The Economy Simulation page states the rule: NPCs never conjure money. This page answers the question that rule implies but does not settle — how much money is there in the first place?

Bestia’s answer is unusually literal. The world generator already places gold deposits with real tonnages, and those tonnages already reconcile with the ore voxels a player can actually break. So the money supply is not a tuning constant somebody picked. It is the amount of gold in the ground, and this page derives it.

Sizing the supply link

A fixed money supply has to be sized against something, and the honest anchor is how many players the world is meant to hold. Everything else follows from that.

Genesis — the 128 km world zone-server ships with — targets 300 concurrent players. That sets a total supply of 300,000,000 coins, half of it minted into NPC hands when the world is created. The calibration constant is therefore:

  1,000,000 coins of supply per concurrent player
  

Working from a player count rather than toward one matters. Sizing the supply first and asking afterwards how many players it supports is how a world ends up with either five players’ worth of coin or a thousand times more than it will ever need.

What the generator already fixes link

Three numbers come out of worldgen and are not ours to choose:

That last figure is the important one: 857 ore voxels make a ton of metal, and the generator enforces the relationship to within 0.1% (Invariants.kt, the tonnage tolerance check). A deposit’s stated tonnage and the number of ore blocks a player can dig out of it are the same fact expressed twice.

Genesis holds roughly 35 t of gold — 24 t of lode, because the three-deposit floor dominates the 8 t the abundance alone would give, plus about 11 t of placer traced downstream of those lodes. That is ~30,000 ore voxels.

The one free parameter link

Thirty thousand voxels have to mint three hundred million coins. That fixes everything else:

  1 gold ore voxel      = 10,000 coins
1 gold_ore item       = one voxel's yield
2 gold_ore + 1 coal  -> 1 gold_bar          (mirrors the existing IRON_INGOT recipe)
1 gold_bar            = 20,000 gold_coin    (one way only)
  

30,000 × 10,000 = 300,000,000. The chain is a recipe rather than a special case precisely so that it is inspectable by the same boot checks as every other recipe.

What that gives at each world size link

Small worlds are over-provisioned, and deliberately so. Genesis gets 24 t of lode gold against an abundance that asks for 8 t, because the three guaranteed deposits are each floored at 8 t. Any function that derives a world size from a player count has to model that floor rather than assume gold scales linearly with area — below roughly 375 km it does not.

Genesis comes out at 298 rather than a round 300, and that two-player gap is the honest width of the model: placer gold is counted as a share of the lode total rather than traced down the rivers it actually follows. 300 is the design target, 298 is what the estimate returns, and the difference sits far inside the error on the placer term.

At 300 players holding around 100,000 coins each, players circulate about 30 M, a tenth of the supply, with some 120 M still unmined. That headroom is the room the world has to grow into.

Sizing the next world from a player count link

The relation runs backwards too, which is how a later world gets its dimensions: a target population fixes a coin supply, that fixes a tonnage, and that fixes an area. WorldSizing searches the forward model rather than inverting it, because the guaranteed deposits make the gold total flat across every world below about 375 km - three deposits that may not be drawn empty already hold more than the abundance asks for at that size. An algebraic inverse would answer confidently in exactly the range where the true answer is “gold is not what decides this”.

Genesis is smaller than the 222 km its own target suggests, and is not wrong to be: below the crossover the floor hands a small world the gold a much larger one would get, so 128 km already carries very nearly the 300 it was sized for. Past the crossover the suggestion is the binding one.

The three-tier purse link

Gold is not a single global number. It sits in three places, and it only ever moves between them.

graph TD
    Reserve["World reserve<br/>(un-minted gold + everything recovered)"]
    Settlement["Settlement treasury<br/>(SettlementLedger.treasury)"]
    Float["NPC float<br/>(one merchant's working purse)"]
    Player["Player wallets<br/>(gold_coin items)"]

    Reserve -->|"daily, by population key"| Settlement
    Settlement -->|"hourly"| Float
    Float -->|"hourly"| Settlement
    Float -->|"sells goods"| Player
    Player -->|"buys goods"| Float
    Player -->|"fees, upkeep, repairs"| Reserve
    Reserve -->|"mining and minting"| Player

• The world reserve holds gold that has not been minted yet, plus everything the world has taken back. It is the counterparty that makes the loop closed rather than notional.
• Settlement treasuries already exist and already work. What changes is where their income comes from: today they revert toward a derived target out of nowhere, which is the one place in the running code where gold is created.
• The NPC float is what one merchant carries. It is refilled from the town’s treasury and returned to it, so a shopkeeper can genuinely run out of coin on a busy afternoon while the town behind them is still solvent.

