Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[GoblinCookie]

First, there is nothing to prevent a demand curve from crash-diving after satiation.  Willingness to pay can actually go negative (You want me to store extra food beyond what I want for myself?  You pay me to rent my space.) without any modification to the underlying system.

Second, supply curves and demand curves do tell the computer what to do.  Individual-level demand (based on utility functions) is aggregated to the site level.  This lets the computer "know" how much of each good would sell at any given set of prices.  The supply curve is constructed from the site's resources including capital (available reactions), labor and materials.  This lets the computer "know" what is the best mix of goods to produce at any given set of prices.  These are high-dimensional curves that are difficult to visualize, but a "shadow" of the two curves reflecting a single good would look like the supply-and-demand curves in that Wikipedia article.  Importantly, that pair of curves is conditional on all of the other goods' supply and demand.

For example, if we're looking at the price of cave wheat and off-screen I manipulate the supply of plump helmets, I would expect a change in the demand curve for cave wheat.  How big an effect depends on how closely the two work as substitutes or complements.

The high-dimensional point where the supply hypercurve intersects the demand hypercurve gives you the set of quantities traded and their prices.  The computer sets production to match the supply implied by that point, and trade occurs at those prices.

When you look at the whole set of goods in the economy as one big optimization problem, some of the weirder results in economics actually pop out as perfectly rational outcomes.  Your example about overweight paupers is a version of the Giffen Behavior effect.  If bare sustenance food (which in industrialized countries means highly processed junk because "who has time to cook?") takes up a big fraction of your income, you can't afford to supplement it with healthier food.  As the price of crappy food does up, you actually end up buying more of it because you're substituting out more and more of the healthy stuff (this ends abruptly when the price of crappy food gets remotely close to that of good food).

The Giffen Behavior can only occur in a multi-good market with budget constraints.  You'd never see that in a simple one-good model, and one would probably never think to hard-code it into a quota system, but it emerges naturally from the kind of integrated economy engine we're discussing here. 

Edit: typo.

Even my system admittadly does create a limited demand curve, little as I believe in it the concept; as a result of the fact that as an entity can aquire more of it's total demand it does so, meaning that up to a point as the 'price' goes demand goes up as the entity can acquire more of something without sacrificing demands that are more urgant. 

The mechanical problem with the system, (aside from it's absudity of course) is that it is tautalogical.  If demand is created by a low price and also decides the price then what determines the price or the demand in the first place?  How do we get to the prices first before we decide how much everyone is going to buy without a limit based upon those prices? 

The total amount of labour in the world is finite yet costs nothing at all since they consume the same basic goods employed or employed.  If we create a situation where price falls with supply and demand increases as price falls we will end up with everyone producing a few goods at a very low price, because the goods are not being consumed only stockpiled. 

Ever increasing demand follows an ever lowering price, but the fact that the amount is not being consumed but stockiled means an ever increasing supply and therefore a low price, hence an ever greater demand, hence ever greater stockpiles. As labour is sucked into food production, other things become scarce meaning that demand for those things is also low and nobody has anything to trade for them except the virtually worthless food that they have hoarded because they can.

Basically, this.  There are other basic economic principles that contribute to why the millionaire isn't fat and why the poor person is.  Inferior vs normal goods are a prime example of that.  An inferior good is one where the demand falls as the means to purchase it goes up.  Going back to the idea of food, first, because it was used in the original example and, second, because it really is a clear example of it, just because a millionaire has... well, a million dollars... does that mean that he's going to buy and consume a million dollars worth of ramen?  No... ramen is an inferior good.  If you have a million dollars, you aren't going to eat ramen (at least not to the extent that, say, a college student does); you're going to eat fillet mignot and drink expensive champaign.

The increase in wealth and purchasing power, while it does effect the quantity of food to a certain extent, its biggest effect is on the quality of food.  Rich people and poor people eat different things, so their demand curves are going to be different.

If you accept GavJ's ideas then it does.  People should buy a vast quantity of goods and then stockpile them pointlessly if the price is low enough and they cannot use them.  The example is not meant demonstrate any other basic economic principles at work, it demonstrates that the concept that demand increases infinately as the price to the buyer reduces does not exist. 

Given that Bill Gate's does not eat all the cheap food he should then in his world he should have a vast, ginormous grain mountain somewhere where he hoards all the food he demands because to him it is so cheap.

In my system however we have represent exactly the situation you are describing.  Firstly entities seek to get their basic and capital demands (a woodcutter needs an axe but also food).  And then they start demanding a stockpile on top of that (a group of 10 woodcutters will want another axe in reserve in case they become 11 or an axe breaks) and then they start demanding more advanced goods (beds, clothing, beer etc).

Nothing is going to convince GoblinCookie.  You are just going to end up typing wall of text posts that he misunderstands.  This thread now has almost an entire page filled with long wall of text posts arguing about basic economics.  I don't mind debating, but it is annoying when the suggestion ideas get buried beneath the arguments.

GoblinCookie, I don't see why you couldn't just start a debate thread so we can argue about economics there instead of turning every suggestion thread even vaguely related to economics into a debate.

I did not wish to get involved in a lengthy economics debate either; it is rather stressful.  The problem is that certain others have always insisted on advocating their economic theories wholesale in the place of game mechanics in opposition to my actual game mechanics. While I give detailed mechanics they simply vaguely assert that Theory X is the answer. 

Although if you offer the millionaire 100,000 cases of ramen at way below normal rates like $0.01 a box, he will still be willing to trade, because he will realize that he can sell them for closer to their market value later and become more of a millionaire.

Both personal consumption and potential feasible resale value can contribute to utility. And also other things (like capital/production value if it's a tool, etc.)

Why would anyone hope to gain from trading for something the price of which has crashed if they do not consume the items themselves?  Remember that whatever is being offered in response has a value greater than the items being offered, if it is being offered for $0.01 that is the price it is being offered for AND IT IS ONLY GOING TO GET WORSE!  Yes the price could return to normal levels, but those will likely be comparatively low and there is an indeterminate amount of time before *that* happens as opposed to a definite greater return for buying a small amount of something expensive that increases in value.

