The public ledger
A track record is only worth something if it cannot be edited afterwards. Every signal Provyan detects is written here before its outcome exists, and the file is built so that changing any past line breaks it.
7f79e3beda3ac5681093a815cc809c3192aaac16a4e165d5223baa9cc5c02089
Replay: the call rule on 12 months it never saw
The live ledger is the only record that cannot be tidied up, and it is young. Until it has months behind it, this replay is the closest thing to a track record: every study was recomputed using candles up to 14 Nov 2025 only, and the exact rule the ledger uses (a proof score of 55 or more, with a lean) was then applied to each signal that fired from 15 Nov 2025 to 7 Nov 2026. Each call is scored on the close 7 days later, after 0.3% round-trip costs.
| Calls | Count | Right | Ordinary day | Average after costs |
|---|---|---|---|---|
| All calls | 596 | 60% | – | +1.5% |
| Up calls | 292 | 45% | 41% | −0.6% |
| Down calls | 304 | 76% | 59% | +3.5% |
| Strong evidence only (score 70+) | 357 | 59% | – | +1.6% |
“Ordinary day” is how often the price moved that way over 7 days on any day in the same period, across all 94 coins. It was a falling year: prices were higher a week later only 41% of the time, so down calls had the wind behind them. Both kinds of call beat their own base rate.
| Month | Calls | Right | Average after costs |
|---|---|---|---|
| Nov 2025 | 17 | 47% | −4.2% |
| Dec 2025 | 33 | 79% | +4.8% |
| Jan 2026 | 63 | 54% | +1.6% |
| Feb 2026 | 70 | 57% | +1.0% |
| Mar 2026 | 35 | 74% | +2.7% |
| Apr 2026 | 52 | 69% | +1.8% |
| May 2026 | 58 | 59% | +5.2% |
| Jun 2026 | 39 | 67% | +7.0% |
| Jul 2026 | 37 | 59% | −6.3% |
| Aug 2026 | 69 | 62% | +2.1% |
| Sep 2026 | 65 | 49% | −1.5% |
| Oct 2026 | 51 | 57% | +1.3% |
| Nov 2026 | 7 | 57% | −0.5% |
Limits: this is a backtest, not the live record. The signal rules and the scoring were written in 2026, after the period was over, which flatters any backtest a little. 21,414 signals fired in the period; 596 met the bar for a call. Rebuilt every week from the same public data.
How it works
- SealMinutes after the daily close, the signals that fired are hashed together with a random salt. Only that fingerprint is published.
- RevealAt the next daily update the signals and the salt are published. Anyone can hash them and check they match the fingerprint.
- Grade7 days after the signal, the closing price is recorded and the call is marked hit or miss. Nothing is removed.
Each line also contains the hash of the line before it. Editing or deleting an old entry would change every hash after it, and the head shown above would no longer match copies other people have saved.
A directional call is only made when the proof score at that moment is 55 or higher. Signals below that are recorded and graded as “tracked”, without counting toward the hit rate. If one coin's signals point in opposite directions at the same close, none of that coin's calls stand. A batch sealed more than six hours after the close is marked late and left out of the record.
Sealing is automatic: a scheduled job runs a few minutes after 00:00 UTC every day, fetches the closed candles, and writes the batch, with no one in the loop. The very first batch, for the close of 13 Nov 2026, was sealed when the ledger opened at 14 Nov 2026, 19:45 UTC, which is why it is marked late and does not count.
Sealed, not yet revealed
| Close of | Sealed at | Signals | Fingerprint (SHA-256) |
|---|---|---|---|
| 13 Nov 2026 | 14 Nov 2026, 19:45 UTC (late) | 4 | dbd93907c621a6291ecd9ff2a0170d581d0e761e6d072dd8d75cc132077a0c2a |
| 14 Nov 2026 | 15 Nov 2026, 00:16 UTC | 58 | 5632629cf9737c7899d182af80a72d3565df4168dbab481c34a6e5ff60e99a3e |
Members can read these now.
Revealed signals
The ledger started on 13 Nov 2026. The first batch is revealed one daily update after it was sealed.
Showing the last 45 days. The downloadable file has everything.