LZLZL/Prediction markets/Kill file
FREEJUDGE IT C · ArbitrageKill file

TWAP tail lock
0 of 4 tenors

2026-08-21 · the information was real; the cost assumption was not

This one is worth studying because of how it died: the information it bet on genuinely exists — I verified that. It was closed as a whole family anyway, because the cost side did not hold: the market had already priced that information in, exactly.

1What it was betting on

Many crypto boards settle not on the closing instant but on an average of the final stretch (a time-weighted average). Which means: the closer you get to settlement, the faster uncertainty collapses — more of the average is already locked, and less time remains to change the answer.

near settlement the outcome is close to mathematically determined so is buying the side that must win, at the quoted price, a sure thing?

The reasoning is not wrong. The information is real. The failure is in the next step.

2Verified: the information does exist

On the shortest tenor, using my own recorded books, buying the locked side at the price actually available at qualifying moments:

MeasureResult
Hit rate of the executable portion90.2%
Average fill price of that portion0.9033
★ Those two lines are the entire answer

Hit rate 90.2%, fill price 0.9033 — essentially equal.

Which is to say: the information exists, and the market already knows all of it, and has priced it precisely. What you buy is a fair price; there is no excess to collect, and the fee makes it a net loss.

I named the family's cause of death "the book is already fair" — not that the information was false, but that it is already in the price.

3And most of the time you cannot buy anyway

MeasuredFigure
Qualifying moments141
Of which actually executable51
Moments with no offer at all on the locked side64%

Equally fatal: at precisely the moments you most want to act, nobody will sell to you. For an obvious reason — the other side also knows that leg is going to win, so why would they hand it over at 0.90?

A textbook case of paper opportunity versus executable opportunity: a backtest filling at quoted prices invents a great many trades that were never available.

4The net, and the verdict

ItemResult
EV of the executable portion, less fees−0.47pp
Against the null modellost to it
First half versus second halfopposite signs
Across four tenors0 of 4 passed

Three independent failure signals stacked: negative net, lost to the null, halves disagreeing. Any one is sufficient; all three together leaves nothing to discuss. And it is not one tenor's problem — all four failed, so the family closed.

5★ The lesson worth keeping

Between "the information is real" and "it makes money" sit three gates

① Does the market already know? — here, entirely: hit rate and price align exactly.

② Can you get filled? — 64% of the time the locked side has no offer.

③ What is left after costs? — −0.47pp.

Most strategy ideas die at the first gate, and their authors typically verified only "the information is real" and thought they were done. "I found a genuine pattern" is a very low bar — it rules out "I am imagining things" and nothing else.

Testing gate one is cheap: compare the advantage you found against the price at the time. If the price already reflects it — as 90.2% against 0.9033 does — stop there and save the entire engineering effort.

6One line to keep

Hit rate 90.2%, fill price 0.9033. Those two numbers being equal is the whole verdict. Finding a real pattern and finding one others do not know about are different things — and only the second pays.

EvidenceCheck it yourself

Information verified on the shortest tenor, the executable portion at qualifying moments had a hit rate of 90.2% against an average fill of 0.9033 — essentially equal, i.e. fully priced.
Executability 141 qualifying moments, of which 51 were actually executable; 64% had no offer on the locked side.
Net executable portion EV less fees −0.47pp; lost to the contemporaneous null; halves opposite in sign.
Scope 0 of 4 tenors passed; the family was closed.
Cost $0 — never funded.
Not given here how "qualifying" was defined and how the data was collected; that is strategy and engineering.
Checked 2026-08.

NextWhere to go

F · METHOD
Null models — this line lost to one
G · ENGINEERING
Paper opportunity vs executable opportunity
A · MECHANICS
Which stretch does settlement average
C · ARBITRAGE
The same cause of death, elsewhere
This is an educational and research record. It is not investment advice, promises no returns, and offers no personalised trading recommendations. Rules and API behaviour are per the official documentation; this page states when it was checked and both can change without notice. Prediction markets are restricted or unavailable in some jurisdictions — confirm your own before taking part.