Tokenomics v2

Owner's request (11/03/2026): tokenomics that is sustainable in the long run and, ideally, growing; the documentation complete before the game starts (exact words in 00). This document is the v2 design and the reason for each rule. 14 says what applies today, in short form; the two must not contradict each other. The numbers come from modelos/tokenomics_v2.py (output in modelos/tokenomics_v2-saida.md, tests in modelos/test_tokenomics_v2.py).

How to read "growing". What grows is real demand for the token, driven by more players, more land, more creators and more chapters. Supply can only shrink. Nothing here promises, projects or announces a price. The price that shows up in the model is a stress test: it is there to see what breaks when it falls, not to say where it will be.

Two owner decisions, on the same day (11/03/2026), shape §G and §I. (1) Nobody on the team injects money: the team contributes only work, and protocol-owned liquidity opens with token only, with a floor and no dollars from the studio. (2) The studio's own catalog is card and PIX only; token pays for creator items, deeds, deposits, the minting fee and the player market. The exact words and the reason are in 00.

Third owner decision, on 11/03/2026 ("pode aplicar"): the nine proposals P1 to P9 from 39 are adopted. They go into §D (season held, capped per account and personal), §E (economic report, economy owner, keys and bridge), §I (newcomer guard as an alarm) and 15 (pasture with a time limit, no early land sale). No v2 rule parameter changed. The record is in 00.

The standing rules still apply: NFTs are only cosmetics, deeds, decoration and titles, never power; gold never turns into token; no sale, airdrop or token with real value before the legal opinion; no staking with interest, no APY, no vote over the treasury; what the player earns is a small, variable extra; the studio does not hold player assets; token only for verified 18+ players.

Summary

1. Diagnosis of v1

1. The burn is undone. In v1, reserve and treasury could release "up to the month's burn". Used to the limit, the net change is zero. In the model, base scenario, the v1 reserve and treasury release 595.8M over 120 months against 893.1M burned: 67% of the burn goes back into circulation. Supply would only truly shrink after the reserve ran out.
2. Velocity. Every use of the token was "buy and spend right away": cosmetics, tax, fees. Nobody needed to hold token. A currency nobody holds supports little value (MV = PQ), no matter how many players there are. In v1, deposit over circulating supply is 0%.
3. Structural sell pressure. 37 §2 shows the treasury selling as the biggest outflow: US$ 68.5k of the US$ 123.3k that leaves each month with 200 thousand active players. The season fund was assumed to be sold in full every month.
4. Legal form. "20% of revenue buys token to distribute" reads as profit sharing: the player receives because of the studio's results. Buying must be tied to consumption inside the game, not to profit.
5. Reflexive price, with no explicit choice. The v1 documents charge in token (15: tax "paid in token"; 33 §5: plot declared at 100,000 tokens), while the v1 model treated flows in fixed dollars. Pricing in token makes the sink cheaper when the price falls (pro-cyclical). With the price at a fifth, the token-priced sink collects a fifth of the money and burns the same tokens; the dollar-priced sink collects the same and burns 5 times more tokens (§H). In the current model's exodus the difference in the price bottom is small (0.50 with token sinks, 0.51 with dollar sinks); the argument that remains is the sink's own.
6. The studio depends on conversion. 37 §3: with 2% paying US$ 8, each active player costs more in server than they pay. The 5% × US$ 15 assumption is optimistic.

2. The v2 design

Each mechanism covers: why, numbers, what the model shows and the v1 problem it fixes. Numbers marked as proposals are on the calibration list (§7).

A. Supply only shrinks

Why. If everything that is burned can come back, the burn is bookkeeping, not reduction. The rule must ensure that, whenever the game has activity, more goes out than comes in.

Numbers.

• Maximum supply: 1,000,000,000, minted at once, with no function to mint more. Burn is permanent; nothing is issued to "match" it.
• Reserve + treasury together release at most 50% of the previous quarter's burn. The season (D) uses at most half of that (25%); the rest covers range liquidity (G) and creator payment shortfalls (E).
• Only exception: in the first year, protocol-owned liquidity (only its own tokens, never the reserve's) can enter the pool in a range without this cap (G).

