THE LEDGER
every claim the study tested — one row each, with the arithmetic that earned its word
This table is the study’s bottom line. Each row is one claim about how a working succeeds; each verdict is a computation that ran, defined in the legend below the table; each row links to its full accounting. The evidence is 529 workings across 10 tradition families whose base success-rates run 23%–69% — that spread is the confound every test adjusts for, so no rate here ever reads across traditions, and the two record sets (150 deep-read home records, re-coded clean; 379 fresh records the study had never seen, 329 of the 529 read with the ending cut away) are never pooled into one bar. Browse every record on THE RECORDS; the blinding and the reader-accuracy license are on THE METHOD.
| the lever | verdict | the contrastarms · Δ · home/fresh lanes | travels?fresh-records Δ | how surereader agreement | the one line |
| the preparation · before you begin — what you bring, and how whole it is | |||||
| one aim only | flat | +7 | not separately scored | One clear aim beats several by 8 points, held within traditions — inside noise. Bring one aim as craft; the record does not demand it. | |
| the wanting (eros) | thin | — | readers often split | Want it for real. The empty rites have still never succeeded — but only 3 readable empty rites survive the blind re-read, so the word waits on more records. | |
| the empty rite | thin | — | readers often split | Nobody works without wanting: so few empty rites exist that the floor rests on 3 readable records (the frozen home view ran 0 of 19). | |
| the undivided will | likely | +46 | not separately scored | Bring your whole self — united wanting ran 40+ points ahead of divided in every view. The numbers are proven-shaped; the question is young — pre-registration and one fresh run make the word. | |
| by the book | retired | — | — | not separately scored | Following the script exactly bought nothing at home, measured twice. The call was cut at the freeze; a repair ticket stands (the recipe carries you to the state; the state does the work). |
| the rite · inside the working — the state, the signs, the release | |||||
| the click | proven at home | +10 | readers agreed | Trust something answering back in the rooms built to summon one — séance, evocation, remote viewing. On the wider records the gap shrinks to +10, inside noise: a local lever. | |
| the ripening | proven | +42 | not separately scored | Fire full, not early — releasing an unripe working failed in both record sets, within tradition. | |
| the firing | unreadable | — | — | readers disagreed | No tradition teaches that the operator can watch the arrow in flight, and neither could our readers: two blind readers agreed barely above chance (α 0.37). The demote rule fired; THE CONCEPTION is its repair. |
| living it as done | thin | +14 | not separately scored | Perform the wish as already true. The posture leans right everywhere, but which idiom you use tracks your tradition, and the just-asked arm is 26 records. | |
| a road for the result | retired | — | — | not separately scored | Naming the result’s road was cut at the freeze — the call kept measuring which tradition wrote the record, not the working. A repair ticket stands. |
| the aftermath · after you fire — what you do with your hands | |||||
| the cord | proven | +46 | readers often split | Fire it and let it go. Clutching afterward — the old warning against lust of result — killed workings in both record sets, within tradition. | |
| what you link with | retired | — | — | not separately scored | What the link is made of never decided holding — measured flat at home and cut at the freeze. What you link with matters as craft; the record scores only that you linked at all. |
| the levers together · the levers, stacked — the closing question | |||||
| the levers together | thin | — | — | not separately scored | Do several right things, and avoid the faults — the levers co-fire rather than trade off. Signs read good climb 0% → 15% → 51% → 68%; faults stack down 58% → 49% → 32% → 14%, and every clean-read working carrying three faults failed. |
the preparation
before you begin — what you bring, and how whole it is
one aim only
flatDoes a working with one clear aim beat one with several, or a drifting one?
One clear aim beats several by 8 points, held within traditions — inside noise. Bring one aim as craft; the record does not demand it.
Counted strictly, the tradition-held gap is Δ +8; set the half-worked records aside and it is Δ +3; count them as wins, Δ +0 — the claim RIDES on the rule.
show the accounting
| every record, all traditions | n |
| the whole corpus | 529 |
| this call not readable on the page | 16 |
| readable, but no succeeded–failed verdict | 68 |
| read on this card | 445 |
| succeeded | failed | rate | 95% range | |
| one clear aim | 202 | 167 | 55% | 50–60% |
| several aims, or drifting | 37 | 39 | 49% | 38–60% |
• held within tradition families (the study-grade read): Δ 8 pts across 10 families
• the home records, under this fresh read: one clear aim 40% (n=100) vs several aims, or drifting 28% (n=18); the fresh records: one clear aim 60% (n=269) vs several aims, or drifting 55% (n=58) — the two sets are never pooled into one bar; the family-held Δ above averages within, never across.
browse the 513 records this card reads, each with its call, on THE RECORDS →
the wanting (eros)
thinDoes wanting it badly beat not wanting it at all — and how much wanting is enough?
