Scoreboard
Generated 2026-09-07 17:25 UTC. multiples of holding Bitcoin; counts; dates. Hold-BTC is always 1.000.
Live book
BTC book only (Kraken). Capital is never shown. The on-chain pot is a separate USDC ledger. Half-size at 15 % drawdown, halt at 25 %; not halted. Tokenised-equity pot: 1.000 × USDC, its own ledger.
Shadow books
Paper books on the same prices and fee model as the live book, since 2026-09-07. 90 days, then must beat hold-BTC on both multiple and drawdown to be promoted. Promotions to date: 0.
Rebased 2026-09-07 17:23 UTC. A structural flaw was found while preparing this page: until then the books ran on the engine's default parameters (an older brain configuration, unlimited alt slots, defensive sleeve on) on their own paper base from 2026-08-31, saw the live book's holdings as their own, and the invariant compared raw balances, so it went blind at the 2026-09-02 capital change. Every book was restarted from the live book's holdings with the live parameters; each row is now the live book with one change, and the invariant is a ratio of multiples since the reset. The pre-reset week is archived, not shown. Rows identical to book_default mean the switched feature has not fired yet.
| book | × hold-BTC | max DD % | trades | fees % | days |
|---|---|---|---|---|---|
| alt_basket_off Live rules with no alt sleeve at all. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| book_default The live book on paper: same rules, same holdings at the reset, same prices, same fee model. The invariant; it must track the real book. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| constant_mix_70_30 Control. A fixed mix rebalanced inside a band; no signal at all. | 1.000 | 0.00 | 3 | 0.150 | 0.0 |
| dca_reentry Live rules, but exits back to BTC in weekly tranches instead of at once. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| gex_regime Control. Long or flat on the options dealers' gamma regime. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| ichimoku_switch Control. Long or flat on a daily Ichimoku composite. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| liq_reversal Control. Buys after long-liquidation flushes, expecting the bounce. | 1.000 | 0.00 | 1 | 0.060 | 0.0 |
| ma200_switch Control. Long BTC above the 200-day average, out below it. The naive investor's rule. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| maxpain_gap Live rules plus an options max-pain rule near expiry, pre-registered from a public claim. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| no_blackout Live rules without the event blackout. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| no_regime Live rules with the regime forced to neutral. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| pillar_basis Live rules plus one extra input: the futures basis. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| pillar_cb_premium Live rules plus one extra input: the Coinbase premium. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| pillar_mvrv Live rules plus one extra input: MVRV-Z. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| pillar_ribbons Live rules plus one extra input: hash-ribbon capitulation. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| single_classifier_defensive Live rules with a single classifier instead of the ensemble, no weekly gate. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| tranche_entries Live rules, but entries in daily tranches instead of at once. Reset artefact: on 2026-09-07 it treated the held coins as fresh entries and stepped them down to tranche 1/3. | 1.000 | 0.00 | 2 | 0.108 | 0.0 |
| voices_weighted Live rules plus one extra input: the scored stance of monitored analysts. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
| xstock_off Live rules with no tokenised-equity sleeve. | 1.000 | 0.00 | 0 | 0.000 | 0.0 |
Research ledger
Windows: post-ETF epoch, Jan 2024 to today: fit Jan 2024 to Aug 2025 (the only ranked window); holdout, the last 12 months (PASS/FAIL, never ranked); pre-ETF from 2020 and from 2015 as stress: can fail a line, never promote one. Every trial is pre-registered; the ledger is the denominator for the deflated Sharpe ratio.
Decision ledger
Every live proposal — approved or vetoed — scored against the asset/BTC ratio at fixed horizons. Too few rows to mean anything yet; shown anyway. Rows scored: 14.
| verdict | horizon (d) | n | mean ratio return % |
|---|---|---|---|
| vetoed | 1 | 2 | -0.95 |
| approved | 1 | 5 | 3.16 |
| vetoed | 3 | 2 | 3.21 |
| approved | 3 | 5 | 9.21 |
Kitsunebi
72 rows are retro (existed before the ledger; their proof dates the anchoring day, not the git date). First non-retro registration 2026-09-04. The ledger.