Fixed-schedule groups: 27% (v1: 30%).

What the model shows. In all scenarios and in the 500 Monte Carlo runs, supply never rises. Over 120 months, the base scenario burns 309.7M (31.0%); the worst quarter releases 43% of the previous burn (cap 50%). Across the named scenarios the burn ranges from 29.3% (stagnation) to 31.2% (growth). In the Monte Carlo, the net change in supply falls between −32.6% and −17.2% (P10 to P90), with a median of −29.6%. The burn varies little across the named scenarios because, with supply that only shrinks, the price is what adjusts: what is released ends up burned.

Fixes: diagnosis 1.

B. Deposit: demand that stays

Why. For a currency to hold value, someone has to hold it. The deposit locks token while a right exists in the game and returns the same nominal amount when the right ends. It earns nothing: it is not staking, it pays no interest, it does not count toward rewards. The more land, clans and creators, the more token locked. This is the main demand mechanism that grows with the game.

Numbers (proposal).

• Form: hybrid. The value is in dollars, converted to token at the TWAP on the day it is locked, but the number of tokens cannot go more than 2× above or below the annual reference (average of the previous 12 months). It returns the same tokens that were locked, never more. Why: in pure dollars, a price manipulated upward shrinks the deposit to almost nothing and a price on the floor makes it prohibitive; in fixed tokens, it turns to dust when the price falls (§H).
• Exit queue: whoever gives up the right gets the deposit back in 30 days, and total refunds across the game are limited to 20% of everything in deposit per month; anything above that waits in the queue, in order.
• Cooldown: the same seat or plot cannot be taken back by the same account or clan within 90 days.
• The deposit only goes back to the wallet that posted it. It never serves as loan collateral and never counts toward the season.
• With no token (gate I not met), there is no deposit: land and seats follow the gold rules in 15 and 19.

What the model shows.

• The deposit locks 1.4% to 3.0% of supply (months 60 and 120, all scenarios), but that is 19% to 42% of what circulates (base: 26% at month 60 and 36% at month 120). In the Monte Carlo, deposit over circulating supply falls between 4% and 55% (P10 to P90), with a median of 26%.
• It grows with the game: in the growth scenario (a chapter every 6 to 9 months), the deposit goes from US$ 0.2M at month 6 to US$ 1.9M at month 36. Since it is in dollars, when the model's price rises it locks fewer tokens: the deposit locks dollars of demand, not a fixed number of tokens.
• Land locks little (there are few plots per chapter, on purpose). Creators and clans lock the most.

Fixes: diagnosis 2.

C. Creator economy

Why. Token buying has to come from someone buying something inside the game. A player-made cosmetic sold in the studio's shop is exactly that: the buyer pays for another player's work, and the creator is paid in token. It works like Roblox's creator payouts, but in reverse: the studio buys the currency to pay, instead of buying it back from those who want to leave.

Numbers (proposal).

• In the creator-item shop, the buyer pays by card, PIX or token, at the same dollar price. The studio's own catalog is card and PIX only (I).
• Creator share: 30% of the net price, always in token. If the buyer paid in token, the creator share comes out of that token. If they paid by card or PIX, the studio buys the token on the market at that moment, as a cost of that sale.
• The payment is held for 60 days in a public contract (card chargebacks, PIX MED refunds) and only then goes to the creator's wallet. The studio does not hold the token.
• A purchase by the creator themselves (same KYC) or by a linked account (same clan, device, payment method or repeated pattern) pays zero.
• Above US$ 500 per month, a creator's payment goes through manual review before release.
• Creators' share of cosmetics sales: a scenario, not a rule. In the model, 20% at launch, +4 points per chapter, up to 40%.
• Resale of a creator item on the token market: 2.5% royalty to the creator. Resale of a studio item: 2.5%, burned (the studio does not use token to pay bills). Outside the model; it is small.

