Scoreboard
A read nobody checks is entertainment. Each one is graded once its session has closed, against a rubric fixed in advance. Misses stay up.
Bull primary on 7 of 16. 6 sessions went the other way.
| Session | Grade | Primary | Happened | Path |
|---|---|---|---|---|
| Tue 01 | A | ▼bear | ▼bear | bear primary live 13:08, holds to close |
| Wed 02 | B | ▼bear | ▲bull | bull third live 10:01, holds to close |
| Thu 03 | A | ▲bull | ▲bull | bull primary live 9:31, holds to close |
| Fri 04 | B | ▲bull | ▼bear | bear third live 10:54, holds to close |
| Tue 08 | A | ▼bear | ▼bear | bear primary live 9:51, holds to close |
| Wed 09 | B | ▼bear | ▼bear | bear primary live 11:11, holds to close |
| Thu 10 | A | ▼bear | ▼bear | bear primary live 9:31, holds to close → bear third live 9:31, holds to close |
| Fri 11 | B | ▼bear | ▲bull | bull third live 9:31, holds to close |
| Mon 14 | B | ▼bear | ▲bull | bull third live 11:44, holds to close |
| Tue 15 | A | ▼bear | ▼bear | bear primary live 10:22, holds to close |
| Wed 16 | A | ▼bear | ▼bear | bear primary live 14:33, holds to close → bear third live 14:53, holds to close |
| Thu 17 | C | ▲bull | ▲bull | bull primary live 15:52, holds to close |
| Fri 18 | C | ▲bull | ▼bear | bear third live 12:24, holds to close |
| Mon 21 | A | ▲bull | ▲bull | bull primary live 10:23, holds to close |
| Tue 22 | C | ▲bull | ◆inside | no branch went live — price stayed inside every level the read drew |
| Wed 23 | B | ▲bull | ▼bear | bear third live 10:11, holds to close |
Primary is the direction of the scenario the read ranked first — it publishes three, and this is the one it led with. Happened is the first branch the market actually took. Inside means none of them went live: price stayed within every level the read drew. That is recorded as a miss rather than a draw, because the read still has no way to say “today balances” as a scenario that could be scored right.
It grades the quality of the description, not profit. There are no trades behind these reads, no simulated fills, and no track record being implied. The question the grader answers is narrow: did this read describe the session that actually happened, and did it point at what turned out to matter?
Everything above is the forward record: reads published before the session opened, on a date you can check, graded afterwards. It is the only number here that is a track record, and it grows one session at a time.
Separately, the framework is run over historical sessions to test the method at a scale the forward record cannot reach yet. Those reads are blinded the same way — every figure cut at 08:05 on the morning in question — but they were written after the fact, and there is no way to prove that the model writing them had no residual knowledge of how those sessions resolved. So they answer a different question: does the method work? rather than is the record real?
They are reported apart and never summed. A backfilled read is marked as one, is excluded from every figure on this page, and is not offered as evidence of anything except the framework itself. Combining the two would produce a larger number and a smaller claim.
A grade is a machine opinion, not a measurement. The verifiable part is the read itself: the levels were published before the open, and every figure in it can be checked against the tape. Weigh the grades accordingly.
The pre-market read is published pre-open. Same structure every time: what the auction did, what it is trying to do, how well it is doing it, and what it is likely to do next.