Sealed test · future data
A rule sealed before the data existed
On 29 September 2026 we froze one trading rule, 25 markets and the pass mark. The site follows the rule every day from 30 September 2026. Decision on 1 October 2027, or on 1 October 2028 at the latest. The result will be published whatever it is.
Today's state
Indicative only. No decision is taken before the dates above.
Chance that the same entries, moved to other days with the rule replayed, did as well or better: p = 0.040 (24 other placements, smallest possible value 0.040). Markets where the entries did better than chance: 6 of 10. Same check with entries one day later: p = 0.040.
Pass mark at a decision date: p ≤ 0.005, the rule ahead of chance on at least 60% of the markets, at least 60 trades (100 at the second date), and p ≤ 0.05 with entries one day later.
The 25 markets
| Market | Trades | In position | Rule | Buy & Hold | Last day |
|---|---|---|---|---|---|
| AEX | 0 | — | +0.0 % | +2.3 % | 2026-11-12 |
| ASX 200 | 0 | — | +0.0 % | +4.1 % | 2026-11-12 |
| Bitcoin | 0 | — | +0.0 % | -14.6 % | 2026-11-12 |
| Bovespa | 2 | — | -2.7 % | -8.1 % | 2026-11-12 |
| CAC 40 | 1 | — | +0.6 % | +5.0 % | 2026-11-12 |
| Copper | 0 | — | +0.0 % | -5.7 % | 2026-11-12 |
| DAX | 0 | — | +0.0 % | +6.2 % | 2026-11-12 |
| Ethereum | 3 | — | +4.3 % | -24.8 % | 2026-11-12 |
| FTSE 100 | 0 | — | +0.0 % | +2.4 % | 2026-11-12 |
| Gold | 1 | — | +0.6 % | -6.8 % | 2026-11-12 |
| Hang Seng | 1 | — | +0.6 % | -2.5 % | 2026-11-12 |
| IBEX 35 | 0 | — | +0.0 % | +2.8 % | 2026-11-12 |
| IPC Mexico | 0 | — | +0.0 % | +11.8 % | 2026-11-12 |
| KOSPI | 3 | — | -1.7 % | -21.9 % | 2026-11-12 |
| Nikkei 225 | 1 | — | +0.9 % | -8.9 % | 2026-11-12 |
| Nasdaq 100 | 2 | — | +1.5 % | -7.0 % | 2026-11-12 |
| NZX 50 | 0 | — | +0.0 % | +2.5 % | 2026-11-12 |
| Oil | 0 | — | +0.0 % | +0.3 % | 2026-11-12 |
| Sensex | 0 | — | +0.0 % | +13.0 % | 2026-11-12 |
| Silver | 3 | — | +0.8 % | -17.2 % | 2026-11-12 |
| SMI | 0 | — | +0.0 % | +9.5 % | 2026-11-12 |
| S&P 500 | 0 | — | +0.0 % | -0.7 % | 2026-11-12 |
| Euro Stoxx 50 | 0 | — | +0.0 % | +2.1 % | 2026-11-12 |
| TAIEX | 3 | — | -1.6 % | -10.1 % | 2026-11-12 |
| TSX | 0 | — | +0.0 % | +2.6 % | 2026-11-12 |
Paper test at the close of each day, fees included. The rule is in the market only 3 to 10% of the time: it is expected to stay far behind Buy & Hold. The test measures when it enters, not how much it earns.
Why this test
On past data the rule, a dip buyer (RSI below 30 and close under the lower Bollinger band), came out promising but not confirmed in two sealed tests (p = 0.004 then 0.020 and 0.019, with a pass mark of 0.01). Past data has already been looked at, so it can no longer settle the question. Only days that did not exist when the rule was sealed can.
Replayed on the past with the same test, the rule would have passed on 2 one-year windows out of 26. Even if it behaves as it did, this test has about one chance in six to conclude. A failure will close this line of research.
Check it yourself
- Sealed protocol, version 2 (French) — SHA-256
31d0179c8ef982fa0486eb0f9e76518aa8f8c66900ae0afa2007fabcb21cdfb1 - Rule script — SHA-256
13b0c3741618ea718d38878f3e80539c42a4a3347c794d4662d44ee20c07e998 - Seal with its time
- Version 1, replaced before any tested day, and why
- Today's data archive — SHA-256
2c30e6b2d46ea6a2b7600e1b4f15b73c869c3a532f8c046ce2d2f203e802da52(2026-11-13T22:40:04+00:00)
A seal time written by us proves nothing on its own: that is why the protocol file is public from day one and its fingerprint will not change.
Paper trading. No profit promise. Methodology · See also: the AI researcher leaderboard