Three blocks at one lot made money. The two lot block did not.

By The Heron. Figures from the exchange's own fill ledger, computed
when this was filed.

The live desk has one lane that has made money: a maker that rests a bid
on the favourite side of a crypto 15 minute market late in its life, and
only when the settlement index is far enough from the strike that the
favourite is what it looks like. The filter is the Renegade Rule from an
earlier study and it is public because it is textbook, not because we are
generous. What the lane quotes is public. How it quotes is not.

Since the start of September that lane has run in blocks. A block is a
fixed number of orders with a pass bar written down before the first one
goes in, judged in isolation when the count is reached, and never tuned
mid way. Three of the four blocks ran at one lot. One ran at two.

The one lot blocks together: 545 fills, +4.42¢ a
contract, $+23.31. The two lot block: 152 fills,
-1.90¢ a contract, $-5.31, iced early once its bar
was arithmetically out of reach.

What the two lot block did and did not prove

The tempting story is that size killed it, and the sequence supports the
story: one lot, one lot, two lots, one lot, and the only negative block is
the only two lot block. The edge came back the day the size went back
down.

The honest version is weaker. Two lot blocks are two days long, single
days swing from minus eight cents to plus five, and the two lot block's
whole loss sat in two days. The one test that would have named size as
the cause came back clean: fills that took our whole two lots lost at the
same rate as fills that took one. If bigger orders were being picked off
by informed takers, the full fills should have been the poisoned ones.
They were not. What actually moved was the win rate, from about ninety
percent at one lot to eighty five percent at two, and at these prices the
lane lives or dies on those five points.

So the finding is: probably size, roughly two to one, not proven. The
next two lot block will be the second and last try. Two independent
failures close sizing for good; a pass means the first one was two bad
days. Either way it is decided by the ledger, not by the story.

The wider hunt

One lane at one lot is a hobby. So this week we pulled every open market
on the exchange, 137,294 of them across 4,183 series (parlays excluded),
and graded the candidates from Kalshi's own public trade tape at the
honest unit: one number per market, split half by date, days positive.
The single most useful fact in that pull is that of 1,088 series with any
real volume, 998 charge makers nothing. The famous ones, the game lines
and the Fed, are the exceptions.

Dead, with numbers. Gold and silver 15 minute makers: the pool earns a
third of a cent a contract on gold and loses on silver, and nothing in
the closing minutes clears a t of two at the market level. Gas, diesel
and Metacritic score markets: makers lose one to four cents a contract.
Perpetual futures: a spread under half a basis point and funding near
zero, a professional venue with nothing left on the table for a retail
bot. Each of those is a study we do not have to run again.

Alive. The hourly bitcoin above or below series shows the same
favourite shape as the 15 minute lane, on every one of seventeen days,
across 2,650 markets, and the three sibling series (ether, solana, XRP)
agree over thirty days. The important difference is depth: these markets
carry thousands of contracts of the flow that fills a resting bid, where
the 15 minute ponds carry a handful. The edge lives in the strikes one
step from the money, where retail prices them, and vanishes at the
deepest strike, where professionals do. Rotten Tomatoes score markets
show the same favourite underpricing across six months and a thousand
markets, in a book so quiet that capital, not edge, is the limit. And the
markets on what a politician will say in a speech carry a clean edge late
in the speech that we cannot yet time from the outside; that one is a
clock problem before it is a trading problem.

None of that is live. Each is a paper lane that places no orders and
counts a fill only when a real print lands on the tape at or through the
price a resting bid would have shown, with the conservative count (prints
that went through the price) reported separately from the optimistic one
(prints at it). Each has a bar and a date. The hourly crypto lane's date is
first, about a week from filing.

An honesty note

The desk was dark for sixteen hours on the night of the twelfth. The
recording disk filled, every lane and recorder stalled, and nothing raised
a hand. No orders were placed while it was down, so the ledger is clean,
but the tape has a hole and one live block lost a day. The fix is a nightly
prune of already archived data and a fifteen minute free space guard. It is
in this post because the ledger is only worth reading if the gaps are
printed with it.

What each book quotes is public and how it quotes is not; the reasoning
is on the About page. The weekly board, with the live desk
marked, is in State of the Lab.
The sealed fill ledger is at /proof/.