrastructure class to keep track of aggregated Army information, personally I would rather just count the number of armies from here .

On switching in the middle of building units, I think we should leave that decision until playtesting. Certainly, as LGJ says, in terms of real life it doesn't make much sense to be able to take resources that are already building certain types of weapons and just magically change those weapons over to other types of weapons. But I think player satisfaction is more important than realism here...

[This message has been edited by Mark_Everson (edited February 03, 2001).]

[Mark_Everson]

Hi Axi!

There is no preferences code per se yet... that's just way too detailed for me at this point. Perhaps I will be able to have some sort of real preferences set up by the demo following this current one, which might be out in a few months at a guess. Right now the people just: eat food, and with the remainder invest a little in kapital for the economic sectors, and put the remainder into a single infrastructure class called "consumer stuff". And the government buys things in the way that we have been discussing in that thread, although many of the details are TBD. The things I am coding now in the infrastructure area are pretty much none of them final, they are either placeholders, or experiments. The real purpose in my mind of the Demo 5 part of the econ model was to get production and buying and selling of stuff squared away. Everything else is mostly trying to use those in a productive way.

I'm not to the point of the people building their own military units yet. There also is no government level other than the centralized one. I'll get these things in place at some point in the future. And the people Will be able to build TFs by themselves, which I believe I already said previously in this thread... "People's preferences can create a needed infra class to build units (elements) any time they are called for. This is just like the class being created as needed when the govt (player) orders building of a unit. The AI will need to have some lookup function to figure out what Sort of military stuff the people should be building." I think this will support your approach just fine...

I basically agree with the stuff you put in the Econ II thread. I'll comment on some details later.

-Mark