So in reality what we see is the opposite; people who are wealthy hoard for investments the most valuable and expensive items that they can find, not the cheapest items they can acquire in massive quantities.  That is why we might have Bill Gates Art Collection but not Bill Gates Grain Mountain.  The reason is that if you buy something that is of a high price as an investment, you are likely to end up on the right side of the bubble.  This is rather a good thing too, because the cheapest things tend to be basic goods needed for survival. 

It is when due to famines for instance that the price of basic goods suddenly goes up that Bill Gates Grain Mountain starts to appear and poor people start to starve to death in large numbers as a result. 

[GavJ]

Goblincookie stuff:

Spoiler (click to show/hide)

The mechanical problem with the system, (aside from it's absudity of course) is that it is tautalogical.  If demand is created by a low price and also decides the price then what determines the price or the demand in the first place?

Neither the demand nor the supply curve rely on knowing final price to draw them -- they are how people WOULD act at ALL prices. Then final price is determined by their intersection point. That's... not a tautology.

The problem is that certain others have always insisted on advocating their economic theories wholesale in the place of game mechanics in opposition to my actual game mechanics.

No, we are insisting on advocating the economics of the real world, endorsed by the vast majority of economists, used successfully for hundreds of years, with thousands of research studies supporting it. We are using this AS game mechanisms since it is a realistic civilization simulator game, and that's how economies realistically work.

Versus your also game mechanics, but based on... I don't know. Goblincookie's dreams and envelope scratchings? I'm not sure where you're getting any of this from, honestly. You seem to be making it up on the fly, which is why everybody is disagreeing with you, because your random untested intuitions simply do not deserve equal consideration to an entire, respected, tested, academic field.

If you want an idea of how this comes across to me, it's sort of like the equivalent of "Well I don't believe in your traditional view of volcanoes. I believe that the planet desires to have a certain quota of lava met, and it keeps a scheduling book somewhere to keep track, and it keeps it in this cute little oak cabinet because reasons, and we should make the game do it just like that, and if we don't, then everythings going to BREAK AND BURN AND FAIL. Oh and no, I can't provide any references for that being how volcanoes work, but it totally is."

Again, does anybody actually have an idea what the next step is for the main OP / trying to design this system economically? It's just gonna keep being a discussion about this silliness until we have a better goal. I'm personally not sure. I laid out pretty detailed algorithm suggestions in the other thread, and don't know where else to go from there without actually coding it up. If nobody has anything we should probly just let the topic sink until/when we do.

[Dirst]

One other point of what was in GavJ's spoiler there, then hopefully we can just talk about other stuff.

Spoiler (click to show/hide)

The supply and demand curves reflect what people would buy at all possible prices.  The demand curve for a site full of historical figures (like a player fort) is simply the sum of all the individuals' demand curves, and those are derived from utility/happiness functions as discussed earlier in the thread.  Non-historical figures are handled abstractly (based on the chance of each person having a particular preference) and summed the same way.  Supply curves for a site are a tad more complicated but answer the question "At this set of prices, what would be the most valuable combination of things we could produce with the resources we have?"  The point at which the demand hypercurve intersects the supply hypercurve gives you the set of prices and quantities for each good.  Calculating the intercept is pretty simple, calculating the demand and supply curves is more intense (the latter needs to take into account the costs of sending caravans and such).

The only special cases you need to handle are import and export bans (setting the demand or supply curve to a vertical line after local needs are met) and places where the supply and demand curves don't cross at all (no one is willing to pay a price that would motivate anyone to make the good), at which point you fix that quantity to zero and move on.  That's it.  The hard part is specifying the building blocks for demand and supply, and the rest is just math.

Again, does anybody actually have an idea what the next step is for the main OP / trying to design this system economically? It's just gonna keep being a discussion about this silliness until we have a better goal. I'm personally not sure. I laid out pretty detailed algorithm suggestions in the other thread, and don't know where else to go from there without actually coding it up. If nobody has anything we should probly just let the topic sink until/when we do.

We could go down the rabbit hole of information asymmetry.  Should NPCs withhold news about supply or demand changes?  Should everything an NPC says be credible to the player or other NPCs?  Would NPCs ever be smart enough to realize when it is in their interests to warn someone about an imminent threat?  Personally, I'm fine with all of these issues getting handled in NPC social interaction AI... NPCs who like you are honest and those who don't are tight-lipped.  An interesting question down the road would be when should an NPC actually lie.  "Oh no, I just came from there, and the Forgotten Beast was slain by some adventurer a week ago.  Go ahead and send your caravan.  The resulting lack of competition in the other towns will benefit me greatly."

I think a more fruitful topic of discussion would be figuring out supply curves, which probably need a way to generate an I/O table from the raws.  Explicit reactions are relatively straightforward to extract, hardcoded ones less so.

[GavJ]

well I think we need the following things for supply curves:

- How paranoid the people are, how much they want to keep on hand in different market conditions before selling (as a scaling factor, not a specific number). In other words, some algorithm for deciding where to draw the line between consumption and sales in the same industries.
- Some measures of basic strategic trading like... not selling armaments to people you were just at war with (The United States doesn't seem to have worked this one out yet, though, lol.)
- We need some representation of tools / means of production, treated like an implied/invisible reagent in reactions that does not get consumed. It could be as simple as just literally "tools" treated equally across all job types. Sort of like how "catapult parts" are all just labeled homogenously in game even though of course in reality they would be different specific pieces.  OR you could actually do tools for different industries specifically.  In the player's fort, this is more like you building your workshops and stuff, but in NPC forts, it can be abstracted to just a number representing their production level and tools.
- We need to of course measure the resources available at a site, which might actually be tricky since worldgen doesn't actually fully generate NPC sites.
- We need some algorithm for determining how much stuff you can gather, based on labor and gathering tools currently.
- We need some algorithm for how quickly you can change your production levels.
- Forts need some better and plausible system of getting information than just the caravans in order to not make terrible predictions about what to produce. I made another thread recently about pigeon or bat messenging, but anything better than what we have now, really.
- We need algorithms for parsing production trees and deciding whether it is more lucrative to sell raw ingredients or finished goods, and what proportions of ingredients to allocate to different recipes and so on. This sounds the most complicated to me, but somebody (I think Dirst) seemed to be implying that there are already well known simple solutions to this.

