Every method this project has verified against its printed source,
and what the implemented ones say about the market right now.
Built 2026-09-18 22:23 UTC.
Status: INSUFFICIENT. Nothing on this page is a signal to trade on. The forward record is 29 trading days (39 calendar days, 2026-08-10 to 2026-09-17), graded once.
Every one of the 3524 grades below was produced on
2026-08-13, 2026-08-17, 2026-08-18, 2026-08-19, 2026-08-20, 2026-08-21, 2026-08-24, 2026-08-25, 2026-08-26, 2026-08-27, 2026-08-28, 2026-08-31, 2026-09-01, 2026-09-02, 2026-09-03, 2026-09-04, 2026-09-07, 2026-09-08, 2026-09-10, 2026-09-11, 2026-09-14, 2026-09-15, 2026-09-17, 2026-09-18, over claim windows
opening 2026-08-10 and closing 2026-09-17 —
21 instruments, 438 distinct instrument‑window
cells, all inside one market episode. 3524 is a count of rows, not a
sample size. Instruments observed over one overlapping fortnight share the
same market‑wide move, so the independent observations here number closer to
one than to 3524, and every significance figure on this page inherits
that.
Over that fortnight the lift was
-3.44pp against the engine's own measured nulls
(847 correct, 2473 incorrect,
204 non‑discriminating) — worse than chance.
The bullish/bearish split inside that number is explained by a rising two‑week
grading window rather than by the doctrine: a drift‑matched placebo carrying
zero doctrinal content reproduces it (z = +1.03, p = 0.14),
and the tape rose across the whole window — every “down” call
missed (2 of 39) while “up” was the only positive cell in the table.
That is one market direction refracted through five labels, not five results.
An effect that survives deletion of the doctrine is not evidence of the doctrine
— and eleven trading days is not evidence of anything, in either direction.
675 of 3524 grades (19%) were produced by a grader whose version was
never recorded. Of the 3524 grades in this ledger, 675 carry
grader_version: analytic-retired-unrecorded — the scoring
rule was versioned only after they were written, and the retired rule was not
stamped. Those grades are not known to be wrong. They are unattributable:
the lift above cannot be assigned to a known scoring rule for
19% of the rows behind it, and neither can any defence
of it. Read together with the eleven‑day span, the headline number now has
two independent reasons not to be read as an edge.
23,823verified passages
47books read
6authors
66rules implemented
4,810live claims
Read this before the numbers above.
23,823 passages are verified, which means the exact words were
found character‑for‑character on the page named. It does not mean
they were tested. Only 66 rules are implemented and
emitting claims. The rest is doctrine we have read exactly and turned into
nothing — usually because the source withholds the operative number.
Current readings
How far each reading stands from the printed page.
A rule is not the same thing as the sentence it came from. Of the
4810 live claims below, 1043 (21%) come from a rule whose
implementation is verbatim — it does what the source says, in the
source's own terms. 3202 come from a
paraphrase, and 565 from an
inference, where the implementer supplied a step the author never wrote
down. Every claim below is labelled. The inferred ones are ours, not Gann's.
How old the market data behind these readings is.
Every level below was computed from the price history as it stood on
2026-09-18 —
0 days ago.
The windows are still open, so the claims still stand and are still gradeable;
but a reading is only as current as the bars it was priced from, and this page
is built more recently than the last scan. The build date at the top of this
page is when the page was made, not when the market was last looked at.
4,810 claims currently in force across
21 instruments and 66 rules — every claim
whose window has not yet closed, not just those recorded most recently.
Newest genuine snapshot 20260918T222318Z. Each claim states a level or a date, a window it must
resolve in, and the tolerance it is judged on — all fixed at emission and
never adjusted afterwards.
AAPL — 258 claims in force
loading…
AMD — 340 claims in force
loading…
BITO — 257 claims in force
loading…
CANE — 195 claims in force
loading…
CORN — 172 claims in force
loading…
CPER — 232 claims in force
loading…
GLD — 243 claims in force
loading…
GS — 261 claims in force
loading…
JPM — 202 claims in force
loading…
NVDA — 299 claims in force
loading…
QQQ — 243 claims in force
loading…
SLV — 241 claims in force
loading…
SOYB — 164 claims in force
loading…
SPY — 172 claims in force
loading…
TLT — 162 claims in force
loading…
UNG — 300 claims in force
loading…
USO — 274 claims in force
loading…
UUP — 87 claims in force
loading…
VXX — 267 claims in force
loading…
WEAT — 219 claims in force
loading…
XLE — 222 claims in force
loading…
What every implemented rule says about the market now
This section is not evidence and is never graded. The claims above were
recorded before their windows opened, which is the only thing that makes them
gradeable. The readings here were recomputed from the price history as it
stood on 2026-08-11. Scoring them against what the market then did
would be scoring a rule against the data it was calculated from — the
oldest mistake in this field, and the reason the two sections are kept apart
rather than added together.
