So watching/reading the latest update on the economy, a couple thoughts come to mind about potential problems and solutions. Essentially, a pure supply/demand AI will create unavoidably inefficient allocations of resources because of a lack of temporal knowledge- that is, a memory of price fluctuations and the ability to predict future values. In the real world, the reason there aren't often massive changes in the prices of commodities is that if the valuation of a good falls below it's estimated 'worth,' speculators will buy it up in anticipation of turning a profit when the market recovers, and the price will then go up.

Similarly, in the video, the price of Vitrium crashed to 1/unit while Josh was mining. What should ideally happen in this situation is a frenzy of AIs buying it up in anticipation of a market resurgence- otherwise, the smartest player strategy will likely be to do just that, taking advantage of the AI's lack of knowledge about the past/future. If you model this seamlessly enough, you wouldn't get massive crashes/booms at all, the volatility over time would be significantly reduced, and the overall model would be much more realistic.

Of course, this requires that AIs have some ability to model a good's value based on its historic price over time; this isn't that hard mathematically, but I don't know what your performance constraints are. Just a thought!

It also requires storage, and storage should have a cost associated with it. That's one way to impede rampant speculation!

Speculation isn't necessarily a bad thing, if you can sufficiently penalise people who speculate *badly*.

One of the rules of economics sims is that "fun to play as a capitalist overlord" and "fun to live in as a mere lackey" are often at odds. If it were the real world I'd be arguing for counter-cyclical policy, stability, a robust welfare state, and other things designed to take the power away from capitalists, especially the power to bubble and bust and leave tiny shards of hitherto functional economy all over the metaphorical floor. But, it's a game, so that sort of stuff is fun!

I am pretty sure Josh mentioned the volatility was entirely because the "world" was new and demand was high, filled, then high again... he believes (or has maybe tested already? wasn't clear in the video) the market will stabilize after some time has passed when factions start to get their supplies and structure road maps worked out...

The initial booms/busts are based on "oh crap we got nuttin... and need all this stuff to do this other stuff". So what you were seeing is a very limited market because of limited resources (not in volume but in number of types) and limited activity availability (limited list of things to accomplish). it was a small scale example of where the game is going in that (economic) area.

I understand what you are getting at, but at the same time determining an outcome based on a 20 minute activity in a newbie universe is unlikely to lead you to a realistic conclusion.

He also said that the NPC supply/trading algorithms were very very basic. A lot of that volatility came about because none of the NPCs were paying attention to things like rolling price averages, forward curves, or expected future demand. Slow the simulation down and give the NPCs more analysis tools, and a lot of the volatility should smooth out.

Edit: for example, if you or I looked at that market, our first instinct would be to just buy out all the Viritium on the market at $1 until the price changed. Then we'd set all our sell orders at $5 and wait for them to sell - which they would because the NPCs have no way of telling where the price is likely to go, but we do.

Note that Josh hasn't implemented an agent (at least to my knowledge) that engages in speculation. Furthermore, the system is vastly sped up for the purposes of the demo, which increases volatility (quadratically, I think). Setting parameters to more "realistic" values, as well as damping volatility by giving the planet an "unsellable" reserve (commodity in storage, in other words) will solve both the volatility problem and the violation of the "law of one price" (the fact that the markets for the station and the outpost are vastly different).

In fact the market shown highlights the need for market makers (people who engage in exactly this -- buying low, selling high). In most games, the player is the only one; there's no reason why NPCs can't be market makers also. Of course, making market making/arbitrage instantaneous also kills the fun factor -- OK in real life, where the markets are controlled by machines with sub-microsecond latencies (and even violates the speed of light on occasion, sort of); not OK in video games.

Overall, this update delivered everything I was looking for, and more. I have a particular interest in this topic (check my post history).

which they would because the NPCs have no way of telling where the price is likely to go, but we do.

Past performance is not indicative of future performance. ~Every mutual fund prospectus ever.
That said, NPCs could easily rely on a concept of "history" to identify value. Additionally, if there are other assets that depend on the ore, "anyone" could backcalculate a "fair" price for viritium based on the value of its outputs (in fact, people take advantage of this in real life -- it's called arbitrage).

This problem could be solved very easily.

Josh has mentioned that NPCs will have personalities which determine what sort of activities they are interested in; soldiering, climbing the corporate ladder, etc.

Wouldn’t it make sense then to have NPCs which due to their personality would be inclined towards speculating.
This could also have the added effect that instead of rapid and unstable fluctuations you might get the bubble scenario in which prices steadily climb over a few “years” and then they fall of the cliff, causing a depression.

But this depends on how smart the AI is or how much it is willing to gamble.