Bestia’s economy is entirely player-driven: players find the gold, mint the coins and set the prices. That works right up to the point where NPCs start handing out money — quest rewards, purchases, world-event payouts — because an NPC that pays out gold it never earned is an inflation faucet with no bottom.

The World Treasury Director is the server-side system that closes that hole. It is a single authority per world incarnation that owns the NPC side of the money supply, and it exists to enforce one rule.

The Conservation Invariant link

If an NPC pays out gold, some other NPC must have earned it first.

There is exactly one faucet in the world — a player turning raw gold into coins with the Minting skill — and everything else is a transfer or a sink. NPCs are not a faucet. They hold a share of the money supply, they move it around, and the Director makes sure the share only grows because players sold them something, never because a payout was invented.

Concretely, the Director maintains a single NPC gold pool for the world — the world reserve — and every NPC-side payment is an entry against it. How large that pool is, and why, is derived from the world’s own gold deposits on the Money Supply page: half the supply is minted into NPC hands when a world is created, and the rest is still in the ground waiting to be mined.

Sinks return to the reserve rather than deleting money: the auction listing fees and the commission posting tax take coin out of circulation and put it back in the world reserve, where it re-enters slowly through the settlements.

In the short run that is indistinguishable from burning — the coin is gone from player hands and stops pushing prices up. Over a world’s lifetime it is not. A burn makes the supply monotonically decreasing, and since minting is bounded by the finite gold in the ground, a world that burns has an ending it was never designed to reach.

Treasuries link

A “treasury” is a named sub-account of the NPC gold pool, not a separate pot of magic money:

• Settlement treasuries collect the NPC share of local trade and fund local work — defending the town, clearing the roads, the Defend the City! type of event. They are also what a settlement pays its imports and its rebuilding with; see Townsfolk & Settlement Simulation.
• Faction treasuries collect sacrifices and temple income and fund faction-aligned work.
• The world treasury is the remainder, and it is what backs rare large-scale events when no single settlement could pay for them.

A treasury can run low, and when it does the world reacts honestly: a poor village genuinely cannot afford a big bounty. This is a feature. It gives regional wealth a visible consequence and it means players who have been trading with a settlement have made that settlement better able to pay them later.

Budgeting Payouts link

The Director does not simply pay whatever a quest generator asks for. Each payout is checked against a budget derived from the pool’s health:

  pool_ratio      = npc_held_gold / configured_supply
payout_modifier = clamp(pool_ratio / 0.50, 0.25, 1.5)
  

0.50 is the target share — the half the pool is seeded with. A pool that has drifted above target pays generously and drains back toward it; a depleted pool tightens up until player trade refills it. The clamp keeps both ends survivable: NPC work never dries up completely, and a flush treasury cannot triple wages.

The modifier is a preference; the reserve is a constraint. payout_modifier shades what NPCs are willing to pay, but nothing may pay out coin the reserve does not hold, whatever the modifier says. An empty reserve means settlements stop being topped up and shops start refusing to buy — which is not a failure state, it is the pressure that makes mining worth doing.

Two further throttles apply on top:

• Daily redistribution. The reserve pays out to settlement treasuries once a game-day, weighted by population, so a city receives more than a hamlet. This deliberately ignores real-world economics; it exists so coin keeps circulating into player hands instead of pooling wherever it happened to land. Weighting by population rather than splitting evenly is the one concession to realism: a city of twenty thousand does more trade in a day than a hamlet does in a season.
• Regional repeat decay. Repeated world events in the same region pay a diminishing share, so a treasury cannot be farmed by players who manufacture the emergency they then get paid to resolve.

Price and Wage Feedback link

The same pool ratio feeds what NPCs are willing to pay and charge, which is the mechanism that keeps inflation in check without an invisible hand:

• What NPCs pay for resources and items scales with payout_modifier. When money is abundant on the NPC side, selling to NPCs is attractive and gold flows back out of the pool.
• What NPCs charge scales inversely. When the pool is thin, NPC goods get expensive, players buy from each other instead, and NPC coin is conserved.
• Per-NPC honor modifiers are applied afterwards, so a player’s reputation still moves their personal prices on top of the world-wide baseline.

The result is a soft, self-correcting band rather than fixed price tables — and, importantly, no scripted “gold sink event” is ever needed to rescue the currency.

Prior art link

Almost every decision on this page has been tried by somebody, and two of them have been tried and failed. This section records which games we borrowed from, and — more usefully — which failure we are deliberately walking back toward.

