Methodology

Each tile on the homepage (gold, equities, stablecoins) divides a live CoinGecko category market cap by an independent, real-world denominator:

• Gold — the tokenized-gold category market cap, ÷ 219,891 tonnes of above-ground gold (World Gold Council / Metals Focus, end-2025) × the live PAXG price (PAXG is redeemable 1:1 for one troy ounce, so it doubles as a live gold price with no separate metals-data vendor needed).
• Equities — the tokenized-stock category market cap, ÷ the World Bank's global listed-domestic-company market capitalization, aged forward by a global-equity index move since that figure was published. ETFs are excluded — they mix bond/commodity exposure and would distort the denominator.
• Money — the stablecoins category market cap, ÷ global M2: live US M2 (FRED M2SL) plus a pinned rest-of-world estimate. FRED has no maintained non-US M2 series (the IMF/OECD sets that exist stopped updating years ago), so a max-observation-age guard rejects any stale series rather than silently averaging a years-old number in as if it were current.

Category market caps are never summed across sectors — CoinGecko's RWA categories overlap, so summing double-counts the same coin.

The Yield Matrix's spread = the best available on-chain APY for a tokenized RWA, minus SOFR (Secured Overnight Financing Rate, the Federal Reserve's benchmark risk-free rate — via FRED). On-chain APY is sourced from DeFiLlama's pool data, gated to pools with real TVL (currently ≥$1M for the headline figure) so a thin, barely-liquid pool can't produce a misleadingly large number — the same principle applied to per-asset yield opportunities and the stablecoin peg-deviation stat below.

The headline APY measures what holding the token earns. Single-asset yield (lending, vaults, staking, yield-bearing tokens) comes first, using the lower of today's base rate and its 30-day average, so a one-day spike can't set the number. Only when a token has no single-asset yield does a liquidity pool count, using its 7-day average fee APY and labeled “LP fees, 7-day avg”. A pool's single-day fee figure is roughly one day of trading annualized and swings widely, and a pair with a volatile token mostly earns from that other token, so it never sets the headline. Every pool, including riskier ones, still appears in each asset's yield opportunities.

The same underlying asset — NVIDIA, gold, Tesla — is often tokenized by several competing issuers as separate coin ids (Robinhood, Ondo, xStocks, Dinari, and others). RWAstats groups these back into one canonical page via CoinGecko's /rwas/* endpoint family: aggregate market cap and volume prefer CoinGecko's own first-party total over a manual re-sum, and access tier / APY are never blended across issuers that genuinely disagree — shown as "Mixed" with the real per-issuer numbers one click away, rather than averaged into a single misleading figure.

A holder count is the sum of each deployed chain's own holder count — it is not a de-duplicated count of unique wallets. The same wallet holding a token on two chains is counted twice. This is a limitation of the underlying on-chain data (per-contract-per-chain), stated wherever a holder count appears on the site.

Deviation = a stablecoin's live price, quoted in its own peg currency, minus 1 — e.g. a token pegged to 1 JPY that currently trades at 1.02 JPY reads as +2.0%. The widest-deviation stat on the homepage is gated to actively-traded stablecoins (24h volume ≥ $50,000) — an illiquid token's stale last-traded price can show an extreme deviation that reflects a lack of trading, not a real market signal.

• Global stats + the discovery-hub asset matrix: revalidated every 60–90s server-side; client refetch every 60–120s.
• FRED benchmarks (SOFR, Treasury yields, M2): revalidated hourly.
• DeFiLlama pool data (on-chain APY, yield opportunities): revalidated roughly every 10 minutes.
• Stablecoin peg comparison: revalidated every 5 minutes.
• Newly-tokenized feed, canonical-RWA groupings, and per-asset on-chain networks/holders: refreshed daily via a scheduled harvest.

Every API response the site makes internally carries a source field ("live" or "mock") — "mock" only appears when an upstream key genuinely isn't configured (e.g. local development without secrets), never as a silent way to fill a gap in live data. See the { API } inspector for the exact call behind any number on the site.

CoinGecko, GeckoTerminal, DeFiLlama, FRED, the World Bank, and the World Gold Council.