[Dirst]

- We need algorithms for parsing production trees and deciding whether it is more lucrative to sell raw ingredients or finished goods, and what proportions of ingredients to allocate to different recipes and so on. This sounds the most complicated to me, but somebody (I think Dirst) seemed to be implying that there are already well known simple solutions to this.

The well known simple solution determines the total production required to produce a final set of output.  So it knows that you need iron ore to make iron, and iron to make steel, and can back out how much of each input is needed include how many actions are needed from each profession.

Limitations are: (1) it only deals natively with consumables, (2) substitute inputs are tricky, and (3) the optimization to find the best vector of final outputs is iterative with the market-clearing algorithm.

For (1), the introduction of capital means that the technology matrix has changed.  The ability to use a woodcutting axe or particular workshop is like having a particular labor available, but the costs differ: labor is free to acquire and has an ongoing cost to maintain (food, drink, etc.) while capital has a cost to acquire and no maintenance.  This can work without massive headaches if we stipulate that the market state is updated too often for a capital good I'm making "now" to be used to make final goods in the same period. 

For (2), should there be a difference between a microcline chair and a sandstone chair?  So long as the mix of goods in a category roughly comport with the distribution of preferences in that category, it's perfectly safe to have a single good called "chair" (which in turn is a capital good used for making offices).  We can also lump them together if we're going to let the craftsdwarf's preferences dictate the choice of materials.  But any case that allows more than one set of inputs to create the same output good (for example, billon bars) may prove problematic.

For (3), we need to bootstrap this in some way because the I/O model feeds into the supply curve which is an integral part of the optimization.  It can be calculated from first principles by starting with zero final output, working out the most profitable first unit to produce, and iterating... but that will take a really really long time to reach a decision.  The solution I think is similar to "book openings" in chess programs: there is no point having the full weight of predictive analytics encumbering the first couple moves in a game.  Procedurally generated starting vectors (which then get fine-tuned by the optimization engine) would help.  Maybe a "temperate farming start" and a "mining start" and a "crafting start" and so on as seeds.

I've seen a decent starting point cut down processing time of an optimization from hours to less than a minute.  There's also an entire branch of research into optimizations that give you an answer "at least X% as good as the global optimum" in a tiny, tiny fraction of the time.  Combined, those can bring market clearing into the realm of periodically updating production and prices in game with reasonable FPS.

Those starting points can be procedurally generated, but can't be hard-coded.  Otherwise a single missing plant or modded raw file would break the whole system.

[GoblinCookie]

Again, does anybody actually have an idea what the next step is for the main OP / trying to design this system economically? It's just gonna keep being a discussion about this silliness until we have a better goal. I'm personally not sure. I laid out pretty detailed algorithm suggestions in the other thread, and don't know where else to go from there without actually coding it up. If nobody has anything we should probly just let the topic sink until/when we do.

I have given the question of how value should be determined in relation to the human player and I have come up with this. 

The total of the goods that are on offer and the total amount of goods that are demanded are broken down into a %.  A trade will always happen if both sides recieve an equal % of the total goods on offer, so the 100% trade will always work.  The AI does not to do this (saves on computing power I guess) because it always trades 100% for 100%.  The problem has always been that the human player does not work that way and may simply figure out how to manipulate the system to get everything for practically nothing.

The % is a percentage of a total number sum.  Each item on both sides has a number which may be a fraction which determines the amount that it is worth relative to the other items as part of the % breakdown.  The AI does not use these numbers as prices, instead it uses them to determine if it is being offered a partial deal (less than 100% on both sides) how much % the items are worth.

The numbers should be determined like this.  The amount that the average worker in their society (not yours) would take to produce of a given item in a given time.  Added to that is the amount of labour that was used to produce the raw materials involved in production. 

The total % value of the numbers of the goods you are demanding out of all the good on offer is compared against the % of the goods that you offering out of the total goods that they demand is what is used to determine the trade.  Having a good trader allows you distort the value of the items, raising the total % that the items you are offering are percieved to be worth against the other items being demanded.

The great thing is that we can simply replace the existing dwarfbucks based system with a % breakdown and do the same about each item that is being demanded.  People will figure out easy enough that they have to offer the AI an equal or greater % of their total in order to get a trade.  It also motivates the player to align their own production decisions to the demands of their trading partners, since they will always get everything if they can offer everything.

[Dirst]

Again, does anybody actually have an idea what the next step is for the main OP / trying to design this system economically? It's just gonna keep being a discussion about this silliness until we have a better goal. I'm personally not sure. I laid out pretty detailed algorithm suggestions in the other thread, and don't know where else to go from there without actually coding it up. If nobody has anything we should probly just let the topic sink until/when we do.

I have given the question of how value should be determined in relation to the human player and I have come up with this. 

The total of the goods that are on offer and the total amount of goods that are demanded are broken down into a %.  A trade will always happen if both sides recieve an equal % of the total goods on offer, so the 100% trade will always work.  The AI does not to do this (saves on computing power I guess) because it always trades 100% for 100%.  The problem has always been that the human player does not work that way and may simply figure out how to manipulate the system to get everything for practically nothing.

Up to this point we've mostly been concentrating on NPC-to-NPC trade, because that will be the overwhelming majority of the economy.  Demand curves (based on happiness/utility of individual creatures) and supply curves (based on technology and available resources) can be complicated to derive, but once they exist NPC-to-NPC trade just works.

These curves can be used directly with the player.  For any price you could set for your +rutile mug+, the NPC caravan will have a certain number it would be willing to buy (or value it would be willing to trade)... and for any price you're willing to pay, the NPC caravan will offer a certain number of pine logs.  If the player is selling something common or buying something in high demand, the price wouldn't have to move much before the NPC says forget it because he can trade elsewhere.