Want it for real. The empty rites have still never succeeded — but only 3 readable empty rites survive the blind re-read, so the word waits on more records.
The floor (no charge → nothing succeeds) is the firmest thing we have. Don’t read the ‘hot road looks weak’ gap as real — it is almost entirely one failed diary (Crowley’s Rex de Arte Regia); the Paris Working, just as hot, held. And a reliability caveat earned the hard way: on the wider records, two independent blind readers often called the charge differently (they agreed on about a third of the calls they both made). The aggregate gap held under the second reader; the per-record call is soft. One more honest check (2026-07-03): we let three readers infer the wanting on 40 of the unread records from stakes, cost, and persistence — they recovered labels for half. Too few records to say whether a success-gradient travels with the inferred labels (the arms were 6 to 11 records; only an enormous gap could have shown). What we can say: the lever is PROVEN only where the wanting is written; the unread two-thirds are unread, not cleared. Counted strictly, the tradition-held gap is Δ +31; set the half-worked records aside and it is Δ +49; count them as wins, Δ +61 — the claim does not ride on the rule.
show the accounting
| every record, all traditions | n |
| the whole corpus | 529 |
| this call not readable on the page | 320 |
| readable, but no succeeded–failed verdict | 21 |
| read on this card | 188 |
| succeeded | failed | rate | 95% range | |
| strong | 31 | 21 | 60% | 46–72% |
| some | 80 | 53 | 60% | 52–68% |
| none — the empty rite | 0 | 3 | 0% | 0–56% |
• held within tradition families (the study-grade read): Δ 31 pts across 3 families
• the home records, under this fresh read: strong 39% (n=23) vs none — the empty rite 0% (n=1); the fresh records: strong 76% (n=29) vs none — the empty rite 0% (n=2) — the two sets are never pooled into one bar; the family-held Δ above averages within, never across.
browse the 209 records this card reads, each with its call, on THE RECORDS →
“Store up energy until it does the work itself.”Crowley, John St. John (1908)
| record accounting | n |
| in this card’s analysis | 67 |
| set aside for this card (n/a or not read here) | 52 |
| active corpus | 119 ✓ |
| level | succeeded | failed | success-rate | 95% CI |
| strong | 12 | 11 | 52% | 33–71% |
| some | 12 | 21 | 36% | 22–53% |
| none — the empty rite | 0 | 11 | 0% | 0–26% |
the empty rite
thinDoes a working nobody wanted — rote, dutiful, a test run — ever succeed?
Nobody works without wanting: so few empty rites exist that the floor rests on 3 readable records (the frozen home view ran 0 of 19).
The floor is the study’s oldest finding and its thinnest arm at once: every readable empty rite failed, in every era of the data, and there have never been enough of them to say so with force. A no-wanting sourcing draw is the standing fix. Counted strictly, the tradition-held gap is Δ +51; set the half-worked records aside and it is Δ +65; count them as wins, Δ +73 — the claim does not ride on the rule.
show the accounting
| every record, all traditions | n |
| the whole corpus | 529 |
| this call not readable on the page | 320 |
| readable, but no succeeded–failed verdict | 21 |
| read on this card | 188 |
| succeeded | failed | rate | 95% range | |
| any visible wanting | 111 | 74 | 60% | 53–67% |
| none — the empty rite | 0 | 3 | 0% | 0–56% |
• held within tradition families (the study-grade read): Δ 51 pts across 3 families
• the home records, under this fresh read: any visible wanting 41% (n=73) vs none — the empty rite 0% (n=1); the fresh records: any visible wanting 72% (n=112) vs none — the empty rite 0% (n=2) — the two sets are never pooled into one bar; the family-held Δ above averages within, never across.
browse the 209 records this card reads, each with its call, on THE RECORDS →
the undivided will
likelyIs the whole operator behind the working, or is a part holding back?
Bring your whole self — united wanting ran 40+ points ahead of divided in every view. The numbers are proven-shaped; the question is young — pre-registration and one fresh run make the word.