What the model shows. In the base scenario, the studio's buying to pay creators goes from ~US$ 6k per month in the first year to ~US$ 12k at the end. It is smaller than the sinks, but it is buying tied to consumption, not to profit, and it grows with the number of creators. In the model, 50% of creator items are paid in token. Creators sell 70% of what they receive (assumption); even so, it is one of the smallest sources of selling.

Fixes: diagnosis 4. Replaces the 20%-of-revenue buyback as a source of buying.

D. Season: small and capped

Why. The season is the "extra" for those who play. If it is too large, it becomes the main sell pressure and the reason bot farms exist.

Numbers.

• Paid only from the reserve, at most 25% of the previous season's burn (season = quarter).
• Weighted by verified play, as in 14 §8: distinct players traded with and variety of activity; accounts from the same clan, IP or repeated pattern count little or nothing; one KYC, one eligible account.
• Automatic pause in the month the price falls more than 30% in 30 days. What is not paid stays in the reserve and does not become a debt.
• No distribution paid with buybacks.
• Held for 30 days (P4, adopted on 11/03/2026). What the season pays an account stays 30 days in a public contract before going to the wallet. An account banned in that window loses the held amount, which returns to the reserve (it is not redistributed to other accounts). The 25% cap counts the amount granted, not the amount released. Banning after the 30 days does not recover what already left; that is why the report in §E publishes the season paid to accounts banned later.
• Per-account cap (P5, adopted on 11/03/2026). No account receives, in one season, more than 10 times the median of the paid accounts in that season. The 10 is a proposal, to calibrate in alpha with the real distribution. What goes over the cap stays in the reserve.
• Personal season (P6, adopted on 11/03/2026). Eligibility is tied to the person's KYC and cannot be delegated or lent. A lent account (the manager "scholarship" of Axie, Thetan and Pegaxy) loses eligibility, and this goes in the terms of use. A banned account also loses the held amount, as in P4.

What the model shows.

• The average per active account sits at ~US$ 0.12 to 0.19 per month in the base and growth scenarios; with 30% bots, US$ 0.08 to 0.13. It is a small extra, as it should be.
• Lowering it from 25% to 10% barely changes the exodus (worst 30-day fall of −20.0% against −21.7%; bottom of 0.45 against 0.51). The season is not what brings the price down; the 25% cap stays.
• The pause does little in the scenarios (on for 0 to 1 month out of 120). It is insurance against the extreme case, not the main stabilizer.
• P4 and P5 are not simulated per account: the model is aggregate and has no per-account season distribution, which we will not invent (output §13). P4 appears only in the aggregate: if half of the bots' season is paid to accounts banned within 30 days (assumption, no data), the bots keep 25.9% of the season in the worst case, against 51.7%, and about US$ 1.0 million over 120 months returns to the reserve (output §15.3). The humans receive the same. P6 is a terms-of-use rule, outside the model.

Fixes: diagnosis 3 (the fund part) and 4; and, from 39, the account farm (Alien Worlds), the giant account (Gods Unchained) and the side scholarship (Thetan, Pegaxy, YGG).

E. The treasury never sells token

Why. A treasury that sells to pay bills is permanent sell pressure and makes the token price part of the studio's budget.

Rule.

