Tuesday, May 17, 2022
All models are wrong
In order to better understand complex realities, people frequently use models. Some models are very complex by themselves, but more frequently people use much simpler modeling tools, for example spreadsheets. A spreadsheet with all the income and expenditures is a simple model for the business venture it represents, and a lot easier to understand. However, that spreadsheet is most certainly wrong. If you had two similar businesses, e.g. two coffee shops in the same area, spreadsheets could tell you that one of them was doing better than the other. But chances are that the spreadsheet can't tell you why, because that might depend on complex factors that aren't in the model, e.g. one of them having friendlier staff because the boss is treating his employees better. The very reason we use models, to simplify things and make them easier to understand, always result in the model stripping away complexities, and thus not being ablet to explain everything.
Prosperous Universe is a game that can be played with a rather small number of clicks and keyboard entries per day, if you consider only the actual game client. But the list of community resources has a lot of tools and spreadsheets which you can use to model your virtual company, with the goal of optimizing it. I'm using the first spreadsheet on that list, called Rain's Master Base Planner. You enter the data of your base, like what buildings you have, what recipes you are producing, and what efficiency bonuses you have from various sources, and the spreadsheet tells you how profitable your base is. So when I have some cash accumulated, I can use the spreadsheet to simulate an expansion: Should I buy another water rig, one more farm, or rather a food processor? Depending on my situation and the market prices, the spreadsheet will tell me the best option. A good part of "playing the game" thus happens on the spreadsheet, not in the game client.
After some hits and misses using the spreadsheet to plan bases, I realized why in game some of the base builds that the spreadsheet said were fine didn't actually work out all that well in the game: The model of the spreadsheet was based on market prices, as if those markets had infinite capacity and frictionless trades. Apparently the previous "universe" actually had commodity exchanges on planets, so people on these few select planets could directly trade between their base warehouse and the market. And if you are a new player, and make a relatively common product in small quantities, your trades aren't going to have a huge impact on the market price. So the spreadsheet "worked" as a model to some extent.
In the current universe, commodity exchanges are now on orbital stations, so even people on planets in those systems have some shipping cost. And the further away you are from the 4 main commodity exchanges, the more important shipping becomes. You can actually design a base full of PP1 prefab plants a few jumps away from the next market, and the spreadsheet would tell you that it is very profitable; only that the amount of iron and limestone you would need to import is actually more than you can handle with your two ships, so the base isn't really working. You can also, on the spreadsheet, design a very profitable base mass producing some rare item; but if you actually did that, you'd find that there isn't all that much demand for that rare item, and your real profit would be much lower, or even non-existent. The spreadsheet also uses prices as they are right now, so unless you start overriding those market prices manually, market risk of changing prices isn't well covered.
Once you understood the limitations of the model, you can still put the spreadsheet to good use. You just need to look out for more things than just net profit: Import and export volumes, an "off sheet" estimation of shipping cost, manually putting in some alternative prices to check sensitivity to price variations, and the like. But the question I am asking myself is how limiting it is to the model that intermediate goods are considered to be at market price. That is not something that a real world company necessarily does: It is often much easier to treat intermediate parts as being worth whatever it did cost to make them, rather than being worth what the market would pay for them.
One example in Prosperous Universe would be the production chain leading to padded work overalls PWO. Now PWO are a so-called luxury consumable, but for the most basic form of workers, pioneers. Thus a very large number of people use PWO to increase the efficiency of the workforce. Only that if you run the numbers in a spreadsheet, the production of PWO from cotton fabric COT is not profitable at all. However, that is because there is no liquid market for cotton fabric. It seems most people who make cotton fabric, a material you need more advanced workers (settlers) to make, also have the less advanced building that turns COT into PWO. One would arrive at a better model for the profitability of the whole chain by treating each intermediate (water -> raw cotton -> cotton fabric -> padded work overall) as costing only whatever labor cost goes into them. By treating the intermediates as "cheap", all the profit of the overall chain appears on the last step, the production of PWO, which is the good you actually sell.
My observation in Prosperous Universe is that when I am trying to find a production that involves settlers using the regular base planner spreadsheet, none of them are really profitable. Which is counterintuitive, because obviously there are a lot of final products around that involve settlers in one of the steps. I just haven't figured out how to make that more clear in the spreadsheet / model that I am using.