The design we are repeating, and why it failed link

Ultima Online’s first economy was, almost line for line, the one described on this page: NPC shops that adjusted price and stock from local market data, bought speculatively from players, and differed between towns so that hauling paid. Raph Koster’s retrospective records what happened to it. Skills advanced by repetition, so crafters mass-produced goods they did not want in order to level, dumped them on shops, and shop buy prices collapsed to nothing. The fix was to periodically wipe the shops’ market data — which is to say, to give up on the simulation.

The lesson is not that the price model was wrong. It was that a progression system elsewhere in the game fed the economy an input it was never designed to absorb. Our defence is in three parts, and all three are already built: prices move per unit rather than per transaction, so a large dump moves the price against itself; a warehouse ceiling caps what a town can physically hold; and a town refuses rather than paying with coin it does not have. What is still missing is Wurm’s per-player payout cap, and it should land with the first NPC that buys anything.

The same retrospective records the other failure mode, and it is the one that actually threatens a closed supply. UO’s currency was closed and held its value for three years. What eventually broke it was not inflation but hoarding — gold flowed to a small number of players and stopped moving, and the game needed an in-fiction campaign to shake it loose. A fixed supply does not fail because there is too much money. It fails because the money stops circulating.

What we took from Wurm Online link

Wurm Online is the only MMO to have shipped a genuinely closed currency at scale, and the three-tier purse on the Money Supply page is its design. Two details are what make it survivable rather than merely closed: an NPC trader holds a finite purse that refills over time, and a player can only earn so much from one per hour. The first makes “the shop is out of coin” an ordinary event instead of a crisis; the second is what stops the whole reserve draining into whoever grinds hardest.

The smaller borrowings link

• RuneScape prices from a base stock: each item has a quantity the shop wants to hold, the price moves with the distance from it, and stock drifts back one unit at a time. Worth knowing that OSRS eventually made shop stock per player to kill queue contention and bot camping. We do not need that, but it is the escape hatch if D6 ever fires.
• EVE Online seeds market orders for goods no player can supply yet, then stops refreshing them and lets them expire. Seeding is a bootstrap with an exit, not a permanent fixture, and a seeded price is deliberately a floor. That is exactly the posture our starter consumables should take.
• Albion Online’s Black Market posts buy orders that climb until somebody fills them. An NPC expressing demand is much safer than one holding stock: it cannot be flooded, and it cannot be bought out.
• Star Wars Galaxies made gear decay irreversibly, so crafters had permanent demand. No other game copied it, and no other game’s crafting economy lasted as long. A player-driven economy needs an item sink at least as strong as its gold sink, or every crafted good is made exactly once.

What we turned down link

Path of Exile deleted currency altogether: its money is a consumable crafting reagent, so the sink and the use are the same act and inflation cannot happen. It is the most elegant solution to the problem this page exists to solve, and we are not taking it, because a world without coin has no shopkeepers, no treasuries and no reason for a town to be rich. The whole texture we want — a merchant who runs out of money, a wealthy city that pays better — depends on money being a thing you can run out of rather than a thing you spend on crafting.

Further resources to read:

• Gold sink — the vocabulary, and a survey of what other games drain with
• AGC: MMO economies — why faucets outrun fixed drains
• Market Interventions in a Large-Scale Virtual Economy — measured effects of a transaction tax and an item sink in Old School RuneScape

Dangers to watch link

Every entry below is a way this design is known to fail, taken from a game it actually failed in. Each names the symptom, the number that would show it, and the lever we would pull.

The measurement is the deliverable, not the lever. The Old School RuneScape study cited above found that a transaction tax barely moved trade volume, an item sink inflated luxury prices without reducing volume, and the illicit gold market shrugged off both. Interventions in virtual economies do not behave the way the design intuition says they will. So every row here is instrumented before anything is tuned.

Three of these deserve more weight than the others. D1 is the failure that actually killed a shipped game. D2 is new — it is created by closing the loop, and no game we borrowed from has it in this form, because none of them ration NPC income against a world-level pool. D10 is not a bug at all but an unmade decision, and it has to be made before mining ships.

Observability link

Because the whole economy hangs off one invariant, it has to be measurable at runtime. The Director exposes, per world:

• Total money supply, and the amount minted, still unmined, and held by NPCs
• Pool ratio and the current payout_modifier
• World reserve balance, and its days of cover at the current payout rate
• Flow rates per category (NPC purchases, broker cuts, quest payouts, world events, fees recovered)
• Per-region treasury balances and event-decay counters
• The metric named against each entry in Dangers to watch

A drift in the money supply that minting does not account for is a bug, and this is where it becomes visible.