Many funds have rules about what they may hold: no stock below a certain price, no bond below investment grade, nothing outside the index. When a security crosses that line, the thesis says, the fund has to sell on that day, whatever the security is worth. That would create a seller with no view on value in small, unloved names, the kind of seller a retail buyer can stand against.
forced flow / fund mandate
Mandate-constrained forced sale
The idea
How it was tested
No backtest. The entry test went to the source: fund prospectuses filed on EDGAR from 2022-2026, searched for the clauses that would bind a fund to sell. Two triggers were checked. For the sub-$5 price rule, three phrasings were tried. For rating downgrades below investment grade, the clauses found were read and classified by whether they bind the fund or leave the decision to the adviser.
Where it died
Rejected on 2026-09-20. The sub-$5 rule returned zero filings under all three phrasings, while the same search found 1,452 filings for the rating wording, so the absence was real and not a broken query. Of 77 downgrade clauses, 2 bind the fund to sell. The rest give the adviser discretion in writing: the fund "will consider such an event" when deciding whether to keep the security. A fund that decides case by case is not a forced seller. The one binding channel, the index rulebook, belongs to a separate mechanism with its own test.
The instrument almost reached the wrong answer first. A text classifier labelled 72 of the 77 clauses "undetermined", because it looked only for "may continue to hold" and missed the most common construction. Reading a sample of each class by hand caught it.
The record
Thesis
Funds barred by their prospectus from holding a security once it crosses a line — a sub-$5 price, a rating downgrade, an index deletion — must sell on that day regardless of value, creating a price-insensitive seller in small, unloved names.
Gates
Entry test — obligationfail
Is someone actually obliged to trade badly here, and unable to stop?
Why it died
The prospectuses were read, and the obligation is not in them. Across fund prospectuses filed 2022-2026, the "sub-$5 price" trigger returns zero filings under three phrasings, while the same search returns 1,452 filings for the rating phrasing — a real absence, not a broken query. Of 77 clauses on a downgrade below investment grade, 2 bind the fund to sell; the rest give the adviser discretion in writing, e.g. "will consider such an event in determining whether the Fund should continue to hold the security", or "may sell the security or continue to hold". A fund that decides case by case is not a forced seller. The one channel that does bind — the index rulebook — is a separate mechanism tested on its own.
What it taught
The instrument nearly got this wrong. A first-pass text classifier labelled 72 of the 77 clauses "undetermined" because it only looked for "may continue to hold" and missed the dominant construction ("will consider ... in determining whether"). Reading a sample of each class by hand, not the aggregate count, is what caught it. And the cheap question that predicts this failure: the obliged party wrote its own prospectus, so it kept itself an exit.
Verdict history
- 2026-09-20Rejected at the entry test after reading fund prospectuses on EDGAR — no binding sell clause Rejected before testing
- 2026-09-30Counted among five rejections with the same shape (obliged in aggregate, discretionary in the specific action), which produced the 'who wrote the rule?' pre-filter