Re-defined as undividedness it IS scoreable and escapes the old distress trap (distressed succeeded MORE). The wider 379 lean the same way: united 78% vs divided 31%, tradition-held Δ +48 (+20 on the fully-blind reads, a 5-record divided arm). Exploratory still — the divided arm stays thin everywhere we can read it. Counted strictly, the tradition-held gap is Δ +44; set the half-worked records aside and it is Δ +48; count them as wins, Δ +33 — the claim does not ride on the rule.
show the accounting
| every record, all traditions | n |
| the whole corpus | 529 |
| this call not readable on the page | 385 |
| readable, but no succeeded–failed verdict | 14 |
| read on this card | 130 |
| succeeded | failed | rate | 95% range | |
| united | 64 | 30 | 68% | 58–77% |
| divided | 9 | 27 | 25% | 14–41% |
• held within tradition families (the study-grade read): Δ 44 pts across 6 families
• the home records, under this fresh read: united 53% (n=34) vs divided 20% (n=20); the fresh records: united 77% (n=60) vs divided 31% (n=16) — the two sets are never pooled into one bar; the family-held Δ above averages within, never across.
browse the 144 records this card reads, each with its call, on THE RECORDS →
| record accounting | n |
| in this card’s analysis | 79 |
| set aside for this card (n/a or not read here) | 40 |
| active corpus | 119 ✓ |
| level | succeeded | failed | success-rate | 95% CI |
| united | 33 | 39 | 46% | 35–57% |
| divided | 2 | 5 | 29% | 8–64% |
• among the not-clutching: united 45% (29/64) vs divided 40% (2/5)
• by tradition, the “united” success-rate ranges 21%–76% — it tracks which records you draw from, not just the call
by the book
retiredDoes following the script exactly — the right hours, names, materials — predict success?
Following the script exactly bought nothing at home, measured twice. The call was cut at the freeze; a repair ticket stands (the recipe carries you to the state; the state does the work).
the rite
inside the working — the state, the signs, the release
the click
proven at homeWhen something answers back mid-working, does the working go on to succeed?
Trust something answering back in the rooms built to summon one — séance, evocation, remote viewing. On the wider records the gap shrinks to +10, inside noise: a local lever.
The strongest single lever in the study, and the one confirmed leg on never-seen records. It fires where contact is the mechanism — calling and waiting, evocation, remote viewing — and goes flat where there’s no one to answer: ‘did the channel open’, not a universal law. Counted strictly, the tradition-held gap is Δ +27; set the half-worked records aside and it is Δ +24; count them as wins, Δ +17 — the claim does not ride on the rule.
show the accounting
| every record, all traditions | n |
| the whole corpus | 529 |
| this call not readable on the page | 69 |
| readable, but no succeeded–failed verdict | 62 |
| read on this card | 398 |
| succeeded | failed | rate | 95% range | |
| something answered | 107 | 72 | 60% | 52–67% |
| nothing came | 117 | 102 | 53% | 47–60% |
• held within tradition families (the study-grade read): Δ 27 pts across 9 families
• the home records, under this fresh read: something answered 57% (n=56) vs nothing came 22% (n=65); the fresh records: something answered 61% (n=123) vs nothing came 67% (n=154) — the two sets are never pooled into one bar; the family-held Δ above averages within, never across.
browse the 460 records this card reads, each with its call, on THE RECORDS →
The admissible mid-working signs are a closed list (the reason readers agree on this call at α .90–1.00): a presence, a voice, an image, a body sign, an instrument moving — arriving DURING the working. The ending never counts as its own sign.
| record accounting | n |
| in this card’s analysis | 91 |
| set aside for this card (n/a or not read here) | 28 |
| active corpus | 119 ✓ |
| level | succeeded | failed | success-rate | 95% CI |
| something answered | 21 | 24 | 47% | 33–61% |
| nothing came | 10 | 36 | 22% | 12–36% |
• by tradition, the “answered-mid-working” success-rate ranges 35%–75% — it tracks which records you draw from, not just the call
the ripening
provenDoes a working fired full — matured, complete — beat one released early?
Fire full, not early — releasing an unripe working failed in both record sets, within tradition.