• The studio pays bills in cash (card and PIX from cosmetics).
• Treasury token is only for: protocol-owned liquidity (G), creator payment shortfalls (when the oracle circuit breaker is active, the treasury advances the payment and the studio buys in the following month to replace it) and burn.
• The treasury receives 1 point of the 5% token market fee.
• Public monthly report, in two parts, on the same day:
  • Treasury: balance and addresses, inflows, outflows by use, liquidity put into the pool, burn for the month and the quarter, release against the 50% cap, season paid or paused, creator payments held and released.
  • Economy (P1, adopted on 11/03/2026): gold and token sources and sinks for the month; on-chain circulating supply against the published schedule, with any difference explained; newcomer share of the dollars that buy token (the guard in §I); season paid to accounts banned later and what was clawed back (§D); months with sales stuck at the floor and tokens with no buyer (§G); verified human actives, not wallets. The first report comes out before the beta (in alpha, with the test currency, as a rehearsal).
• Economy owner (P2, adopted on 11/03/2026). A named person, the project owner at first, signs the report and is the one who triggers the levers already published: the gold sink controller (x0.8 to x2.0, 33 §10), the 5% and 10% bot triggers and the reviews of the cash floor and the newcomer guard (§I), and the season pause and cap (§D). Creates no new lever and changes no rule alone: a rule change goes through 00. The name goes in the report. Applies from the beta; before that, it is the owner.
• Bridge and treasury keys (P7, adopted on 11/03/2026). Multiple signatures with independent signers: no single entity alone has enough keys to sign (on Ronin, 4 of the 5 stolen keys belonged to one company, 39 §1); a timelock on large treasury outflows, so there is time to see and stop them; a public alert on every treasury and key movement. Numbers in Phase 2 (how many signatures and from whom, timelock length, what counts as a "large outflow", per-withdrawal limits), together with the contract audit (34 §8).

What the model shows. Treasury selling: zero, in every run (it is a test: test_tesouraria_nunca_vende). Cost: without selling token, the studio does not turn into cash what the player pays in token (§I). P1, P2 and P7 are operational rules and stay outside the model.

Fixes: diagnosis 3; and, from 39, the report that only looked at the treasury (EVE, MIR4, Pixels) and the concentrated keys (Ronin, Horizon, Play Bridge).

F. Longer unlock

• Team: 12-month cliff + 72 linear, counted from the public launch.
• Partners: by milestones, cap of 60 months.
• Testers: as in v1 (25% at launch + 12 months).

What the model shows. Fixed-schedule peak: 4.3M per month (v1: 5.9M), ending at month 84 (v1: 60). Even so, the largest monthly net sale in the base scenario comes from this schedule (month 25, US$ 59k, 38% of that month's selling). This is the token's risk window, as 37 already said.

G. Liquidity

Why. The team contributes only work (owner's decision, 00). So the protocol pool cannot depend on anyone on the team putting in money: it opens with token and only collects dollars when players buy.

Rules.

• Protocol-owned liquidity, opened with token only. At launch, the pool receives 5M tokens and US$ 0 from the studio. The tokens sit in a one-sided range that starts at the floor price (in the model, index 1.00; the floor's dollar value is a proposal to calibrate in beta and is published in the contract). The dollar side starts empty and only grows with player purchases. The LP position stays with the treasury.
• Floor. Below the floor there is no buyer in our pool. A sale that finds no buyer does not happen: the token stays with whoever wanted to sell. Only the money that buyers put into the pool can leave it. Result: in our pool there is no spiral below the floor, but there is also no exit below it.
• Honest caveat: the floor only exists in OUR pool. Another pool or an exchange can trade below it, and we do not control that. The floor is not a guaranteed price and no material may present it as one (§5).
• No paid liquidity incentive (no "liquidity mining").
• Range above the price: when the price goes above 1.5× the 12-month average, liquidity tokens enter the pool, and the dollars that come in stay in the pool, never going to bills. It counts toward the 50% cap (A), except in the first year.
• Oracle circuit breaker: the shop charges at the 30-minute TWAP (24 hours in the first year, below). If the spot price moves more than 10% away from the TWAP, or the TWAP moves more than 25% in 24 h, the shop pauses payment in token (card and PIX continue), deposits and tax use the last valid TWAP, and creator payments come from the treasury. Public rule and limits.

Launch-phase safeguards (first 12 months), because the pool is shallow. These are proposals; the model has no oracle and does not simulate these three.

1. Longer TWAP for the dollar-priced sinks: 24 h instead of 30 min in the first 12 months. After that, back to 30 min if calibration shows a deep pool.
2. Oracle circuit breaker, the one above, with the 10% and 25% limits.
3. Pause on movement: if the spot price moves more than 10% in 1 hour, the shop pauses payment in token until the price settles (card and PIX continue). X = 10% is a proposal, the same limit as the circuit breaker. With a 24 h TWAP, the 10% limit against the spot price trips more easily; calibrate in alpha.