This would be valid from a rational^* economics standpoint, but not fun.  It leaves no room for trade skill, haggling, reputation, schmoozing, etc.  Basically, the player can't get a better deal than an NPC could get, and it would take a tedious series of test offers to figure out what deal that NPC would get.

So with dynamic prices in the background and keeping the game fun in the foreground, how should a caravan interact with the player?  The problem I have with your system is that it seems to depend on how much junk the player brought to the depot (to figure out what 100% means to the player).  If I read that correctly, exploits are simple: bring one log to the depot, and trade it for the entire wagon load of caravan goods.

That's probably not how you intended it to work.  Might as well just base the 100% on how the NPC values the stuff.

The % is a percentage of a total number sum.  Each item on both sides has a number which may be a fraction which determines the amount that it is worth relative to the other items as part of the % breakdown.  The AI does not use these numbers as prices, instead it uses them to determine if it is being offered a partial deal (less than 100% on both sides) how much % the items are worth.

The numbers should be determined like this.  The amount that the average worker in their society (not yours) would take to produce of a given item in a given time.  Added to that is the amount of labour that was used to produce the raw materials involved in production.

This is already incorporated into the supply curve.  An NPC normally would not sell an item for less than it cost to make that item.  "Normally" because there might be extraordinary circumstances when dealing with a player (such as a huge difference in trade skill, or a highly profitable trade earlier in the session).  A below-cost trade between NPCs would only occur if we get really clever about attitude manipulation (buying friendliness) or razors-and-blades business models (I'll give you an anvil and supply you coke).

The total % value of the numbers of the goods you are demanding out of all the good on offer is compared against the % of the goods that you offering out of the total goods that they demand is what is used to determine the trade.  Having a good trader allows you distort the value of the items, raising the total % that the items you are offering are percieved to be worth against the other items being demanded.

The great thing is that we can simply replace the existing dwarfbucks based system with a % breakdown and do the same about each item that is being demanded.  People will figure out easy enough that they have to offer the AI an equal or greater % of their total in order to get a trade.  It also motivates the player to align their own production decisions to the demands of their trading partners, since they will always get everything if they can offer everything.

If a trade session only takes the NPC's 100% into account, we could circle back to the existing trade system (pure barter) except with dynamic prices.

The decision is where to insert the broker's trade skill.  It could let the player get an across-the-board price adjustment.  I think we could go a step further and let the broker act as an adviser to the player.  This is because the broker has available to it all of the logic that an NPC trader would use.

The player selects some items to sell, and the broker could suggest what to ask in exchange.  The player selects some items to buy, and the broker could suggest what to offer.

Trade skill weights the broker's surplus calculations away from her own preferences through racial averages and then toward the specific circumstances of the fort and the caravan.  An inexperienced broker overpays for things in her favorite color; a master broker knows that sheep wool is particularly popular in the caravan's homeland (a chance grouping of historical figures' preferences).

"I'll trade you this *microcline figurine of cows* for that ≡iron warhammer≡."
"How about you trade me some of whatever your smoking?"
"But... but... it's blue!  And it has cows!  It's like the embodiment of haunting moos."
"*sigh* Throw in that and that and that and that, then we'll talk."

^*"Rational" is a technical term in economics, basically implying that the actor weighs all possible alternatives and goes with the one that yields the highest surplus for her.  Rational does not mean emotionless, since surplus includes things like happiness.  Rational also does not mean everyone gets together to craft the socially optimal solution that makes everyone best off, since each actor is trying to maximize her own surplus rather than the total amount of surplus in the economy.  Finally, rational actors are typically modeled with a consistent discount factor for future events that prevents them from constantly changing their minds.

(Edit: typo)

[GavJ]

This is already incorporated into the supply curve.  An NPC normally would not sell an item for less than it cost to make that item.  "Normally" because there might be extraordinary circumstances when dealing with a player (such as a huge difference in trade skill, or a highly profitable trade earlier in the session).  A below-cost trade between NPCs would only occur if we get really clever about attitude manipulation (buying friendliness) or razors-and-blades business models (I'll give you an anvil and supply you coke).

Or they may have over-invested in something, only to find out later that all their partners also have the means to make those things or aren't willing to demand as much anymore (and maybe won't ever again), so they're liquidating to cut their losses.
Or they're trying to gain a monopoly by undercutting competitors out of business then jacking prices later (such as Japanese television/VCR manufacturers did).

Probably other reasons too. The monopoly one and the razors and blades we probably don't want to simulate. The liquidation and the buying friendliness we probably do want to simulate.

Friendliness is probably pretty easy to model by just incurring an arbitrary cost to dealing with strangers, especially of other races, at first. Which will just make everyone less efficient and lower trade volume (if I'm mistaken, then change some other variable that does achieve that). Then familiarity brings you closer to rational trading. Screwing them over takes you further away. And offering overly good deals obviously helps. The game already mediocrely does some of these things.

Liquidation is probably easy to do and just to keep in mind while programming supply curves.

It leaves no room for trade skill, haggling, reputation, schmoozing, etc.

Generic haggling skill is easy to model -- much simpler than some intentional misinformation asymmetry thing. All it has to be is that when a trader is trading with somebody of higher skill than him, the value of all the higher skilled dude's stuff gets an arbitrary multiplier (i.e. stuff originating from him. It retains the multiplier once in the buyer's inventory). This only affects in the moment trading dynamics during a caravan visit, not long term strategy.

You could make it so that you have an option to use haggling aggressively or not. If you use it strongly, then when the caravan gets home, they'll realize you sort of fleeced them and it hurts reputation. NPCs may be aggressive with perceived weaker trading partners and less so with stronger ones, or rarer versus more common caravans in their area, or what have you.

Schmoozing and reputation, see above part of the post. NPCs base on similar things like relative threat and how often they see you (it's more worth it to invest in strong friends and return customers)

[heydude6]

This might be slightly derailing but i think something that will help is if we focus on player to caravan trading for a bit. You guys might get a few ideas from this.