821 readings from 49 rules across
22 instruments, computed over 233,845 daily bars
reaching back to 1927-12-30. 743 of them resolve
before the end of this year; the rest run to 2035-08-06.
Every reading carries the reasoning that produced it and the page of the book it
came from.
34 of the 83 implemented rules say nothing at all, on any of
these 22 instruments. That is the intended behaviour and it is
reported rather than hidden: a rule whose trigger condition is not met abstains,
and a corpus in which two rules in five are currently silent is a corpus that is
not manufacturing signal to fill a page. The silent ones include every Wyler
node, Woods's demand areas, four of the five Martin trend-action rules and four
Gann squaring variants — checked individually against a fully populated
market view, with volume present, so they are abstaining on the market and not
on a missing column.
Instruments here are the research series, not the trading
universe: ^GSPC, ^DJI and ^IXIC are the indices these authors actually wrote
against, and they carry the century of history the long-cycle rules need. The
bars end 2026-08-11; a level is an anchor that only moves when a new
pivot forms, and a projected date is computed from a pivot months or years back,
so both survive that gap — but the gap is stated rather than smoothed
over.
AAPL — 41 readings
loading…
BA — 40 readings
loading…
CAT — 41 readings
loading…
Crude oil (CL=F) — 40 readings
loading…
Dow ETF (DIA) — 31 readings
loading…
Dow Jones (^DJI) — 28 readings
loading…
Gold (GC=F) — 32 readings
loading…
GE — 47 readings
loading…
GLD — 28 readings
loading…
S&P 500 (^GSPC) — 26 readings
loading…
IBM — 41 readings
loading…
Nasdaq (^IXIC) — 39 readings
loading…
KO — 38 readings
loading…
Nasdaq 100 ETF (QQQ) — 45 readings
loading…
Silver (SI=F) — 38 readings
loading…
SLV — 34 readings
loading…
S&P 500 ETF (SPY) — 29 readings
loading…
USO — 44 readings
loading…
XOM — 41 readings
loading…
Corn (ZC=F) — 38 readings
loading…
Soybeans (ZS=F) — 39 readings
loading…
Wheat (ZW=F) — 41 readings
loading…
How readable is the source sentence behind each reading?
This project will not let a reader quote from a page whose text scores below
0.955 on its word-recognition gate — that is how 620
pages were routed away from the text readers. Applying the same gate to the
sentences that license the rules: 21 of 240
rule records are licensed by a quote that fails it, and those rules produce
745 of the 2489 claims below whose source
sentence can be scored at all — and 188 of the
553 graded claims in the record above.
The worst is nirenstein.sensitive_numbers at
0.680, which alone carries 164 grades.
Its licensing sentence reads, in full:
'If968 is thc high thcn pcrhaps, Thereforg Lhey &rc equal the sam: vibralioma] state summationally: These jra alsd sensitive numbers;aS thc SLIm is Uhc Smc; 896 689 698 869" I568 is Ihc Jow Ihen; perhaps, 658'. 586: 856 685 865
A damaged quote is not the same as a wrong rule — “The 10-year
oyole oquals 120 months” is unambiguous once you see it. But a rule
marked verbatim against a text like this is doing something weaker than
the word promises, and the label has no field to say so. It does now.
How much of this could be graded at all
Of 23,823 verified passages, 428 are conditional level
statements — a trigger and a target, stated forward
(“crossing 56 will indicate 65”). 338 of those name an instrument.
375 give the target as a figure; 53 give it only as a
direction — “much higher”, “lower prices”
— and are not falsifiable as printed, however confidently they read.
The rest of the corpus is not untested because we ran out of time. It is not
gradeable in principle: doctrine has nothing to check, and a procedure whose
operative number the author withheld cannot be run by anyone. That is the
honest denominator for every other number on this page.
None of the 428 is graded here. A regex can find a sentence that says
what will happen; it cannot tell whether the sentence was written before the
event or after it in a teaching example, and every row is held pending that
ruling. Calling them predictions first would repeat the mistake this page
already made once, when 3524 rows were read as 3524 trials.
The verified corpus
Every entry carries the author's exact words and the page they
were read from. A quote that could not be found on its page,
character‑for‑character after Unicode normalisation, was discarded and the
method with it. Open a book to load its passages.
10,724 of these passages state a procedure — and
7,159 of those name a market instrument or a timeframe.
The gap is not padding. Baumring's Gann Harmony volumes are bound
anthologies: he reprinted the acoustics, geometry and number theory he taught
Gann's “law of vibration” from, alongside his own seminar
material. A procedure for suppressing chosen partials on a vibrating string is
a procedure, and it is correctly recorded as one — it is not a trading
rule. Volume 1 has 60 procedures and 6 that name an instrument; Gann's
Master Stock Market Course has 653 and 629. The split below is the
readers' own recorded judgement, not a classifier: they were told to fill
those two fields only where the page states or demonstrates them, and to leave
them blank rather than guess.