Nothing behind the curtain.

Hold at both ends of the round.

The threshold is 1,000,000 SIA. A wallet must meet it at both the opening and closing snapshots. New holders may enter the public directory immediately and begin reward eligibility next round. Transferring tokens between wallets cannot carry opening eligibility to a different address.

Performance, adjusted for capital.

Scoring version sia-v1.0 combines capital-adjusted return (35%), realized return on supported cost basis (25%), drawdown resilience (25%) and profitable active-day consistency (15%). The result is multiplied by data confidence. Returns are normalized as clamp(5,000 + 2 × return in basis points, 0, 10,000); resilience is 10,000 minus maximum drawdown in basis points. Consistency is profitable days / active days. The final integer score is on a 0–10,000 scale, displayed out of 100. Minimum requirements: 14 days of history, 10 closed trades, five active days and 80% coverage. Ties resolve by data confidence, then closed trade count, then lexicographic address. SIA trading never contributes.

Transfers are not profits.

Supported trades use FIFO lots, integer quote micro-units and all supplied supported venue fees and gas. Routed swaps must be normalized to one economic trade per route before ingestion; duplicate route IDs are ignored. Deposits and airdrops create unknown-basis lots. Withdrawals consume lots without creating a gain or loss. Sales with missing cost basis are excluded, never assigned zero cost. Unrealized P&L is shown only with reliable marks. Capital-adjusted returns use Modified Dietz; drawdown requires external-flow-adjusted unitized equity.

Know the boundaries of the evidence.

No chain, venue or price provider is verified yet. The default analysis window is 30 days, not lifetime or cross-chain history. Coverage measures supported, complete priced activity within the stated window. Missing history, unreliable prices or too-small samples prevent a ranked reward entry. Provider adapters must document finality, supported assets, liquidity thresholds and price freshness before launch.

Funded receipts. Never assumed yield.

The intended model is a 3% trading fee dedicated to holder rewards. The launchpad and exact fee structure are not selected or verified. This is not 3% of creator fees or 3% yield for holders. Platform deductions must be reconciled against actual creator receipts. Only confirmed net receipts allocated to rewards enter the funded budget. Operating expenses and gas are separately funded. No guaranteed returns.

One closing snapshot. One allocation.

Beta rounds run daily, with an intended 00:00 UTC close. No round is scheduled before setup. Allocation: 30%, 20%, 15%, 10%, 5%, then 4% each for sixth through tenth. Empty-place allocations and integer rounding remainder carry forward. Rewards use frozen rankings and both eligibility snapshots. A controlled test round is required before real funds can be released. Payouts move through queued, prepared, submitted, confirmed or failed states; ambiguous responses are reconciled, never blindly resent.

A flag is a reason to review.

Rapid equal-quantity buy/sell reversals in under 60 seconds trigger a review signal. Missing cost basis and unreliable prices reduce confidence or prevent ranking. Provider evidence of shared funding, synchronized positions or common counterparties opens wallet-splitting, self-trading or wash-trading review cases. These are not accusations. Suspected wallets await review and are not automatically labeled fraudulent. Unsupported assets, unresolved review and insufficient history do not receive a reward allocation.

Public addresses. Private ownership.

Anyone can search a public address. Ownership verification is required only for personal preferences. No seed phrase or unnecessary transaction approval is requested. Wallets are not linked to real-world identities. AI receives bounded evidence to explain results, cannot alter scores or reward amounts and has no private-key access.

The agency, in stages.