The record · 記録
One frozen rule, marked to market.
Including every drawdown.
Strategy v1.0 is the site's published rule: hold the top 5 of the 15 majors by 60-day momentum while BTC is stacked above its 50-day average, skip the most crowded perp-funding longs and anything under $5M daily volume, rebalance every 14 days, else sit in cash. This page recomputes that frozen rule over public bars every day and shows it against just holding BTC. It is a backtest with a growing published tail, not a promise, and none of it is advice.
Read this first. Everything before 2026-07-13 is a recomputed backtest; live results are always lower. The board the rule reads from is attested daily since 2026-07-13, and its selection panel publishes this rule's output since 2026-07-17, so the record accrues in public from there. Costs of 20bps per side are charged; the window covers two historic bull markets and two multi-year drawdowns.
Parity gate: locked validation = 70.9x through 2026-07-15; this build recomputed 70.9x. The builder refuses to publish outside a narrow band around the locked number.
Log scale, $100,000 base. — v1.0 · - - BTC buy-and-hold · dashed vertical = publication begins. Data through 2026-07-17.
| year | 2019 | 2020 | 2021 | 2022 | 2023 | 2024 | 2025 | 2026 |
|---|---|---|---|---|---|---|---|---|
| v1.0 | -19.2% | +75% | +1,682.9% | -28.9% | +140% | +113.2% | -9.4% | -21% |
| BTC | -13.2% | +301.7% | +57.6% | -65.3% | +154.5% | +111.8% | -7.3% | -28% |
Recent holding spells
| coin | entered | exited | days | return |
|---|---|---|---|---|
| LINK | 2026-04-21 | 2026-06-02 | 42 | -10.9% |
| TRX | 2026-04-21 | 2026-06-02 | 42 | -0.1% |
| BTC | 2026-04-21 | 2026-06-02 | 42 | -12.5% |
| DOGE | 2026-05-05 | 2026-06-02 | 28 | -19.2% |
| BNB | 2026-05-19 | 2026-06-02 | 14 | +1.7% |
| ETH | 2026-04-21 | 2026-05-19 | 28 | -9.2% |
| XLM | 2026-04-21 | 2026-05-05 | 14 | -10.8% |
| XLM | 2026-03-24 | 2026-04-07 | 14 | -5.9% |
| AVAX | 2026-03-24 | 2026-04-07 | 14 | -2% |
| TRX | 2026-03-24 | 2026-04-07 | 14 | +3.1% |
| LTC | 2026-03-24 | 2026-04-07 | 14 | -1.8% |
| BCH | 2026-03-24 | 2026-04-07 | 14 | -6.6% |
| SOL | 2026-01-13 | 2026-01-27 | 14 | -12.6% |
| BNB | 2026-01-13 | 2026-01-27 | 14 | -4.8% |
| TRX | 2026-01-13 | 2026-01-27 | 14 | -4% |
| BCH | 2026-01-13 | 2026-01-27 | 14 | -2.7% |
| ETH | 2026-01-13 | 2026-01-27 | 14 | -9% |
| SOL | 2025-10-07 | 2025-10-21 | 14 | -15.6% |
| AVAX | 2025-10-07 | 2025-10-21 | 14 | -30% |
| LINK | 2025-10-07 | 2025-10-21 | 14 | -19.2% |
Newest 20 of 224 closed spells since 2019-10-01; open positions marked. A spell is one coin held between rebalances under the frozen rule.
The full derivation, the validation battery this rule passed, and everything that failed it are in the research library: what survived the gate covers the funding filter, walk-forward validation the method. The regime gate itself is on the dial page, and today's selection on the board.
Questions
Is this a live track record?
It is two things, split honestly on the chart. The whole curve is a backtest: the frozen v1.0 rule recomputed over public daily bars, net of 20bps-per-side costs and real perp funding exclusions. The published era is the part you no longer have to trust us on: Koryu's board has been attested daily since 2026-07-13 and the strategy's daily selection has been published on it since 2026-07-17, so from there the record accrues in public, one committed snapshot at a time.
Does Koryu trade this?
No. Koryu holds no assets and executes nothing. The page marks a frozen rule to market because a measurement site should show what its own published rule would have done, including every drawdown, rather than implying it.
Why show BTC buy-and-hold next to it?
Because for crypto, beating a savings account is not the question; beating just holding bitcoin is. Every figure here is shown against BTC over the identical window, and the comparison stays on the page in the years it is unflattering.
What stops the numbers from being quietly retuned?
Three things. The spec is frozen and published (the exact rules are on this page). The builder recomputes the locked validation result before every publish and refuses to write if it cannot reproduce it; the parity check is printed above the chart. And the board snapshots the rule reads from are committed to a public hash chain before outcomes are knowable.
Is a 70x backtest realistic to expect?
No. Backtests overstate live results, this window contains two historic bull markets, and the rule spends most of its days below a previous peak, including multi-year drawdowns near -45%. The honest reading is the shape, the discipline, and the gap to passive holding, not the end value. Decisions are yours.