What the model shows (output §10b). The pool with no team money collects real dollars only from player purchases: US$ 135k at month 3, US$ 218k at month 12 and US$ 326k at month 36 (base scenario). In months 1 to 12, with only the testers and the pool circulating, the price index reaches 2.53 at month 8 and closes the first year at 1.89.

The tokens column adds up sale attempts with no buyer: the same token that goes back to the wallet and is offered again counts again.

The cost, said plainly. In the exodus, 22 months have sales stuck at the floor: whoever wants to leave cannot leave through this pool. That is the trade-off of a pool with no team money: no spiral below the floor in our pool, but no exit there. The shock (an outside sale that would drop the price 80%) and the whale (10M tokens, 1% of supply) hit the floor: both end with a bottom of 0.45 of the month-29 level (−55%). With dollar sinks, the same activity burns 1.51× more tokens in the bottom month, and at month 36 the index is at 1.03× that of month 29. This risk must be written in the whitepaper (16). A shallow pool is a real launch risk: that is why the three safeguards exist.

H. Dollar prices for studio sinks

Why. With dollar prices, when the token falls, the same activity burns more tokens: the sink works against the fall. With token prices, it gets cheaper along with the price and speeds up the fall.

What the model shows. Same activity (base scenario, month 24), price ×0.2: with dollar prices the burn goes from 1.8M to 9.0M per month and the money coming in does not change (US$ 119k); with token prices the burn stays at 1.8M and the money coming in falls to a fifth (US$ 24k). A finding that changed with the pool that has no team money: in the exodus, the bottom is almost the same with dollar and token sinks (0.51 against 0.50), and in the hard exodus (−80% in 3 months) too (0.46 in both). In this model the exodus bottom does not depend on dollar pricing; what weighs more is the deposit (§B: hybrid 0.51 against 0.38 in fixed token; hard exodus, 0.46 against 0.36). Dollar pricing stays because the sink keeps working in favor of the price (it burns 5 times more tokens with the price at a fifth) and collects the same money, but it is no longer the strongest stabilizer in the design.

Fixes: diagnosis 5.

I. Token launch gate, studio catalog and the gold economy

• Gate (adopted from 37 §3): the token is only launched if beta measures gross cosmetics revenue ≥ US$ 0.40 per active player per month. Below that, the game ships without a token: cosmetics by card and PIX, and nothing changes in gameplay (gold, instanced land and professions do not depend on the token, 14 §1).
• Studio catalog in card and PIX only (decided on 11/03/2026). Token pays only for: creator items; deeds, deposits and the minting fee; and the player market. In the model, STUDIO_CATALOG_TOKEN = False. It was a recommendation in 00; it is now the owner's decision, for the same reason as the pool (the team contributes only work).
• Cash floor (proposal, now an alarm): if cosmetics revenue by card and PIX stays below US$ 0.20 per active player per month (server cost) for two months in a row, the owner reviews prices and costs. The earlier lever (taking token out of the catalog) has already been applied by decision, and the team injecting money is off the table.
• Gold: the sink controller chases the band (BAND_CTRL turned on in economia.py, lever from x0.8 to x2.0), as 37 §1 proposed.
• Bots, operational triggers: above 5% estimated bots in active accounts, bans get priority; above 10%, no new price sink goes in before a ban wave.
• Newcomer guard, as an alarm (P3, adopted on 11/03/2026). If newcomers (in their first three months of play) pass 20% of the dollars that buy token in a quarter, after month 12, the economy owner (§E) reviews and publishes the result in the report. The first 12 months are left out because at launch everyone is a newcomer. There is never a mandatory purchase at entry: nothing is paid to play or to enter the season (00 E1). It is an alarm, not an automatic lock. In the model, the same limit is a test: test_novatos_abaixo_da_guarda (output §15.1) is the implementation of the alarm in the model. The test uses 20% over months 13 to 120 and 25% in any single month; in base the number is 11.7% (worst month, 13.0%), and an entry kit of US$ 3 per newcomer blows through it (mutant). In production the alarm looks at the quarter.

