Filed before it happened
On November 4, 2026 we wrote down four rules: the three that passed on the shelf, and one survivor of a search of about 640 rules, filed as a test of luck (by our own count of tries, one survivor is about what luck gives). They are filed exactly as they were run on data to October 30, 2026, with the ranges each one’s own record shows, and sealed the filing with a fingerprint. Every trade from November 2, 2026 on is one nobody could have fitted. This is not a forecast. It is a dated commitment that can be checked, and it is checked after each fresh data download, a miss reported as readily as a hit.
Filed by J. Berkowitz, for Edgeproof LLC. Who signs what
What this can and cannot claim
- Three were chosen because they passed on the years before; the fourth because it survived a search. Filing them fixes the rules from here on; it does not undo the picking that found them. Hundreds of rules were tried on this site; these four are what got through, and one of them is about what luck alone gives.
- The ranges below come from each rule’s own history: what it made over every stretch of 1 to 24 months since 2008, nine times in ten. They know nothing of a changed market, so they are probably too narrow.
- A month or a year inside a range is not proof the rule works; one outside is not proof it is broken. It is the first evidence that was not part of the test.
- Nothing here is traded by us, and nothing is advice. Backtest results are hypothetical; the months after the filing are the only test that is not.
The filing
Filed November 4, 2026, 2:26 PM Eastern, on data to October 30, 2026; the months ahead count from November 2, 2026. The filing’s fingerprint (the SHA-256 of the filed document, the same one written in its dated record): d7feb66ab197e90d9aa621d77c72e5aef3e988c28b47c641869c43a09ed26ba8
Witnessed outside this site: the Internet Archive saved this page, every fingerprint on it, on November 8, 2026 at 15:42 UTC, four days after it was filed: the archived copy. Compare its fingerprints with the ones here; a filing changed after that day would not match.
The morning move on NQ (shelf)
As run: 716 trades from 2008-01-04 to 2026-10-14, $113,468 after costs, 90 out of 100, Passed; about 38 trades a year; the longest wait for a new high 46 months.
Over the 12-month stretches in its own record: between -$1,841 and $23,156 nine times in ten (the middle $2,398); 22% of its 12-month stretches lost money. Over any one month: between -$2,261 and $4,810.
the rule’s fingerprint (the SHA-256 of its settings as filed): 28491b77bd7af55d5171f9a96fea1dc53aa2b7c2b923ae6024ac0de8c43aa4ac
The morning move on RTY (shelf)
As run: 797 trades from 2008-06-27 to 2026-10-14, $40,473 after costs, 92 out of 100, Passed; about 42 trades a year; the longest wait for a new high 36 months.
Over the 12-month stretches in its own record: between -$3,382 and $10,492 nine times in ten (the middle $1,375); 24% of its 12-month stretches lost money. Over any one month: between -$1,351 and $1,972.
the rule’s fingerprint (the SHA-256 of its settings as filed): 5290903b8117df35fc7564471064c6e5ce2c0b5044b60fa5f1f3495aaf6b1d19
The morning move on YM (shelf)
As run: 479 trades from 2008-01-04 to 2026-10-23, $27,428 after costs, 89 out of 100, Passed; about 25 trades a year; the longest wait for a new high 61 months.
Over the 12-month stretches in its own record: between -$2,796 and $5,497 nine times in ten (the middle $1,250); 33% of its 12-month stretches lost money. Over any one month: between -$901 and $1,740.
the rule’s fingerprint (the SHA-256 of its settings as filed): ebe379b050aa45922460a5196e2fa9919c4662771427e817931529db18b27e8b
The VWAP stretch fade on NQ, on days inside yesterday's range (SEARCH 10)
As run: 529 trades from 2008-01-04 to 2026-10-30, $39,338 after costs, 86 out of 100, Passed; about 28 trades a year; the longest wait for a new high 48 months.
