SpreadBoard separates market discovery from execution evidence. Unknown inputs stay unknown instead of being replaced by optimistic defaults.
The standard $50 probe estimates each leg against directional depth. Grid Discovery separately checks native minimums, contract multipliers and the same base quantity on both legs at $10, $25 and $50.
(sell VWAP ÷ buy VWAP − 1) × 100A tiny best quote cannot make an entire position executable. When target depth is unavailable, the matched-size value stays unresolved; any current top-book reference remains separately labelled and the calculator does not invent a fill.
The multi-horizon 24h outlook weights the current-rate projection 35%, settled 1d 30%, dailyized 7d 25%, and dailyized 30d 10%, then renormalizes across only the horizons that exist. A current/history sign conflict cuts the blended magnitude by 25% and is shown explicitly; history can keep a cooled leader on the radar but cannot masquerade as a live rate.
Funding stability, persistence and rate form the funding thesis. Entry size, observed convergence, half-life and basis risk form the spread thesis. Multi-horizon expected funding adjusts the spread score; evidence-backed convergence adjusts the funding score. Positive means the shown hedge receives.
The Watchlist estimates a collateral reserve for each futures leg from hourly prices: rolling maximum adverse 24-hour excursions, full-history and recent realized volatility, 99% parametric shocks, basis widening, and leg correlation. The larger supported shock is used rather than an average.
The personalized margin planner combines that public stress with the member's exact maintenance tier, leverage, isolated/cross equity, other positions, fees, borrow, gas and slippage. It returns stress headroom and refuses missing required inputs; it does not call that an exchange liquidation price.
Matching tickers are not enough. Market base/quote metadata and shared contracts or mints are used where available. Implausible ratios, leveraged products, and ambiguous DEX identities are held out.
A route without enough identity evidence is research context, never silently upgraded into an executable claim.
Spot transfer routes require withdrawal from the buy venue and deposit to the sell venue on a compatible network. Generic green deposit/withdraw badges are not sufficient.
DEX quotes preserve chain, contract, quoted notional, route plan, gas, slippage tolerance and provider price impact when supplied. Gas adjusts the opening edge once; price impact embedded in quoted buy/sell prices is not charged twice. Larger size, changing gas, MEV and transaction failure remain explicit limitations.
Versioned score observations are frozen before 8h/24h outcomes exist. A separate daily public-market backfill uses only each observation's preceding 24 hours and excludes historical DEX rows because old chain, contract and rail evidence was not retained. It cannot satisfy identity, exact-cost or activation gates. A future model must use purged chronological train-calibration-test splits with a 24-hour embargo and at least three walk-forward folds, pass leakage and class-balance checks, beat prevalence baselines, produce calibrated probabilities, survive drift checks, and run in shadow mode for at least 14 days.
Only explicitly contributed completed positions can supply route-level median total and component lifecycle-cost percentages. Fees, borrow, gas, transfer and measured slippage are included for research economics; slippage remains embedded in actual portfolio fills and is never deducted twice from portfolio PnL. Exact DEX chain and contract must match before DEX evidence is reused.
No OpenAI, Anthropic, Gemini, Ollama or predictive model is configured. Activation remains impossible until at least 5,000 labeled 24-hour outcomes across 100 routes and 30 days exist, account-cost coverage is sufficient, and a tested deterministic fallback is present. Eight-hour labels are monitored separately and cannot inflate this gate. LLM margin predictions are explicitly excluded.
Current public API or streaming book inside the freshness window.
A completed venue funding payment, not a projection.
A scenario using explicitly displayed assumptions.
Evidence is missing; no positive assumption was inserted.