Warbandry · the economy, with the levers in your hands
What a roll costs, and what that decides
A roll is credited per 1,000 tokens bought, and the tokens come back when you sell. So a roll costs the swap fee, not the tokens — and the fee, the rarity split and the coin's value decide everything downstream: what a card is worth, what a season costs, what launch day mints. Move the levers; every figure recomputes.
From the analyses of 2026-08-25 and 2026-08-26 · every fee is a parameter, not a decision
1 · The roll
The buyer pays the swap fee on the way in, the swap fee on the way out, and two gas bills shared by however many rolls one swap carries. The 1,000 tokens themselves are recovered. This setup — value, buy fee, sell fee, batch — feeds every section below.
Scenarios
30 k$ at the left, 100 M$ at the right
the platform's dynamic band is 0.3 / 0.5 / 1.2 %
1 · 10 · 50 · 500 — buying n × 1,000 tokens in one swap is n rolls
cents; an L2 figure
Face value of a roll
Sunk cost of a roll
Share of face lost
Rolls for 100 $ of fees
Any mythic, on average
Recommendation
Keep the roll priced by the fee, and say so in the copy. The tokens come back and no gate on a balance survives one transaction, so the fee is the only price a roll can have — and the fee is a lever we hold. The roll threshold stays at 1,000 tokens; it only sets the grain of the gacha.
What the scenarios show
Launch, calm, one at a time — the fee is a fifth of a cent and the gas is four cents: gas is the whole gacha. A whale batching fifty pays forty times less per roll.
1 M$, fixed 1 % — a roll worth 1 $ costs 2 cents. A mythic is a dollar of fees.
100 M$, asymmetric — the keeper pays 0.5 % of a 100 $ face; the discarder pays 3.5 %. A mythic costs 175 $ to find and 100 $ to hold.
10 M$, settable high — 1 $ a roll, 50 $ a mythic: the gacha finally costs what "fifty rolls" sounds like, and every trade pays the same 5 %.
2 · The rarity split
Six tiers, a share each; three factions at parity. A named creature of a tier has that tier's share divided by how many creatures the tier holds (9 / 9 / 12 / 6 / 6 / 3). Traits ride on the roll: 30 % positive, the trait's tier on the same split, about twelve positive traits to choose from. The split is a multiplier on every cost below it.
Splits
The six shares are normalised to 100 % as you move them; the labels show the normalised value.
Target
Odds per roll
Rolls, on average
Rolls for 90 %
Fees, at the setup above
Expected rolls per target, log scale
Recommendation
Launch on the epoch-0 split. It puts a mythic at fifty rolls — a purchase, not a wall — and keeps the creature-and-trait pairs in the thousands, which is where the collection's real rarity lives. A steeper split doubles every price with nothing gained in play; a flatter one empties the top. The split is epoch data, so it can move later for future rolls only.
What the splits show
Epoch 0 — a mythic every 50 rolls, a named mythic every 150, a named mythic with one chosen positive trait every 6,000. The last is a windfall, not a hunt.
Flatter — halves every search: mythics become common enough that the top of the collection stops being a top.
Steeper and thin top — double every search and every card premium with it. A named legendary at 150 to 200 rolls is a purchase; at 100 M$ and 1 % it is 300 to 400 $ of fees.
3 · Cards and farms
A card's floor is the lock it carries. Its fair price is the floor plus what the creature cost to find: rolls × sunk roll cost, plus the carding fee. Because the sunk cost is a share of the face and the face is the lock, the premium collapses to rolls × (buy fee + sell fee) — a number with no price in it. Under about +10 % nobody hunts and nobody farms; over +100 % a farm lives off the spread and a player is better off buying the card than rolling for it.
Fee setups
tokens frozen per creature; the vault lever, 10 to 1,000
burned in coins, share of the lock; cap 10 %
Mythic card · floor
Mythic card · fair price
Mythic card · premium
Named legendary · fair price
Named legendary · premium
Mythic card premium against the round-trip fee (buy + sell), at this splitpremium · the dot is your setup
Recommendation
Buy low, sell high: about 0.5 % / 3 %, each under a 5 % cap written in code. The premium lands between +100 % and +200 %, a real card market, and it is the only shape where the player who keeps what they roll pays a seventh of what the discarder pays. The carding fee then taxes the farms at the exit, never the player who recruits.
What the setups show
Calm 0.3 / 0.3 — a mythic card is worth a third over its lock. Thin margins, no farm, and no reason to hunt anything either: rarity is worth nothing on the market.
Fixed 1 / 1 — the card doubles its lock. A market exists; a farm covers its costs.
Asymmetric 0.5 / 3 — the premium is +177 %, a full market, and the player who keeps what they roll pays a seventh of what the discarder pays. This is the setup the second analysis recommends.
Settable high 5 / 5 — six times the lock. Cards are expensive, farms are rich, and every ordinary trade pays 5 % — the trading that a low fee invites is gone.
4 · The arena
Both players stake; the winner takes both stakes less the rake. A player at fifty per cent loses exactly rake × stake a match, whatever they do. Stakes are set by hand in tokens and follow the price of one creature; the rake is set per bracket. The rule that keeps a season affordable as the coin grows: lower the lock so a creature stays near a dollar.