Let's go say that the fortress is located very far away from the rest of all potential trade partners. A caravan arrives, and all you do is offer mugs. In RL that caravan will probably never come back because there are better places to buy mugs.

Now let's say that there is a famine in all of the local lands (remember famines haven't been implemented yet) the caravan might come back for your food, with the logic that whatever horrible disease destroyed their food probably hasn't reached your fort yet. And so the only thing they want is food, its at the bottom of the hierarchy of needs.

You give them the food they want.

How will this affect future trading relations?

Now let's say that you decide not give them any food at all but you decide to give them 3 golden statues for all the goods they have on the caravan except the food?what would happen if you gave the statues for free?

What if they were instead statues of food?

How will this effect future trading relations?

I hope this helped

[Dirst]

@GavJ, I was sloppy with my terminology there.  No one would plan to sell below cost unless it was buying friendliness or a razors-and-blades model.  There's another category of goods were the value is affected by how many use it (called network goods) which doesn't apply in DF.  There are reasons to sell at least some network goods below cost.

Coming back to regular non-network goods: once the goods are manufactured, yes they could be liquidated below cost if the market changed.

The monopoly scenario is called dumping, and it would only be effective in a very dynamic economy where firms actually go out of business or at least change their product lines.  DF technology isn't really suited to that kind of strategy, and I agree it's not worth trying to make it work.  (A lesser form of dumping is called limit pricing, where you still make a profit but make it uneconomical for anyone else to enter your market.  Capital goods are so cheap in DF that limit pricing would be virtually impossible even if we wanted to allow it.)

@heydude6, if the site sending the caravan has an acute shortage then its willingness-to-pay for the scarce goods will be really high and really low for everything else due to budget constraints.  It makes sense that the player should give them what they want, but of course the player doesn't have to.

If we want the visiting broker to somehow know what is in the fort but not at the Trading Post, then maybe it would be good if he can request specific things be brought to trade.  This would need to be sensitive to how much the fort needs for itself.  If the visitor requests "surplus" goods and the player refuses, this should harm relations.  If the visitor is requesting the player's own supplies, he should be ecstatic to get anything.

As for taunting the visitors with statues of food, that's probably outside the scope of DF AI for the foreseeable future.

[GoblinCookie]

Up to this point we've mostly been concentrating on NPC-to-NPC trade, because that will be the overwhelming majority of the economy.  Demand curves (based on happiness/utility of individual creatures) and supply curves (based on technology and available resources) can be complicated to derive, but once they exist NPC-to-NPC trade just works.

These curves can be used directly with the player.  For any price you could set for your +rutile mug+, the NPC caravan will have a certain number it would be willing to buy (or value it would be willing to trade)... and for any price you're willing to pay, the NPC caravan will offer a certain number of pine logs.  If the player is selling something common or buying something in high demand, the price wouldn't have to move much before the NPC says forget it because he can trade elsewhere.

This would be valid from a rational^* economics standpoint, but not fun.  It leaves no room for trade skill, haggling, reputation, schmoozing, etc.  Basically, the player can't get a better deal than an NPC could get, and it would take a tedious series of test offers to figure out what deal that NPC would get.

So with dynamic prices in the background and keeping the game fun in the foreground, how should a caravan interact with the player?  The problem I have with your system is that it seems to depend on how much junk the player brought to the depot (to figure out what 100% means to the player).  If I read that correctly, exploits are simple: bring one log to the depot, and trade it for the entire wagon load of caravan goods.

That's probably not how you intended it to work.  Might as well just base the 100% on how the NPC values the stuff.

The caravan arrives already with a fixed amount of goods it aims to purchase from your fortress.  The AI caravan arrives with a list of 100 logs that it wants from you and offers a fixed quantity of goods.  If it offer it one log then it will offer you only 1% of all the goods that it has on offer.  If you offer 100 logs then you get to walk away with the entire caravan. 

It gets worse in that it creates a custom model based upon your trading that it uses to fit you into the system.  If you keep selling it only 1 log, it will conclude that the only surplus your settlement is capable of producing is 1 log, meaning that you will only ever get any traders if everything else is massively overabundant except logs; in which case you will end with vast stockpiles of unsellable items in return for 1 log. 

This is already incorporated into the supply curve.  An NPC normally would not sell an item for less than it cost to make that item.  "Normally" because there might be extraordinary circumstances when dealing with a player (such as a huge difference in trade skill, or a highly profitable trade earlier in the session).  A below-cost trade between NPCs would only occur if we get really clever about attitude manipulation (buying friendliness) or razors-and-blades business models (I'll give you an anvil and supply you coke).

For much less?  Much less of what?  Gold, Silver, Bronze, Food, Water, Beer, Plump Helmets, Wheat, Strawberries?  Remember we are talking about potentially 2000+ settlements here. 

The whole concept of basing a whole economy on dynamic value rather than using it in limited circumstances is a computer's worst nightmare however realistic it may be. We could instead be using my system which based everything ultimately upon a system of exchanges that are precisely based upon finite and measurable values, we are instead going for valuing fluctuating values against other fluctuating values, potentially causing all sorts of memory loops and bugs. 

If a trade session only takes the NPC's 100% into account, we could circle back to the existing trade system (pure barter) except with dynamic prices.

The decision is where to insert the broker's trade skill.  It could let the player get an across-the-board price adjustment.  I think we could go a step further and let the broker act as an adviser to the player.  This is because the broker has available to it all of the logic that an NPC trader would use.

Remember that the AI has already guessed what you should be producing and it is there for a specific reason, to acquire the desired amount of goods.  It offers a subset of goods that it has guessed you should be lacking based upon what you would producing and consuming were you an AI and that IT has in surplus. 

The broker skill is used to inflate the percieved value % of the goods the player is offering against the other goods that are not being offered.  It is used in partial trades when the human settlement cannot or will not give the AI everything on it's list of demands in order to make the goods that are being offered make up a greater amount of the % than would otherwise be the case based solely on their labour value.