What the model shows. What the player pays in token in the studio catalog is burned; it does not become cash. With 50% of catalog purchases in token (v1 assumption), the studio's cash at month 24 of the base scenario is US$ 32k with the token live against US$ 94k with the token at zero; over 120 months, US$ 4.90M against US$ 13.45M. With the catalog in card and PIX only (and 50% of creator items paid in token), cash rises to US$ 79k at month 24 and US$ 10.92M over 120 months. With poor conversion (2% × US$ 8), cash is negative (−US$ 6k at month 24 and −US$ 1.04M over 120 months with the token live; −US$ 3k and −US$ 0.50M with the token at zero), and the gate prevents the token launch.

Sensitivity, measured with the model (base scenario, 120 months):

Closing the studio catalog to card and PIX costs about 4% of the burn (323.1 → 309.7M) and more than doubles the studio's cash (US$ 4.9 → 10.9M). That extra cash is what keeps the team from ever needing to inject money. The pool still collects about US$ 218k of real dollars by month 12, all of it from player purchases. Almost all of the burn comes from deeds, tax, minting, the player market and creator items.

Fixes: diagnosis 6.

J. Communication

See "How to talk about it", below. The model's reference price (US$ 0.02) is only a unit converter and does not appear in any material.

3. What the model shows, on one page

Main limit: in the long run, the model's price drifts upward, because dollar demand does not react to price and no outsider sells beyond the 3% turnover. After the fixed schedule ends, almost nothing new circulates. This is an effect of the assumptions, not a forecast; that is why the model is only read in relative measures (§13 of the output).

4. Exploits and defenses

5. How to talk about it

Every piece of communication about the token carries the same risk warning as the whitepaper (16).

6. What we do not know

• Conversion and spend per payer. The model inherits 5% × US$ 15; the gate exists because this may not happen.
• How much is paid in token. With the studio catalog in card and PIX only, the doubt is the share of creator items paid in token (50% in the model). More token means more burn and less cash (§I).
• How much holders sell (turnover). The model assumes 3% of free circulating supply per month; in the Monte Carlo, 1% to 6%. It is the parameter that moves the price level the most.
• Launch liquidity. The pool opens with no dollars. The floor value, the dollars players actually put into the pool and how much selling goes without a buyer depend on behavior that only alpha and beta measure. With only testers and the pool circulating, the model's index reaches 2.53 at month 8.
• Exit for those who want to sell. Below the floor, in our pool, there is no buyer. How much that weighs on those who join is a product and communication question, not a model one.
• Outside markets. Another pool or exchange can trade below the floor. The model does not have them.
• Creators. How many, how much they sell, how much they hold. The 20% to 40% share of sales is a scenario.
• Real market depth. The model only has the protocol pool, and it has no oracle: the TWAP, the circuit breaker and the launch safeguards (§G) were not simulated.
• Per-account cap and holding period. P4 and P5 are not simulated per account, because the model is aggregate (output §13). The 10× is a guess until the alpha shows the real per-account season distribution, and the 30-day ban rate (50% in output §15.3) is an assumption with no data.
• Behavior in an exodus. The model assumes every refunded deposit is sold (worst case).
• Legal. Deposits, creator payments in token and burn need the legal opinion (34 §10). If the opinion blocks the token, the gate already covers it: the game ships without a token.