Big settlements get more. The daily distribution is weighted by population rather than split evenly, because a city of twenty thousand does more trade in a day than a hamlet of forty does in a season.

Nothing is burned link

This is a change from the earlier design, and it is the point of the whole structure. Auction listing fees, the commission posting tax, repair costs, the gold on a merchant who is killed — all of it returns to the world reserve. Nothing is deleted from the world.

The effect is the same as burning in the short run: the coin leaves circulation and stops pushing prices up. The difference is that it comes back, slowly, through the settlements, which means a busy economy is also a liquid one. A burn would make the supply monotonically decreasing, and a world whose money supply only ever shrinks has a very specific ending.

Minting yield is a recovery rate, not a bonus link

The Minting skill grants what its table calls a yield of +10% to +50%. Read as a multiplier on the coins a bar produces, that skill would create gold out of nothing and break the invariant this page exists to state.

So it is not a multiplier. A bar contains a fixed amount of gold, and minting recovers a share of it: an unskilled master wastes most of the bar, and Minting Lv. 5 wastes almost none. The waste goes back to the world reserve rather than vanishing. The skill therefore makes a player better at extracting value from gold they already hold, which is what the skill is for, without letting them print.

Invariants link

These continue the numbering used by the settlement simulation.

• I23 — Conservation. The world reserve, every settlement treasury, every NPC float and every gold_coin item held by a player sum to a constant, and that constant is the configured supply. A drift is a bug, not a balance issue.
• I24 — Minting is one way. No path converts a gold_coin back into a gold_bar. A coin is money for good.
• I25 — Treasuries are capped. No settlement treasury exceeds its ceiling; the overflow returns to the reserve rather than accumulating where it landed.
• I26 — A float is bounded. Killing a merchant yields pocket change, never a payday, because a float is a working purse and not a bank.

Two places the arithmetic can quietly leak link

Both are in code that already runs, and both are easy to get wrong while changing something else.

• Long catch-ups discard time. A settlement left alone for a year is simulated as thirty days, deliberately. With a reserve as counterparty, the discarded interval must not silently create or destroy gold — the reserve has to be reconciled against what was actually simulated, not against what was skipped.
• Ledger rows delete themselves. A settlement’s row disappears once it has decayed back to its reference, which is what keeps the table small. That stays correct only if the same reversion that erases the row has already made the reserve whole.

The end of the gold link

A finite supply mined out of finite deposits has a terminal state. Once the last gold voxel in the world is broken, no new coin can ever enter, and the economy becomes purely redistributive. At Genesis scale that is thirty thousand voxels, and three hundred players will get there.

This is an open question and the honest thing to do is say so. The candidates, none yet chosen:

• Recycling. Gold recovered from destroyed items returns to the reserve, so the supply is finite but circulates indefinitely. Closest to the rest of the design, since nothing is burned already.
• Slow replenishment. Deposits refill far below the rate they are worked. Preserves scarcity, weakens the fiction.
• New ground. The world grows, and new territory arrives with its own deposits.
• Accept it. Exhaustion is an intended late-game condition, and a world where gold cannot be made any more is a world where gold is genuinely precious.

The choice has to be made before mining ships, because it is the one decision that becomes permanent in the data.

Prior art link

The three-tier purse is taken almost directly from Wurm Online, which is the one MMO to have shipped a genuinely closed currency at scale. Gold flows from players into kingdom coffers through purchases and settlement upkeep, and the coffers periodically distribute money to NPC traders, which is the only route back to players. A Wurm trader has a finite purse that refills over time, and there is a hard cap on how much a single player can earn from one per hour. Both mechanisms are here: the float is the purse, and the payout cap is what stops a closed loop being drained by whoever grinds hardest.

A fuller comparison against Ultima Online, RuneScape, EVE, Albion and Star Wars Galaxies — and the failure modes each of them found — lives on the Economy Simulation page.