The record
A backtest can be argued with forever. What cannot be argued with is a claim written on a date, committed to a public repository, and scored a year later. This page is that: what every instrument on this site said each day, scored once the year has passed; and a register where anyone can enter a rule before seeing its result, to be scored on the same terms from that day forward.
Computed from the daily snapshot for 10 September 2026
The ledger began on 2026-09-08 and holds 3 days. The first row becomes scorable on 2027-09-08; until then it accumulates. The register holds 13 rules, of which 13 are the site’s own, seeded rather than pre-registered.
Every version of both files is kept in the repository’s git history, so no entry can be changed after the fact without the change being visible.
What we said, and what happened
No row is a year old yet. What is shown is what the site is saying now, written down; the scoring column fills in from 2027-09-08.
| Date | Price | Composite | Rules firing | A year later |
|---|---|---|---|---|
| 2026-09-10 | $78,273 | 39 average | Public attention in its lowest sixth; Price below the Stock-to-Flow model; Price below the 2-year average | awaiting |
| 2026-09-09 | $78,444 | 39 average | Public attention in its lowest sixth; Price below the Stock-to-Flow model; Golden cross; Price below the 2-year average | awaiting |
| 2026-09-08 | $79,094 | 40 average | Price below the Stock-to-Flow model; Price below the 2-year average | awaiting |
The register
13 rules registered. 0 pre-registered by readers; at 90% confidence about 0.0 of those would look good by chance. The seeded rules are the site’s own and carry no such credit.
| Rule | Registered | By | Since registration | Full history | Today |
|---|---|---|---|---|---|
| RSI-14 below 30 → a doubling within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 21 eps · 47.6% vs 59.8% · 2 open | quiet since 2026-06-11 |
| RSI-14 above 70 → a 40% fall within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 20 eps · 30.0% vs 38.6% · 1 open | quiet since 2026-09-04 |
| Mayer multiple (price / 200-day average) above 2.4x → a 40% fall within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 8 eps · 25.0% vs 26.4% | quiet since 2021-03-14 |
| Price / 200-week average below 1.0x → a doubling within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 4 eps · 75.0% vs 46.0% · 1 open | quiet since 2026-08-17 |
| Price / 2-year average below 1.0x → a doubling within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes · 1 open | 4 eps · 25.0% vs 44.9% · 1 open | firing |
| 50-day average / 200-day average crosses above 1.0x → a doubling within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 12 eps · 66.7% vs 57.0% · 1 open | quiet since 2026-09-09 |
| 50-day average / 200-day average crosses below 1.0x → a 40% fall within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 12 eps · 58.3% vs 28.6% · 1 open | quiet since 2025-11-17 |
| 111-day average / (2 x 350-day average) above 1.0x → a 40% fall within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 5 eps · 100.0% vs 26.0% | quiet since 2021-04-23 |
| MVRV above 3.7 → a 40% fall within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 5 eps · 20.0% vs 28.1% | quiet since 2021-03-14 |
| MVRV below 1.0 → a doubling within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 5 eps · 60.0% vs 54.8% | quiet since 2023-01-13 |
| Fear and Greed index below 20.0001 → a doubling within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 5 eps · 20.0% vs 51.7% · 1 open | quiet since 2026-07-08 |
| Fear and Greed index above 79.9999 → a 40% fall within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 5 eps · 40.0% vs 39.2% | quiet since 2025-01-22 |
| Hash rate, 30-day / 60-day average below 1.0x → a doubling within a year (seeded) | 2026-09-10 | cryptoexponentials.com | 0 episodes | 15 eps · 60.0% vs 53.3% · 1 open | quiet since 2026-08-20 |
To register a rule: open a GitHub issue with the “Register a rule” template. Choose the rule before you look at what it would have done. It will be scored every morning from the day it is entered, its result kept whether it works or not, and its since-registration record shown separately from its history — because only the forward record is evidence.
Why the register separates two records
A rule’s full history answers “would this have worked”. It is what every backtest reports, and it is where thresholds get chosen after seeing the answer. The record since registration answers “did it”, on episodes that had not happened when the rule was written down. Only the second is a prediction. The register shows both, side by side, so the difference is visible.
Thirteen of the site’s own rules are in the register, marked as seeded — the thirteen that can be expressed in the rule builder’s vocabulary. Two scorecard rules cannot be: LPPLS, which uses a different statistical design, and Stock-to-Flow, which needs the supply series. Nor can the six site-specific rules, which read the composite. The seeded rules’ thresholds were chosen after seeing history; they are not pre-registered and get no credit for being here. They are entered so their forward record runs on the same terms as everyone else’s.
What the ledger is worth today
Nothing, yet. The first row can be scored a year after the ledger began, and the page says on which date. Every day adds a row; every day after the first anniversary scores one. The value is the accumulation, and the fact that every version of the file is in the repository’s history, so no row can be quietly changed after the fact.
The rules whose readings are recorded here are scored on their full history on the scorecard, and a rule of your own can be tried first in the rule builder.
Written daily by fetch/ledger.py and fetch/registry.py; published as ledger.json and registry_scores.json.