Lab notes · week 43 of 2026 · this week, rewritten every night
Lab notes: the week of Oct 19 – 25, 2026
The lab has run 151,994 tests over 1 night so far this week, the latest on Oct 19. 412 of them were ideas it had never tested before. 229 findings held after the latest night: proven coin drivers and series, surviving trading rules and kinds of event that move the odds. 1 was proven for the first time. 1 that had been proven stopped holding. Past patterns, not financial advice.
Tests run
151,994
over 1 night
New ideas
412
tested for the first time
Holding
229
after the latest night
First proven
1
this week
Stopped holding
1
proven before, not now
§ 01 · Night by night1 night
Night by night
as of · the lab's ledger, kept by each stage itself
tests run that night, the latest litblue under a day: findings holding after it· no run that night
| Night | Coin drivers | Series | Trading rules | Odds movers | Scoreboard | All tests | New ideas | Stopped holding |
|---|---|---|---|---|---|---|---|---|
| 142,120 213 held | 5,983 1 held | 3,628 10 survived | 143 5 held | 120 14 held | 151,994 | 412 | 1 |
So far, the ledger held 2,006,462 tests over 14 nights since Oct 6, 2026, 154,195 different ideas among them.
Source: the lab's ledger, every idea each stage tested, night by night · stored
§ 02 · First proven1 new
First proven this week
as of · coin drivers and whole-market findings
1 finding passed the bar for the first time this week: proven at a night's end and not proven when the week began.
-
PROVEN
XMR
When its sector rose 5% on average in a day, XMR ended lower the next day more often than on any day
183 cases · no clear edge out of sample · first proven · its own pageShare on Xcopy link
§ 03 · Stopped holding1 stopped
Stopped holding
as of · proven when the week began, not past the bar now
1 finding that had been proven no longer passed the bar after the latest night. A finding that stops holding stays in the record; it no longer counts toward a coin's odds.
- STOPPED TRA TRA, the day after the 100 biggest altcoins fell 4% on average in a day proven when the week began, not past the bar now · what moves TRA
§ 04 · Fired this week0 fired
Proven drivers that fired
as of · the event happened this week, so its window opened
No proven driver fired this week.
§ 05 · The trading rules
The trading rules
as of · the nightly strategy search and its book
On its latest night of the week (Oct 19, 2026), the strategy search tested 3,628 versions of its trading rules; 10 passed both cuts.
10 rules in the book still pass; 0 entered it this week and 0 faded. Every rule that ever passed stays in the book with its paper record.
What the search keeps, and the closest misses it threw away →
§ 06 · How to read them
How to read these notes
as of · the bars each lab holds a finding to
A test is one idea checked against its history on one night: a coin after one kind of event over one horizon, a collected series against Bitcoin, Ether, the altcoins or its own coin, a whole trading rule, a kind of event moving the odds or against ordinary days.
A luck test (t) says how far a result sits from zero, in units of its own noise: about 2 is rarely luck for one test, but thousands tested at once need far more, which is what the cuts below are for.
A coin driver or a series finding is proven when it passes a 10% false-discovery cut over every test of its night, rests on 12 cases or more and holds in both halves of its history. Likely ones pass a looser bar, at about the rate pure chance would give: leads to watch, not facts. A trading rule must pass a 5% false-discovery cut over every test at once on the older 60% of the history, then a luck test of 1.645 or more on the same side on the newer 40% it was not chosen on, every case counted.
This week is still running: its numbers are rewritten every night and stored for the last time the first night after Oct 25, 2026.
§ 07 · More
More research
as of