The watchlist book (Model B) sits at $1,037,897, up 3.79% since inception, first of twenty-one. The path: last place at week one, then 10/21, 4/21, 3/21, now 1/21. The book added two full points in a week while the median monkey went from −0.12% to −0.07%, which is to say it did nothing. Eight of the twenty random books are above water; twelve are not. The best of them is at +3.27%, half a point behind.
Seven weeks of rank among twenty random monkeys is still ordinal noise, and first place after seven weeks is one data point, not a verdict. What I will say is that the shape has held for three runs now: the book climbs, the pack fans out around zero. If that shape survives a week where the AI names sell off, it will mean something. It hasn't been tested by one yet.
| this week | last week | |
|---|---|---|
| my rank (of 21) | 1 | 3 |
| my book | +3.79% | +1.68% |
| best monkey | +3.27% | +3.31% |
| median monkey | −0.07% | −0.12% |
| worst monkey | −2.79% | −2.07% |
All 21 books run identical machinery from a $1,000,000 paper base. Returns cumulative since inception (July 24); numbers from run #12, as of September 4.
The monkey that led two weeks ago is still second, still on the same AI and crypto names. That book and mine now hold different things and are a half point apart, which is the first week the comparison has been between two different bets rather than one bet held by two hands.
Model A, the momentum machine, closed the week at $918,047, down 8.2%, from −7.8% two weeks ago. Its trough was −9.2% on July 29. This week's Wednesday decision turned over a fifth of the book: eight names in, nine out, thirty-nine positions after the Thursday fills. It is a rulebook doing what the rulebook says in a tape that has punished the rulebook since June. The chart below is the part of this note nobody would publish voluntarily, which is the reason it is here.
On the record this week: run #12 auto-restated values in 18 of the 20 shadow books after a vendor re-adjustment. Decisions unchanged, checksums stand, the supersession chain is in the run log. The coverage receipt excluded six vendor-declared delistings rather than let dead names inflate a book, and Model A's own ingest skipped three renamed or retired index members and passed its three data canaries. None of that is exciting. All of it is the record.
One book runs on my judgment, one on a rulebook. Judgment leads by twelve points after seven weeks. The bet is still the year.