Holders and share of backing
Who holds the tokens BTI measures, and how each token's bytes divide among its holders: the holders appendix, the rule of shares, coverage, and the leaderboard population.
BTI attributes bytes to tokens, never to addresses. To say how much blockspace stands behind an address, something has to know who holds each token. At Blockspace Holdings that is the job of the holders appendix. This page covers what the appendix is, the rule by which a token's bytes divide among its holders, what it can and cannot see, and the population the block.space leaderboard ranks.
The holders appendix #
For every protocol the books measure, it answers who holds each token, how many units, and therefore how many of the token's bytes back each address. It is a daemon beside BTI's tiers: one service, one cursor, one database schema, seeded once from every source and then advanced block by block. The pages read it; it never reads them.
Its rules include:
- The engine is not changed. The appendix reads BTI's tables and never writes them.
- Built once, advanced by deltas. The first pass seeds from every source; each later pass applies what changed. This follows the ruleset's standing law (2026-10-18) that every BTI index is built once and keeps its state; a whole rebuild has two causes only, a ruleset change or recovery after a failed check.
- A pass is one transaction. Readers never see a half-written table, and a failed check rolls the pass back while the last good state keeps serving.
- Scripts, not people. Owners are keyed by scriptPubKey; the address string is a rendering of it. Burn addresses are holders.
- ord is definitional for where sats land. Each pass compares 100 random sat landings with ord's own answer, and a disagreement fails the pass.
The appendix is temporary by construction. The permanent design, BTI's custody draft (0.3, 2026-10-17), makes ownership a T2 fact, "units of token X moved to owner Y in transaction Z", written by each roster unit beside its claims, with per-address books as T3 arithmetic. Nothing of it is built; it is meant to land at the next full resync, when the appendix schema is dropped.
Share of backing #
The rule of shares:
share(X, T) = units(X, T) / units_total(T)
where units_total is the holder total the source reports, never the protocol's stated supply. Each of the token's book components, content, structure and overhead, in bytes and vbytes, is multiplied by the share and summed over the tokens an address holds. So Book C is content, Book B adds structure and Book A adds overhead, and C ≤ B ≤ A holds per address too. This is the address's share of backing.
The rankings documentation states the consequences plainly. An address holding half of a token's units is backed by half of that token's bytes. A non-fungible unit, such as an inscribed sat, has one unit, so its owner carries all of the token's bytes. The share is by units held, not by value.
Every pass ends with checks that must hold before it commits, among them: the shares of each token sum to 1, and the bytes on addresses plus the bytes nobody holds equal the books, per protocol.
What it can and cannot see #
The appendix reads each protocol's holders from its own source: ord's index export plus a block-by-block replay for sats, BTI's adapter state for runes and other UTXO-bound tokens, OPI for BRC-20, Counterparty's state for its assets, and so on. A token whose holders it cannot see contributes nothing to any address, and the rankings rule is that it is "not measured", never 0. The leaderboard's coverage line states what share of the books that leaves out.
As of 2026-10-18, Omni, alkanes and three OP_RETURN script ledgers (brc20-opreturn, op20, brc100) were not measured. Alkanes holders were approved for measurement on 2026-10-19, and the leaderboard's coverage API on 2026-10-20 (pass finished 09:20 UTC, books at block 972,199) listed alkanes among the measured protocols. It gave Omni properties, BRC-100 and OP-20 as not measured, and placed 94.6 % of Book A on addresses.
The leaderboard population #
The first seed landed on 2026-10-19 at 00:08 and recorded 125,438,419 sats and 4,460,161 addresses. On the latest reading the leaderboard ranked a population of 4,491,923 addresses, with holdings as of block 972,199 (as of block 972,199, 2026-10-20 09:22:01 UTC).
The leaderboard ranks one row per address across every protocol the appendix can read, with a composition bar showing how its backing divides among the families. Following the rankings rules, the board says "as of block H", where H is the one block the appendix is synchronised to, and its API carries the exact, unrounded values. See the frontier.
Sats decide the top #
The leaderboard was held back until sats were in, because they decide the top of the table: "a leaderboard without sats ranks wrongly". Sats are the largest family by Book A, and each inscribed sat's bytes go whole to the one address that holds it. See sats and ordinals.
Related #
Sources #
Live figures: holders as of block 972,199, 2026-10-20 09:22:01 UTC. Refreshed hourly.
Updated 2026-10-20 · This page as Markdown