The player selects some items to sell, and the broker could suggest what to ask in exchange.  The player selects some items to buy, and the broker could suggest what to offer.

Trade skill weights the broker's surplus calculations away from her own preferences through racial averages and then toward the specific circumstances of the fort and the caravan.  An inexperienced broker overpays for things in her favorite color; a master broker knows that sheep wool is particularly popular in the caravan's homeland (a chance grouping of historical figures' preferences).

"I'll trade you this *microcline figurine of cows* for that ≡iron warhammer≡."
"How about you trade me some of whatever your smoking?"
"But... but... it's blue!  And it has cows!  It's like the embodiment of haunting moos."
"*sigh* Throw in that and that and that and that, then we'll talk."

Cool as those ideas may be, I think the existing trade system is already time consuming enough and does need to be made worse.  In my system the AI's demands, the list of items, each item's accepted minimum quality and the amount of each item that they will accept can be easily be neatly lined up say on top of the screen. On the left of the screen are the items that the AI is offering for sale and on the right of the screen are the items that you are offering to sell. 

As you offer an item your % offered increases on the right of the screen.  As you demand an item the % demanded increases, trades happen if the total you demand is equal to or less than the % you offer.  This is an intuitive system that people will figure out rather easily.  The broker's skill simply acts invisibly to raise the % value of the items actually being offered against the other items that the AI demands.  The player does not even have to know it is doing that, it does not need to understand the mechanics to play the game (how it should be in any game). 

^*"Rational" is a technical term in economics, basically implying that the actor weighs all possible alternatives and goes with the one that yields the highest surplus for her.  Rational does not mean emotionless, since surplus includes things like happiness.  Rational also does not mean everyone gets together to craft the socially optimal solution that makes everyone best off, since each actor is trying to maximize her own surplus rather than the total amount of surplus in the economy.  Finally, rational actors are typically modeled with a consistent discount factor for future events that prevents them from constantly changing their minds.

(Edit: typo)

Thing is the consequences of rationality is rather different for a government dealing with other governments than it is for one private entity dealing with another private entity.  One major reason for this is that governments are non-profit entitities, private entities pursue profits because they can hand these over to their owners and governments have no owners. 

Another second reason is that there is a potential stability to the relationships between government and political incentives to play nice which do not exist in relationship between private entities.  As an example imagine we have two settlements, a human settlement that produces grain and a dwarf settlement that produces metal. 

If the grain producing settlement strikes a 'bad' deal then the dwarf settlement will grow meaning the supply of metal increases in order to meet it's own future demand as it grows.  If they strike a 'good' deal and part with little grain for much metal the dwarf settlement will stagnate and in future they will be short of metal so they will not grow. What makes the regular obsession with profits particularly irrational is that the human settlement has nothing to do with it's grain surplus other than sell it to the dwarves; provided it has a stockpile sufficiant to deal with any crop failiures that may occur.

As unlike with a private buisiness, there is no seperate group owning the settlement that wants to help itself to lots of 'grain profits', which means we will end up with the kind of system I have been advocating.  The dwarf settlement ends up demanding as much grain as it needs+a stockpile and the human settlement ends up demanding as much metal as it needs+a stockpile.

[Dirst]

The caravan arrives already with a fixed amount of goods it aims to purchase from your fortress.  The AI caravan arrives with a list of 100 logs that it wants from you and offers a fixed quantity of goods.  If it offer it one log then it will offer you only 1% of all the goods that it has on offer.  If you offer 100 logs then you get to walk away with the entire caravan. 

It gets worse in that it creates a custom model based upon your trading that it uses to fit you into the system.  If you keep selling it only 1 log, it will conclude that the only surplus your settlement is capable of producing is 1 log, meaning that you will only ever get any traders if everything else is massively overabundant except logs; in which case you will end with vast stockpiles of unsellable items in return for 1 log.

Okay, so you picture the caravan showing up with a strict shopping list.  Offering them anything not on their list would be of 0% value to them, which seems a bit weird.

This is already incorporated into the supply curve.  An NPC normally would not sell an item for less than it cost to make that item.  "Normally" because there might be extraordinary circumstances when dealing with a player (such as a huge difference in trade skill, or a highly profitable trade earlier in the session).  A below-cost trade between NPCs would only occur if we get really clever about attitude manipulation (buying friendliness) or razors-and-blades business models (I'll give you an anvil and supply you coke).

For much less?  Much less of what?  Gold, Silver, Bronze, Food, Water, Beer, Plump Helmets, Wheat, Strawberries?  Remember we are talking about potentially 2000+ settlements here.

Sorry, I don't see what in my post you are referring to.

The whole concept of basing a whole economy on dynamic value rather than using it in limited circumstances is a computer's worst nightmare however realistic it may be. We could instead be using my system which based everything ultimately upon a system of exchanges that are precisely based upon finite and measurable values, we are instead going for valuing fluctuating values against other fluctuating values, potentially causing all sorts of memory loops and bugs.

There is a trick used in economic modeling called a numéraire.  It is some tradeable good that has a fixed price, and everything else is priced relative to that.  The only items that MUST be present in the game are the 19 hard-coded materials including water, magma, and ash.  I think raw green glass would make a decent numéraire, and with an internal fixed price of 100 there would be plenty of range for prices while staying with integers.

(Edit: Note that the player doesn't need to see things in numéraire units.  Maybe at some point in time a silver coin can buy 8.7 chunks of raw green glass.  It would be simple to display all prices in "silver coins" by dividing the internal prices by 8.7.  Hell, you could even let the player choose the units. "The Overseer has decreed that our new unit of currency is the *obsidian bracelet*.  No, I don't know why.  By the way, does anybody have change for a +steel breastplate+?")

Remember that the AI has already guessed what you should be producing and it is there for a specific reason, to acquire the desired amount of goods.  It offers a subset of goods that it has guessed you should be lacking based upon what you would producing and consuming were you an AI and that IT has in surplus. 