One honest entanglement: the felt crest is also the top of the wanting’s scale, so part of this gap is the charge wearing a second label. The kill-test that split them is on THE METHOD; the behavioral leg (the rite reached its scripted end) is the part that stands apart. Counted strictly, the tradition-held gap is Δ +47; set the half-worked records aside and it is Δ +63; count them as wins, Δ +42 — the claim does not ride on the rule.
show the accounting
| every record, all traditions | n |
| the whole corpus | 529 |
| this call not readable on the page | 334 |
| readable, but no succeeded–failed verdict | 22 |
| read on this card | 173 |
| succeeded | failed | rate | 95% range | |
| fired full | 82 | 49 | 63% | 54–70% |
| fired early | 10 | 32 | 24% | 13–39% |
• held within tradition families (the study-grade read): Δ 47 pts across 8 families
• the home records, under this fresh read: fired full 46% (n=48) vs fired early 0% (n=15); the fresh records: fired full 72% (n=83) vs fired early 37% (n=27) — the two sets are never pooled into one bar; the family-held Δ above averages within, never across.
browse the 195 records this card reads, each with its call, on THE RECORDS →
the firing
unreadableDid the shot land — can a reader see the working arrive?
No tradition teaches that the operator can watch the arrow in flight, and neither could our readers: two blind readers agreed barely above chance (α 0.37). The demote rule fired; THE CONCEPTION is its repair.
Half right: most of the raw gap is the charge, seen one station later (see the kill-test memo).
show the accounting
| record accounting | n |
| in this card’s analysis | 81 |
| set aside for this card (n/a or not read here) | 38 |
| active corpus | 119 ✓ |
| level | succeeded | failed | success-rate | 95% CI |
| a felt peak | 15 | 16 | 48% | 32–65% |
| flat / dutiful | 11 | 39 | 22% | 13–35% |
living it as done
thinIs the wish performed as already true — become, staged, received — or only asked for?
Perform the wish as already true. The posture leans right everywhere, but which idiom you use tracks your tradition, and the just-asked arm is 26 records.
Which idiom is used tracks tradition, not outcome. The old durable piece — bare petition fails — softened on the wider corpus: petition succeeded 10 of 21 raw, 2 of 5 read fully blind; held against tradition the leftover gap is small (Δ +8) — and the just-asked arm is 26 records, so only a large push could have shown. Not proven a lever; not cleared either. Counted strictly, the tradition-held gap is Δ +22; set the half-worked records aside and it is Δ +34; count them as wins, Δ +33 — the claim does not ride on the rule.
show the accounting
| every record, all traditions | n |
| the whole corpus | 529 |
| this call not readable on the page | 18 |
| readable, but no succeeded–failed verdict | 67 |
| read on this card | 444 |
| succeeded | failed | rate | 95% range | |
| lived as done (any idiom) | 224 | 194 | 54% | 49–58% |
| just asked | 10 | 16 | 38% | 22–57% |
• held within tradition families (the study-grade read): Δ 22 pts across 5 families
• the home records, under this fresh read: lived as done (any idiom) 38% (n=116) vs just asked 0% (n=4); the fresh records: lived as done (any idiom) 60% (n=302) vs just asked 45% (n=22) — the two sets are never pooled into one bar; the family-held Δ above averages within, never across.
browse the 511 records this card reads, each with its call, on THE RECORDS →
“though I could feel her mouth sucking up mine, I could simultaneously see the gold filling the room”Crowley, Rex de Arte Regia, Opus XXXIII (11 Feb 1915)
| record accounting | n |
| in this card’s analysis | 96 |
| set aside for this card (n/a or not read here) | 23 |
| active corpus | 119 ✓ |
| level | succeeded | failed | success-rate | 95% CI |
| enacted (any idiom) | 34 | 57 | 37% | 28–48% |
| merely asked | 0 | 5 | 0% | 0–43% |
• by tradition, the “enacted” success-rate ranges 0%–80% — it tracks which records you draw from, not just the call
a road for the result
retiredDoes naming how the result will arrive help it arrive?
Naming the result’s road was cut at the freeze — the call kept measuring which tradition wrote the record, not the working. A repair ticket stands.
the aftermath
after you fire — what you do with your hands
the cord
provenAfter firing, does letting the working go beat clutching it?
Fire it and let it go. Clutching afterward — the old warning against lust of result — killed workings in both record sets, within tradition.