7. Calibrate in alpha and beta

• Gross cosmetics revenue per active player (US$ 0.40 gate) and card and PIX revenue (US$ 0.20 floor).
• Share of creator items paid in token (model: 50%).
• Creator share (proposal 30%), share of creator items in sales, 60-day hold, review above US$ 500.
• Deposit per right (US$ 150 / 80 / 400 / 200 / 500 / 15), 2× band, 30-day queue and 20% per month, 90-day cooldown.
• Season cap (25% of the burn) and pause trigger (30% in 30 days).
• Joint release cap (50% of the previous quarter's burn).
• Token market fee (5%: 4 burn, 1 treasury), 2.5% royalties, US$ 0.50 minting fee.
• Initial liquidity (5M tokens, floor, long TWAP in year 1), range trigger (1.5× the average), circuit breaker limits (10% and 25%) and pause on movement (10% in 1 hour).
• Launch floor in dollars; real dollars entering the pool (model: US$ 218k at month 12); tokens with no buyer and months with sales stuck at the floor.
• Allocation for partners (7%) and treasury (13%) against real contracts and uses; burn the treasury excess in the annual report if use stays low.
• Holder turnover, exit of newcomers and veterans, share of verified 18+ players.
• Bots: real share, 5% and 10% triggers.
• Newcomer guard: 20% limit per quarter after month 12 (P3) against the number measured in the beta.
• Season: 30-day hold (P4), per-account cap of 10× the median (P5) and personal eligibility in the terms of use (P6).
• Monthly economic report (P1): source of each number and the count of verified human actives; name of the economy owner (P2).
• Bridge and keys (P7), in Phase 2: how many signatures, from whom, timelock length, what counts as a large outflow, public alert.
• Pasture with a time limit: 60 days every 12 months per plot with a deed (P8, 15).
• Run modelos/tokenomics_v2.py again with the measured data.

8. Owner's pending decisions

They are in 00, with the recommended answer: token name, network, creator share and team allocation. Decided on 11/03/2026: protocol-owned liquidity opens with token only, with no team money (§G), and the studio catalog is card and PIX only (§I). Decided on 11/03/2026 ("pode aplicar"): the nine proposals in 39, P1 to P9 (§D, §E, §I and 15).

Model

modelos/tokenomics_v2.py, standard library only. It reuses tokenomics.py, economia.py and sustentabilidade.py without changing their assumptions. Monthly step, 120 months, population in cohorts (newcomers leave at 12% per month, veterans at 4%), dollar flows converted at the month's price.

Price: a constant product pool (x tokens × y dollars = k) with the protocol-owned liquidity, opened with token only (5M, US$ 0 from the studio) in a one-sided range that starts at the floor (index 1.00). A sale that finds no buyer does not happen and the token goes back to the wallet. Every buy and every sell in the month goes through the pool, in 4 steps per month; the price is y ÷ x. The model has no oracle: the TWAP, the circuit breaker and the launch safeguards (§G) are proposed rules, outside the simulation. The external term is the order balance (buys minus sells, in dollars) against pool depth, plus holder turnover. It is a stress test device, not a forecast.

Scenarios: base (200 thousand active players, +0.5% per month), growth by chapters, stagnation, exodus (−50% in 6 months), hard exodus (−80% in 3), −80% shock, whale with 1% of supply, 30% bots, poor conversion (2% × US$ 8) and a 500-run Monte Carlo. v1 runs in the same scenarios, minimally reimplemented.

Tests (python3 -m unittest modelos/test_tokenomics_v2.py): supply never rises; total never goes above 1 billion; deposit returns at nominal value, never more; season ≤ 25% and release ≤ 50% of the previous burn; treasury never sells; season pauses on a drop above 30%; creator's own purchase pays zero; the studio puts no money into the pool; the pool never pays out more than it received; the price never falls below the floor; the pool's real dollars are never negative. Each invariant was broken on purpose and the test failed (python3 modelos/mutantes_tokenomics_v2.py).

Other games' failures as stress tests (output §15, analysis in 39): dependence on newcomers (STEPN test), entry stopped from month 36, account farm and speculative land. Six more invariants: with entry stopped, supply never rises and the pool's real dollars are never negative; in base, newcomers stay below 20% of the dollars that buy token (test_novatos_abaixo_da_guarda, the implementation of the P3 alarm); outside pasture mode, no plot stays idle more than 4 months with the same owner; with pasture limited to 60 days every 12 months (P8), no plot stays idle more than 6 months (2 of pasture + 4 of inactivity) with the same owner; and with the clawback of banned bots' season (P4, aggregate only), the sum of all buckets stays at 1 billion. Each has its mutant. P4 and P5 are not simulated per account, because the model has no per-account season distribution (output §13).