The signal is real. It is not tradeable. (and the six silent data faults that hid both) by caganco · Pull Request #8 · caganco/trailingedge

scrapers: a DKB filing that parses to zero transactions is a failure, not an empty filing

caganco

committed

f4e2155

scrapers: pre-2021 filings parsed to zero because the date regex only accepted slashes

caganco

committed

ee25c9e

scrapers: correct the parse floor to 2015-01 - the DUY diagnosis was a confound

caganco

committed

00305ac

scrapers: validate every row against the form's own arithmetic; parse the 2015-2020 blotter

caganco

committed

ba363bd

scrapers: let the checks choose the totals binding; parse prices era-aware

caganco

committed

0b895de

scrapers: quarantine unparsed DKB PDFs instead of discarding the evidence

caganco

committed

fcf0630

scrapers: two more evidence-backed blotter emissions + offline quarantine repair

caganco

committed

705e53a

scrapers: accept a legacy filing only when two independent sources agree to the kurus

caganco

committed

1cc941b

scrapers: stop retrying the WAF inline; recover the disclosures it drops

caganco

committed

625bea0

scrapers: escalate the WAF recovery, batch the existence check, fetch fresh months first

caganco

committed

23686eb

signals: a trade cannot predate its own filing; a third of the clusters cannot vanish

caganco

committed

1041b2d

data: fetch prices per ticker, not per batch - a third of the "survivorship" was a bug

caganco

committed

2fdfe84

data: load the exchange's own bulletin - survivorship bias 31% -> 1.5%, and the 60-day sign flips

caganco

committed

e9c1d7f

scrapers: stop sleeping inside every month; sweep fast, recover in passes

caganco

committed

80bbe41

signals: the edge is real and it is not tradeable - the spread eats it

caganco

committed

1799071

data: load the VBTS tradability flags - and they turn out not to be the constraint

caganco

committed

ffe1f07

docs: the README states the finding, not the aspiration

caganco

committed

4c55467

stage-0: freeze the routine/opportunistic definition before measuring it

caganco

committed

dacf3d2

signals: the routine/opportunistic split does not carry to BIST - there are no routine filers

caganco

committed

f7f0f3e

docs: METHODOLOGY listed as open three gaps that are now closed

caganco

committed

bb62026

scrapers: a chunk gets a deadline it cannot exceed

caganco

committed

babf6bb

scrapers: the three-hour hang was a timezone, not a hang - correcting the record

caganco

committed

11f3617

Navigation Menu

Repository navigation

The signal is real. It is not tradeable. (and the six silent data faults that hid both) - #8

The signal is real. It is not tradeable. (and the six silent data faults that hid both)#8
caganco merged 29 commits into
masterfrom
fix/duy-era-silent-parse-failure

Commits on Jul 12, 2026

scrapers: a DKB filing that parses to zero transactions is a failure, not an empty filing

scrapers: pre-2021 filings parsed to zero because the date regex only accepted slashes

scrapers: correct the parse floor to 2015-01 - the DUY diagnosis was a confound

scrapers: validate every row against the form's own arithmetic; parse the 2015-2020 blotter

scrapers: let the checks choose the totals binding; parse prices era-aware

scrapers: quarantine unparsed DKB PDFs instead of discarding the evidence

scrapers: two more evidence-backed blotter emissions + offline quarantine repair

scrapers: accept a legacy filing only when two independent sources agree to the kurus

scrapers: stop retrying the WAF inline; recover the disclosures it drops

scrapers: escalate the WAF recovery, batch the existence check, fetch fresh months first

signals: a trade cannot predate its own filing; a third of the clusters cannot vanish

data: fetch prices per ticker, not per batch - a third of the "survivorship" was a bug

data: load the exchange's own bulletin - survivorship bias 31% -> 1.5%, and the 60-day sign flips

scrapers: stop sleeping inside every month; sweep fast, recover in passes

signals: the edge is real and it is not tradeable - the spread eats it

data: load the VBTS tradability flags - and they turn out not to be the constraint

docs: the README states the finding, not the aspiration

stage-0: freeze the routine/opportunistic definition before measuring it

signals: the routine/opportunistic split does not carry to BIST - there are no routine filers

docs: METHODOLOGY listed as open three gaps that are now closed

scrapers: a chunk gets a deadline it cannot exceed

scrapers: the three-hour hang was a timezone, not a hang - correcting the record

Commits on Jul 13, 2026

tests: cover the two modules the headline result actually rests on

docs: split-sample stability - the conclusion holds in both halves

cli: --help crashed on a Turkish console, and the README documented a command that does not exist

fix: the cost model was reading a return index as if it were a price

fix: make the exchange bulletin authoritative over yfinance in price_history

report: emit the net-of-cost verdict alongside the gross one

chore: mypy clean, and cover the two inserts that the type fix touched