Powerful in the one diary deep enough to show it; thinner on rougher records. And ‘do the working, then stop caring’ barely beats clutching — real release needs something to release: you have to have cared first. The old third reading — ‘never cared’ — is retired from this card: the readers used it as a default for silence (of 253 such calls, only 3 co-occur with no-wanting), so it measured nothing; the empty rite is read from the wanting’s floor instead. Reliability caveat: on the wider records two independent blind readers often called the cord differently (agreement about one in four where both committed). The direction held — and strengthened — under the second reader; the per-record call is soft. Counted strictly, the tradition-held gap is Δ +40; set the half-worked records aside and it is Δ +41; count them as wins, Δ +31 — the claim does not ride on the rule.
show the accounting
| every record, all traditions | n |
| the whole corpus | 529 |
| this call not readable on the page | 400 |
| readable, but no succeeded–failed verdict | 10 |
| read on this card | 119 |
| succeeded | failed | rate | 95% range | |
| let it go | 41 | 11 | 79% | 66–88% |
| clutched | 25 | 42 | 37% | 27–49% |
• held within tradition families (the study-grade read): Δ 40 pts across 8 families
• the home records, under this fresh read: let it go 50% (n=8) vs clutched 32% (n=22); the fresh records: let it go 84% (n=44) vs clutched 40% (n=45) — the two sets are never pooled into one bar; the family-held Δ above averages within, never across.
browse the 129 records this card reads, each with its call, on THE RECORDS →
| record accounting | n |
| in this card’s analysis | 65 |
| set aside for this card (n/a or not read here) | 54 |
| active corpus | 119 ✓ |
| level | succeeded | failed | success-rate | 95% CI |
| let go | 12 | 10 | 55% | 35–73% |
| clutched | 5 | 11 | 31% | 14–56% |
| never cared (frozen view only) | 4 | 23 | 15% | 6–32% |
• by tradition, the “released” success-rate ranges 50%–67% — it tracks which records you draw from, not just the call
what you link with
retiredDoes the link’s material — hair, photograph, name, nothing — decide holding?
What the link is made of never decided holding — measured flat at home and cut at the freeze. What you link with matters as craft; the record scores only that you linked at all.
the levers together
the levers, stacked — the closing question
the levers together
thinDo the levers compound — or do four good signs just re-measure one good working?
And the faults, counted the same way — each one written ink, never silence (no charge · clutching · released unripe · a divided will · a muddy aim; a working that got no sign is NOT at fault — that is the quiet road):
Set aside, honestly: 268 deed workings (road-less — a trick laid, a charm placed) at 56%, whose good signs are exterior and wait on their own call; and 32 state workings too thinly written to place (fewer than two signs readable), at 28%.
Do several right things, and avoid the faults — the levers co-fire rather than trade off. Signs read good climb 0% → 15% → 51% → 68%; faults stack down 58% → 49% → 32% → 14%, and every clean-read working carrying three faults failed.
This ladder was restacked (2026-07-03): the four signs are interior, so they are counted only on state workings; the click counts only in rooms built to summon a sign; and only signs actually READ count — silence is not zero. The old ladder pooled all of that and its rungs wobbled. The fault ladder’s zero rung is a mixture (no fault VISIBLE), but every fault counted is written ink. Its two three-fault survivors are told honestly: one is a pair of adepts deliberately breaking the sigil procedure as an experiment — the faults WERE the working — and one was read with the ending leaked; the twelve clean, unintentional three-fault workings all failed. Deed workings — near half the graded corpus — wait on an exterior call of their own. Both ladders are descriptive, record sets pooled and labeled as such; the levers’ individual claims stay in their own cards.
show the accounting
Split by record set the restacked staircase holds its shape — the home records run 0% → 0% → 41% → 41%, the fresh records run 0% → 25% → 57% → 84% (none → some → most → all read signs good), and the two sets are never pooled into one claim. The levers are not redundant: two levers rarely travel together, and the click and the cord even pull apart (a working that gets an answer tends to grip; one that lets go tends to get no answer) — so stacking is a real question, not one lever measured four times. The home-tradition read alone says Δ 16.
The honest limit: only 91 workings carry all four levers clearly read at once, so the per-lever increments sit in cells of n=2–6. A richer-diary sourcing pass is the standing fix.
Every rate on this page counts a half-worked record as failed — the strict rule, set before scoring; each card above shows its claim both ways. The house words (eros, gnosis, the click, the cord, the close) are defined on THE METHOD.