Scores written down before the prices moved

Soundings — prospective hypothesis (registered October 18, 2026; Issue 1: Twelve Networks, Four Questions)

Hypothesis H1: networks scoring 8 or more on the four-question screen will show higher 12-month price returns and shallower maximum drawdowns than networks scoring 4 or less, measured from each issue date. Null: no difference between buckets. Secondary: the same comparison at 3 and 6 months.

Rules

• Each issue's scores and CoinMarketCap 23:59 UTC prices on the issue date are appended to ledger.csv and never edited. Corrections to a past issue are logged as new rows flagged "restated", never overwrites.
• Forward returns are computed from the issue-date price to the price 91, 182 and 365 days later (same source).
• Results are reported by score bucket (>=8, 5-7, <=4) with the number of observations beside each figure, from the fourth issue onward; misses are reported with the same prominence as hits.
• The first full 12-month reading is October 2027. Until roughly 24 issues exist the tables are descriptive, not inferential.
• Price, volume and volatility are reported alongside the screen but never enter the four scores.
• If the hypothesis fails, that is published as the finding.

Universe rule (registered October 19, 2026, before any Issue 2 data was pulled)

• From Issue 2, the screen covers every base-layer network with its own token, a measurable fee market and a circulating market cap above $3 billion (CoinGecko, issue date). Layer-2 networks, stablecoin issuers and application tokens are excluded; if any is ever added it is listed separately.
• Issue 1's twelve were a judgment (largest base layers plus Hyperliquid and Canton) and the note says so. On October 19, 2026 the rule would also have admitted Zcash, Dogecoin, Monero, NEAR, Stellar, Bitcoin Cash, Litecoin and TON.
• A network that falls below the threshold after inclusion stays in the ledger until its 12-month reading is taken.
• Where a metric cannot be measured (e.g. Monero active addresses), the rubric scores 0 and the note says why.

Hypothesis H2 (registered October 19, 2026, before any forward data exists)

• Networks whose relative fundamental rate of change (DEFINITIONS.md, addendum of the same date) is positive at an issue will show higher 12-month price returns than those where it is negative, independent of the level score.
• Secondary: the same at 3 and 6 months; and a joint reading of level (H1 buckets) and direction (H2 sign).
• H2 is tested from Issue 2 onward, when roc_fees, roc_users and roc_rel are recorded in the ledger at issue. The Issue 1 prototype (prototype_quadrant_2026-10-19.png, roc_prototype_2026-10-19.json) is kept for the record but is not a ledger row: it was computed on October 19, 2026, after Issue 1's prices, with users history for four networks starting only in September 2025.
• Same rules as H1: results by sign with observation counts, misses reported as prominently as hits, a failed hypothesis published as the finding.

Applications table (registered October 19, 2026, as an addition; the base-layer rule above is unchanged)

• From Issue 2, Soundings carries a second table of application protocols, scored on the same four questions and ranked within its own class. The two tables are never merged into one ranking; one chartbook page sets cap-to-fees for both classes on a single axis, as context.
• An application qualifies if it has its own token, a protocol-level fee stream that the token has a claim on or governance over, and either a circulating market cap above $3 billion (CoinGecko, issue date) or protocol revenue above $100 million over the trailing twelve months (DefiLlama "revenue", issue date). Revenue, not fees: fees are what users pay; revenue is what reaches the protocol or token holders.
• Stablecoin issuers, layer-2 networks, wrapped and bridged assets, and app-chains with their own base layer (e.g. Hyperliquid, which stays in the base-layer table) are excluded. A protocol with no DefiLlama revenue entry on the issue date is not in; being measurable is part of qualifying.
• Why cap or revenue, not revenue alone: selecting on profit would select on the thing Question 1 scores and remove the high-cap, low-earning cases the screen exists to show. The revenue gate admits earners too small for the cap gate; the cap gate keeps the large names whatever they earn.
• Application rows carry the same ledger columns and enter H1 and H2 within their class, with observation counts reported separately for the two classes.