Over the 12-month stretches in its own record: between -$1,614 and $8,980 nine times in ten (the middle $601); 27% of its 12-month stretches lost money. Over any one month: between -$838 and $1,759.
the rule’s fingerprint (the SHA-256 of its settings as filed): 64fb3fb70f28d7f49659726140ae59cf7259c7436db413bb155905f4484a7057
How they are doing
Checked November 4, 2026 against the filing: each rule run again at its filed settings on the data then held. The trades before November 2, 2026 had to land on the filing to the cent first (they did, all four); the trades since are read against what the record implied.
| rule | before November 2, 2026 | since November 2, 2026 |
|---|---|---|
| The morning move on NQ (shelf) | 716 trades, $113,468, as filed | 0 trades, $0; no trades yet |
| The morning move on RTY (shelf) | 797 trades, $40,473, as filed | 0 trades, $0; no trades yet |
| The morning move on YM (shelf) | 479 trades, $27,428, as filed | 0 trades, $0; no trades yet |
| The VWAP stretch fade on NQ, on days inside yesterday's range (SEARCH 10) | 529 trades, $39,338, as filed | 0 trades, $0; no trades yet |
No trading day on or after November 2, 2026 is on the data yet. The first check that can say anything is about three months out (each rule trades two or three times a month). This table is rewritten after every data download.
The second filing: two that nearly passed, with nothing broken
Filed November 16, 2026, 11:15 PM Eastern, on the same data, to October 30, 2026. Of the shelf’s Closes, these are the two whose score passed and whose only problem is too few trades to rule out a loss: each made money on 2008 to 2022 and since 2023, and the second look on its page says why no change is offered. So none is made. The trades from November 2, 2026 to the day before November 19, 2026 came after the data but before this filing, so they are shown apart and not counted; the months ahead count from November 19, 2026. The filing’s fingerprint: 8fa7b33c940a5fe5310c74aebc57c63a290cf01c65fcb2bc6f0062f4c48aafe3
Witnessed outside this site: the Internet Archive saved this page, every fingerprint on it, on November 17, 2026 at 04:30 UTC, fifteen minutes after this filing and two days before the months it counts from: the archived copy. Compare its fingerprints with the ones here; a filing changed after that would not match.
Picking two of the Closes is a choice too, made by that stated rule, not by how they did since. An 89 and an 80 that fall short on their number of trades are what near misses look like; the months ahead are the test.
The morning move on ES (shelf)
As run: 548 trades from 2008-01-04 to 2026-10-14, $43,861 after costs, 89 out of 100, Close; about 29 trades a year; the longest wait for a new high 55 months. 2008 to 2022: $29,201; since 2023: $14,660.
Over the 12-month stretches in its own record: between -$2,703 and $8,824 nine times in ten (the middle $1,450); 25% of its 12-month stretches lost money. Over any one month: between -$1,193 and $2,373.
the rule’s fingerprint (the SHA-256 of its settings as filed): 91f97d64bcae5776fbae60c01cc768b7d5010b5f7c30978b656d03ff10f6d61b
The 38.2% pullback on NQ (shelf)
As run: 1,622 trades from 2008-01-03 to 2026-10-20, $63,478 after costs, 80 out of 100, Close; about 86 trades a year; the longest wait for a new high 119 months. 2008 to 2022: $52,449; since 2023: $11,029.
Over the 12-month stretches in its own record: between -$3,773 and $26,900 nine times in ten (the middle $472); 46% of its 12-month stretches lost money. Over any one month: between -$3,758 and $5,481.
the rule’s fingerprint (the SHA-256 of its settings as filed): c90438e57ea9a8041196fd89d659bc4389e29a9f400fd6984ff054c22d7964e6
| rule | to October 30, 2026 | November 2 to November 18, not counted | since November 19, 2026 |
|---|---|---|---|
| The morning move on ES (shelf) | 548 trades, $43,861, as filed | not checked yet | 0 trades, $0; no trades yet |
| The 38.2% pullback on NQ (shelf) | 1,622 trades, $63,478, as filed | not checked yet | 0 trades, $0; no trades yet |
Not yet checked against a fresh download; the first check is run with the first filing’s, and this table is rewritten then.
Raw market data: FirstRate Data (firstratedata.com). A measurement on a file, not advice.