This wasn't clear from your previous post, but I got that from what you said above.  I still think the inability to go "off-script" is unnecessarily limiting.

The broker skill is used to inflate the percieved value % of the goods the player is offering against the other goods that are not being offered.  It is used in partial trades when the human settlement cannot or will not give the AI everything on it's list of demands in order to make the goods that are being offered make up a greater amount of the % than would otherwise be the case based solely on their labour value.

Cool as those ideas may be, I think the existing trade system is already time consuming enough and does need to be made worse.  In my system the AI's demands, the list of items, each item's accepted minimum quality and the amount of each item that they will accept can be easily be neatly lined up say on top of the screen. On the left of the screen are the items that the AI is offering for sale and on the right of the screen are the items that you are offering to sell. 

As you offer an item your % offered increases on the right of the screen.  As you demand an item the % demanded increases, trades happen if the total you demand is equal to or less than the % you offer.  This is an intuitive system that people will figure out rather easily.  The broker's skill simply acts invisibly to raise the % value of the items actually being offered against the other items that the AI demands.  The player does not even have to know it is doing that, it does not need to understand the mechanics to play the game (how it should be in any game).

It can't be worse than the current system, because it gives you all the options you have now plus an additional "broker's suggestion" available to the player to take some of the tedium out of balancing out a thirty-seven item trade.

Your system would be intuitive if the caravan's shopping list is visible to the player, but it restricts trade to just what the caravan wants.  When it comes down to a choice of making the player happy or the caravan happy, it's better to side with the player.  For one thing, caravans never donate to Bay12!

Thing is the consequences of rationality is rather different for a government dealing with other governments than it is for one private entity dealing with another private entity.  One major reason for this is that governments are non-profit entitities, private entities pursue profits because they can hand these over to their owners and governments have no owners. 

Another second reason is that there is a potential stability to the relationships between government and political incentives to play nice which do not exist in relationship between private entities.  As an example imagine we have two settlements, a human settlement that produces grain and a dwarf settlement that produces metal. 

If the grain producing settlement strikes a 'bad' deal then the dwarf settlement will grow meaning the supply of metal increases in order to meet it's own future demand as it grows.  If they strike a 'good' deal and part with little grain for much metal the dwarf settlement will stagnate and in future they will be short of metal so they will not grow. What makes the regular obsession with profits particularly irrational is that the human settlement has nothing to do with it's grain surplus other than sell it to the dwarves; provided it has a stockpile sufficiant to deal with any crop failiures that may occur.

As unlike with a private buisiness, there is no seperate group owning the settlement that wants to help itself to lots of 'grain profits', which means we will end up with the kind of system I have been advocating.  The dwarf settlement ends up demanding as much grain as it needs+a stockpile and the human settlement ends up demanding as much metal as it needs+a stockpile.

An actor is considered rational if it is maximizing an objective function.  For a firm, the objective function is profits.  For an individual, the objective function is utility (roughly DF's "happiness" metric plus some measure of wealth).  For a government, the ideal objective function is called "social welfare" which is the sum of all profits and utilities under that government's jurisdiction.  Real governments are not ideal, though up to this point we've been carrying around the idea of an ideal government as an implicit assumption.  By the way, real firms aren't ideal profit-maximizers either because the managers within them are maximizing their own utilities.

A site within DF is fairly close to an ideal government because the site basically owns all of the capital and employs all of the creatures.  This varies from a "firm" in that the site cares about individuals' utility as well.  We can model selfish nobles by overweighting their personal utility in the "social welfare" calculation, and to a lesser extent the utility of the noble's friends and family.  We'll get something a bit more nuanced than a production mandate and an export ban, which makes for a more interesting game.

[GavJ]

The AI caravan arrives with a list of 100 logs that it wants from you and offers a fixed quantity of goods.  If it offer it one log then it will offer you only 1% of all the goods that it has on offer.  If you offer 100 logs then you get to walk away with the entire caravan. 

This is an overly clumsy way of trading. They need to re-evaluate all of their trading once they arrive and get an updated true list of what you own. Otherwise they would likely be giving themselves a bad deal for no good reason.

For instance, if the trade partner has more logs than they thought, and thus the partner values logs less than expected, then they would likely demand more logs than they expected to demand for the entire caravan.

Or if there are NO logs, then it would usually be dumb to just leave without trading, since in most cases, there's probably something else they have instead that you'd be still willing to trade for, just not quite as profitably as logs (depending on your whole network of other trade partners, etc. All of which get taken into account).

Basically, the moment your caravan arrives, you re-calculate your civilization's trade strategy, using the new information you just got updated with, and constrained to what you have on hand, and then make up new demand curves and THEN trade based on that.

For much less?  Much less of what?

Value. It is an intangible equatable measurement that spans across all goods. It is what the Y axis basically measures in both supply and demand curves.

Value in dwarf fortress is measured in "dorfbucks" which does not correspond to any coins or anything in particular. It's just wanting something X versus Y amount.

If you do happen to have a currency, then you can map people's values to a certain amount of currency just to make it easier to talk about, but it's not necessary. The true measurement is just psychological value.

Cool as those ideas may be, I think the existing trade system is already time consuming enough and does need to be made worse.  In my system the AI's demands, the list of items, each item's accepted minimum quality and the amount of each item that they will accept can be easily be neatly lined up say on top of the screen. On the left of the screen are the items that the AI is offering for sale and on the right of the screen are the items that you are offering to sell. 

I agree with goblincookie here that for player-NPC trading, it would be really annoying and unnecessary to have the guy suggest random sets of things. That's not how people actually trade.

However, items don't just have fixed values either, as goblin is suggesting -- the value depends on the quantity being traded, so you can't have fixed advertised values. Instead, they dynamically update.

Let's say I want to sell logs for breastplates.  On each side of the trade screen, I click a quantity of items that I'm proposing to buy and sell, respectively. It initially displays the price that it would be to add the first one of that item to your current deal.