- The universe rule above, applied on October 20, 2026, admits eight base layers not in Issue 1: Zcash, Dogecoin, Monero, NEAR, Stellar, Bitcoin Cash, Litecoin and TON. They are scored on the Issue 1 definitions and rubric, unchanged, and enter the ledger with issue date October 21, 2026 and CoinMarketCap 23:59 UTC prices that day. - Scores: Bitcoin Cash 6, Litecoin 6, Zcash 4, Dogecoin 4, Monero 4, NEAR 4, TON 4, Stellar 3. Every one trades at more than 1,900 times its fees; seven of eight at more than 5,000 times. - Bittensor ($3.1B) is excluded: it has no DefiLlama fee entry on the data date, and its registration fees are largely recycled into emissions. Being measurable is part of qualifying, as in the Applications rule. Revisited at Issue 2. - Monero supply uses the protocol's fixed tail emission (+0.85%), not CoinGecko's +2.2%, whose year-ago figure disagrees with the emission rule by more than 10%, per the Q3 definition. - Usage data for the eight is thin and mostly single-source; Zcash and Monero are not measurable by address and score 0 under the universe rule. Each Users score is the lower bucket where providers or dates disagree. - The addendum rows join the Issue 1 rows in H1 and in the Screen Tracker from their own issue date. Forward returns are measured from each row's own issue date. - Data: DefiLlama fees and CoinGecko market cap and supply, pulled October 20, 2026; research notes and sources in the addendum data folder, available on request. Trade Bands portfolio selection (registered October 20, 2026, before any application was scored) - The hypothetical Trade Bands portfolio, when it starts, holds the 10 highest-scoring names across the base-layer and Applications tables together, by total score, with no quota per table. The two tables remain separate rankings and separate indexes; only the portfolio's selection crosses them. - Ties break on the lower Question 1 ratio: market cap to fees for base layers, market cap to revenue for applications. - A name must trade on Coinbase spot for US customers to be eligible; ineligible names stay in the screen and the indexes. - Sizing, signals and execution are set in the portfolio rules, registered before the start date.

Definition change: Question 1 on net fees (registered October 20, 2026, effective Issue 2)

• From Issue 2, Question 1 scores fees net of what the protocol pays back to fee payers for paying: per-fee rebates are deducted, and where payers earn minting rewards for paying, only fees from unrewarded payers count. Gross fees stay printed beside the net figure. Networks that pay nothing back are unaffected.
• Issue 1 rows are not edited. Under the new definition Canton's Issue 1 earnings score would be 0 and its total 1, not 3 and 4: in September 2026, 99.99% of its $51.5M traffic burn came from parties the protocol rewards for paying. Both totals fall in the same H1 bucket (4 or less).

Applications table: readings of the rule (registered October 21, 2026, before any application was scored)

• Stablecoin issuers are excluded whatever else they earn. Sky (formerly Maker), which issues USDS and DAI and earns on the collateral behind them, is a stablecoin issuer and is out; so are World Liberty Financial (USD1) and Ethena (USDe).
• A protocol fee stream means fees for using the protocol. Margin on goods the protocol sells (for example Collector Crypt's trading-card packs) is product revenue, not a protocol fee, and does not count toward the revenue gate.
• Exchange-company tokens (for example WBT, LEO) are not on-chain applications and are excluded.

Applications table, first scores (registered October 21, 2026; issue date October 21, 2026)

• Four applications qualify under the Applications rule on October 20, 2026: Uniswap, Pump (pump.fun), Chainlink and Rain. They are scored on the same four questions, ranked within their own class, and marked "(app)" in the ledger. Prices: CoinMarketCap 23:59 UTC on October 21, 2026.
• Scores: Uniswap 9, Pump 7, Chainlink 5, Rain 2. Question 1 uses trailing-twelve-month protocol revenue (DefiLlama), net of payments back to payers.
• Uniswap's supply change uses CoinGecko's corrected year-ago figure (CoinGecko revised it on November 6, 2025); Pump's uses the supply implied by its recorded burns, where CoinGecko's series shows an erroneous flat stretch. Both follow the Question 1-3 definitions and the Monero precedent in the addendum.
• Uniswap's fee switch began December 28, 2025, so its trailing-twelve-month revenue covers part of a year; it is scored on the rule as written. Chainlink's holders revenue goes into a reserve that holds LINK rather than burning it.
• These rows enter H1 and H2 within the Applications class and are reported separately from base layers. They are eligible for the Trade Bands portfolio under the selection rule registered October 20, 2026.

Each published document is locked with a SHA-256 hash recorded in the manifest at lock. A copy whose hash does not match is not the published version.

• Soundings_01_Twelve_Networks_Four_Questions_v1.0.pdf
  59867aac8d3e5d33b276335c13fbddf3c5d7abb0dac955382269c63f58de44fa
• Soundings_01_Chartbook_v1.0.pdf
  694385ef914b6b8e2a3de6ae29b1afd4eb499f267708039696b822a81da3a8d1

The method files that govern future issues (DEFINITIONS.md, HYPOTHESIS.md) are available on request with their dated revisions, as are the scoring code and the data pulls.