Token rewards. Everything is paid in Token, never in new WT. That needs a cash budget, not a token allocation. Supply is fixed and can only fall — redeeming against the cash reserve retires the WT you redeem.

A chip. You do not bet with WT. Bets settle in Token and the major crypto currencies. Keeping the wagering currency separate from the reward currency is what stops revenue from being denominated in our own token — the loop that broke both losers above. The arithmetic is in the appendix.

1. Minted on house edge, on real-money bets only. Zero-edge games mint nothing. BT bets mint nothing.
2. Split pro rata, every week. The pool buys WT; everyone cashing in that week shares it in proportion. No account takes more than 5% of a week's pool, and anything over the cap is returned rather than forfeited — a capped week delays a claim, it never destroys one.
3. Every BT expires twelve months after it is minted — dated from each mint, so the amount in circulation can never exceed a year of minting.
4. BT tables stay. Losses burn, wins mint, a single win capped at 1% of everything outstanding. A BT win mints BT, not dollars, so the casino's exposure is exactly zero.
5. What BT buys sells back for BT, so the weekly pool stays the only route from BT to WT.
6. BT that is not earned on house edge is budgeted. Prizes and grants are capped together at a tenth of the previous week's earned BT.
7. Boosts decide who gets the pool, not how big it is. They divide a fixed pool rather than multiply a bill, so they cost the casino nothing.

Two guards travel with the mint: the edge used is the one actually charged on that round, and games where another player sets the return are excluded — otherwise the mint would follow a number your opponent chooses.

This is the white paper’s own monthly model, line for line: each plan’s player ramp, the six allocation schedules with the airdrop unlocked by wagering, the launch subsidy, the cash reserve, and bought-back WT held out of the float. Hold staking at 55% and turn buyer interest off, and the payout value lands within 3% of the paper’s fair value in every month, and within 1% at month 24. The gap is the buyback: it buys at the market price here, not at fair value, so it holds a slightly different amount of WT out of the float. The revenue dial scales the chosen plan’s ramp.

The market price is cleared month by month. Buying is the payouts stakers put back into WT, outside interest — a quarter of revenue at launch, growing at your rate — and the buyback’s WT that players keep. Selling is the share of each new release its holders sell, 10–60% depending on who holds it, as the paper assumes, plus turnover of the unstaked float, which rises as the price runs above the payout value. That turnover is set once: on the base plan at month 24, with the paper’s own flows, the market clears exactly at fair value. It is anchored to the paper, not tuned to a result.

The market sees the release schedule coming. The unlock calendar is public, so traders do not wait for a release to land before pricing it: buying and release selling enter each month’s price as a weighted average of the months ahead, with a month six months out counting half. Priced month by month instead, the private round’s first release would crash the price in month 4 and its last would lift it 87% in month 13 — jumps a published schedule rules out.

One limit. Month 0 is the listing month. Half the public round selling into one month of launch buying is not a flow a monthly model can price, so the market line starts at month 1. Use the model to see what the price is sensitive to, not as a quote.

The direction worth noticing: more staking lowers the payout value, because the same dollars are shared more widely, and raises the market price, because it takes WT off the market. And raising the holder share has to come out of operations — the other four claims are fixed.