Players
per bracket; cap 15 %
log scale, 10 k to 10 M — for the burn
Lock in force
Loss per match
Matches until one creature is gone
The season, in money
Burned a year, share of supply
The season's cost across values, at this stake and rakelock at 1,000 lock follows the value
Recommendation
A free bracket, the rake set per bracket, and stakes that follow the lock rule. A player at fifty per cent loses rake × stake a match whatever they do, so the free bracket is how the ladder lives while it is thin; 2 % on small stakes and up to 15 % on large ones lets the whales fund the ranking; and the lock rule keeps a season at the same dollars at 1 M$ and at 100 M$.
What the players show
Casual — thirty matches at 2 % on one creature: 0.6 of a creature a season. With the lock rule that is sixty cents at any value.
Regular — a hundred matches at 5 %: five creatures a season. 5 $ with the rule; 500 $ at 100 M$ without it.
Grinder — three hundred matches on five creatures at 5 %: seventy-five creatures a season. The ranking half of the rake is what they are paying into, and what they are playing for.
Whale — twenty creatures at 15 %: three creatures lost every match. At a million matches a year that bracket alone burns fifteen per cent of the supply.
5 · Launch day
The pool opens with ETH against half the supply. A buy takes tokens out along the curve; selling straight back returns about what was paid, less the fees — so the roll count is the whole story. Rolls are cheap on day one because the coin is cheap, not because the pool is thin: at 30 k$ a dollar is 33 rolls whatever the pool holds. A sniper tax makes them a little less free; an opening value makes them cost something.
Scenarios
15 k$ to 1 M$, log — against half the supply; the opening value is twice this
100 $ to 100 k$, log
a burst triggers the 1.2 % band
Opening value
Rolls minted
Price moved by
Sunk, fees and tax
Per roll
Mythics expected
Recommendation
Accept the rush, set the sniper window for fairness, and choose the opening value on purpose. No tax makes a three-cent roll expensive; only the opening value does. Whether the first day mints two hundred thousand rolls or twenty thousand is a product choice — a collection born in an hour or a game that is discovered — and it is made by the value the pool opens at, not by the pool's depth.
What the scenarios show
15 k$ pool, 1,000 $ — 31,000 rolls for 24 $ of fees; 625 mythics expected for one wallet, price up 14 %.
15 k$ pool, 10,000 $, 30 % tax — 200,000 rolls, the price nearly triples, and the tax takes the roll from 0.001 $ to 0.016 $: still nothing against a face of 3 cents.
50 k$ pool, 10,000 $ — the deeper pool lets the whale in with less impact and more rolls per dollar of impact, not fewer.
500 k$ pool — an opening at 1 M$: the same 10,000 $ is 10,000 rolls, and each one costs 2 cents. The opening value is the lever; the tax is a fairness tool.
6 · Revenue and outside liquidity
The creator receives 75 % of every swap fee. Revenue is value × turnover × average fee × 0.75, where turnover is how many times the supply is bought in a year — one turn is a million rolls. Rolls are credited only on our pool, so the roll flow cannot leave; pure trading can.
Scenarios
the rest is pure trading
Rolls a year
Creator revenue a year
Lost to the other pool
Recommendation
One canonical pool, no staking vault, and the copy says where to buy to roll. Rolls are captive to the pool that credits them, so outside liquidity costs at most a quarter of the revenue; a staking vault would pay holders for holding, which the operator has ruled out in favour of paying the ranking. Watch turnover, not price.
What the scenarios show
Quiet — below 1 M$ there is no business, only a game; at 10 M$ and 1 %, one turn is 75 k$.
Active, half the trading leaves — revenue loses a quarter, not a half. The roll flow is captive to the pool that credits it; that is the whole argument for one canonical pool, and for the copy saying where to buy to roll.
Frenzy — ten turns is ten million rolls a year; revenue is linear in the fee, and so is the card premium, but the trading that makes ten turns is not — a 5 % fee kills it.
7 · Relics, and the frozen supply
A relic is an object of its own, rolled like a creature with its own rarity ladder, frozen with its own lock. Two things follow: one roll in so many is a relic instead of a creature, and every warband carries one more lock. The slot is cheap only once the lock lever has been pulled down.
Scenarios
log, 100 to 100,000
Rolls to any relic
Rolls to a legendary relic
Rolls to a mythic relic
Frozen, 16 creatures
Frozen, with the relic slot
Recommendation
One roll in twenty, and pull the lock down before the slot opens. At a 1,000-token lock the relic slot alone takes a full point of the supply off the market for every ten thousand players; at 100 it takes a tenth. One in twenty puts a mythic relic at a thousand rolls — rarer than any creature, still findable.
Method
Supply 10⁹; a roll per 1,000 tokens bought, gross, per wallet; face value = value ÷ 10⁶. Tokens are not consumed by a roll. A round trip on a constant-product pool loses only the fees when nobody trades between the legs. Every draw is independent: expected rolls are 1 ÷ odds, and the 90 % figure is the count that leaves a one-in-ten chance of still coming up empty.
Roster 9 / 9 / 12 / 6 / 6 / 3 per tier. Traits: 55 % none, 30 % positive, 15 % negative; a positive trait's tier on the same split; about twelve positive traits to choose from. Melt 2 %. Half the rake burned. Creator at 75 % of every swap fee. Player curves are illustrative and nothing here is a forecast.
The written analyses, with every table at every value, are in the repository: 2026-08-25-economic-analysis.md and 2026-08-26-economic-analysis-2.md, and the model that produced the second.