The market system is based on rarity, a calculation derived from the local availability of a good and the total available amount of that good (or at least, the amount that is reachable). Goods are arranged hierarchically into stages, from "raw" materials all the way up to complex goods which require many precursors which themselves are often manufactured.
The system I've written to handle these complex goods works great. I can add "recipes" to a master file and calculate a price for any good I can dream up (and properly research).
A core concept of the system is that of a reference, which represents the amount of a good which can be purchased by 1500 oz of gold (approximately 3,125 gp). For small goods, like crops, a single reference can represent tons of material.
The question I am currently working through is: how to take the bits of information I have and translate them into an availability metric.
The market and price list provide two important functions to the party: they generate striving (to aspire towards a piece of equipment they can't yet afford) and scarcity (equipment that is not available for any cost).
We can, of course, allow the price itself to perform both of these functions. However, all equipment should not be available at all times simply because it exists somewhere in the system. The question is then how to determine how often a good will appear in a given market.
My initial thought is to use the local number of references to determine this rarity. If less than a single reference exists in the local market, then there is a good chance it'll not be available that week (or similar period of time). This means that we can recursively build up a local reference amount for manufactured goods based on the local references of its components.
Once I have this number, I can either use it as a probability straight up, or mitigate it a bit by comparing it to the rarity constant (itself a function of the size of the market). For now, I'm simplifying the final product by taking the log of that number and using that to determine a "rarity score."
\[\mathit{rarity} = \left\lceil-\log_{10}\left(\mathit{ref}_\ell\right)\right\rceil\]
Example: for a good with a 0.002 local references ($\mathit{ref}_\ell=0.002$), the rarity score is 3. That can be further used for a roll as needed. We could say that an item with a rarity of 3 will be available in a market on a 1 in 1d30, for example. This makes rare items difficult to find, but not impossible. The party could either hang around until some were available, or travel to another (and perhaps more favorable) market.
The simplest method to determine local references is to use the minimum of all constituent components, without regard to the amounts needed. I'm not super comfortable with this, because my conception of references is tied more tightly to actual production numbers (which I generate from the working population availability). But it has led to some bizarre cascades. Fodder, for example, is very cheap, but a lot of it is needed to raise a foal into an adult horse. This also means that the number of fodder references needed per horse is very very low. Both of these can quickly get out of hand as very small or large numbers are passed through several iterations of the recipe parser.
I tried dividing local references by the amounts needed, or comparing the local references available to the number needed. Neither are really satisfactory. I think I will use the simple method described above, and if I get a lot of complaints in play then I will revisit it.
The recipe system is built and effective. There are a lot of recipes to write, however.