As I click higher numbers of breastplates I want to buy, the "price" (measured in intangible dorfbucks) may or may not change for them, depending on the NPC's supply curve. If it does change, it will go up with higher numbers.

As I click higher numbers of logs, the price the NPC is willing to pay will go down (or not change), according to his demand curve.

At any point where the sum of prices is acceptable, you can hit trade and have it go through. (this doesn't reset the prices back. They continue to act dynamically as if it were all done in one trade. You can't trick them by just doing 10 small trades, obviously).

"Acceptable" means different things, depending on how much strategy we are coding in. If the NPC wants to buy friendliness from you, for instance, he might be willing to take a slight loss early in your relationship (for example if you are militarily much stronger than them).

In a zero strategy situation, "acceptable" would just mean "equal prices on both sides, or a better deal for him, only" is accepted.

One major reason for this is that governments are non-profit entitities, private entities pursue profits because they can hand these over to their owners and governments have no owners. 

No, governments are for-profit as well in an economic sense. Everybody is always for-profit, there is no such thing as non-profit entities, except maybe literally insane or dumb people.

When you hear "not-for-profit" what it really means is "not for maximum capital dividends to shareholders" which is very different than "not for maximizing economic utility."  Governments still want to maximize utility, but it's just that they have different values and utilities than a corporation does. For instance, a government places economic value on security of its citizens, often even if that comes at a currency deficit.

That doesn't mean that they aren't maximizing value profit. It just means that the non-currency value for them may outweight the role of currency value.

[Scruiser]

For much less?  Much less of what?

Value. It is an intangible equatable measurement that spans across all goods. It is what the Y axis basically measures in both supply and demand curves.

Value in dwarf fortress is measured in "dorfbucks" which does not correspond to any coins or anything in particular. It's just wanting something X versus Y amount.

If you do happen to have a currency, then you can map people's values to a certain amount of currency just to make it easier to talk about, but it's not necessary. The true measurement is just psychological value.

   I think one of the issues is that price and value should be separate (both for the purposes of our discussions and as a finalized in game mechanic).  Value could be hard coded like it is now, just with more categories to cover use cases.  Value modified by individual personalities/preferences, civ values/preferences, and site needs determines demand.  Price is then determined by supply and demand, with something similar to the schemes you have suggested earlier in this thread and in other threads.

Again, does anybody actually have an idea what the next step is for the main OP / trying to design this system economically? It's just gonna keep being a discussion about this silliness until we have a better goal. I'm personally not sure. I laid out pretty detailed algorithm suggestions in the other thread, and don't know where else to go from there without actually coding it up. If nobody has anything we should probly just let the topic sink until/when we do.

Anyway to practically implement, there are several steps Toady can take:
    1st to implement some more categories of value.  I think separate numeric values materials would be good. For example VALUE_BLUNT_WEAPON, VALUE_EDGED_WEAPON, VALUE_ARMOR, VALUE_AESTHETIC_WORN, VALUE_AESTHETIC_FURNITURE, VALUE_OTHER.  This way, you could directly tell the game AI in the RAWs that steel is valuable for edged weapons, blunt weapons, armor, but isn't particularly beautiful (by default, creature preferences could override of course).  Conversely gold would be worthless for edged weapons and armor, okay for blunt weapons, and highly valuable for jewelry and furniture.  As an added bonus, the AI will finally know how to priotize metals in armor and weapons.  Items could also have separate value categories.  A sword and axe may have equal VALUE_WEAPON, but an axe would have an additional category of VALUE_TOOL since it is used for wood cutting while the sword would have a higher VALUE_WEAPON_AESTHETIC (again individual preferences and civ preferences could override this).  This part should straightforward to implement
   2nd implement individual modifiers to value.  Having a preference for gold acts as a multiplier on the base value number being used.  This is pretty straightforward to implement (if it is already handled).
   3rd implement site usage of raw materials and tools and reagents to actually produce final goods.  (Right now a dwarf civ can produce steel without having access to iron).  This is mostly straightforward, although some changes will need to be made to handle the initial bootstrapping of tools (1st Anvil problem) and other similar issues.
   4th implement site AI that tracks what it needs to produce/buy to sustain itself and increase production (both raw materials and capital.)  This starts to get tricky here, but it is still doable, especially if a good job was done in step 3.
   5th Combine all the work of all these step and you now have all basic elements to get supply and demand curves.  This is really tricky.  It could be handled by aggregating individual demands, and then adding in site demands as a separate category.  The end result will need to be theoretically extensible to handle both free and totalitarian societies, and eventually weird cases like hive minds.  I think that for dwarf civs and similar cultures, you have the site purchase all the critical capital, weapons and armor, survival quantities of food, and most of the non-critical capital, then you aggregate the population and treat it as though individuals are buying extra food, commodities, luxury items, and capital for personal enterprise.
   Other misc stuff: add in trade route length as a factor; progressively adjusting the AI to work with more realistic information (start with perfect information in the first iteration, then less information with each improvement); personalities and bargaining as factors in trading; outright manipulation of information (requires AI to have incomplete information)

Thoughts?  Do you guys at least agree on steps 1-4?  I think step 5 is where our discussions usually break down, both because it is the hardest, and because it is in the future at difficult to guess and what leads up to it.

[GavJ]

I think one of the issues is that price and value should be separate (both for the purposes of our discussions and as a finalized in game mechanic).  Value could be hard coded like it is now, just with more categories to cover use cases.  Value modified by individual personalities/preferences, civ values/preferences, and site needs determines demand.  Price is then determined by supply and demand, with something similar to the schemes you have suggested earlier in this thread and in other threads.

Hardcoding is specifically what this whole project is about getting away from...

What is the "inherent value" of a log, outside of it's practical usefulness (i.e. site needs) and one's personal preferences for it (happy thoughts)? Nothing, there isn't anything else left beyond those things. The remaining value is zero after that for everything, i.e. not actually a variable.

Gold is only valuable for people's personal preferences for it. Steel is only valuable for site needs for it if/when they are actually expecting war and need more armament or upgraded armament. Etc.