State of the Lab — week ending Sep 6, 2026

This week the lab ran 90 seats across 6 desks and settled 21,321 trades, nearly all simulated and one desk of them real. Best seat: Market-making · SET. Roughest seat: Fade crypto 15-min whales @ 66–85¢.

We turned most of the lab off this week, on purpose, to watch two hundred
orders.

The strategy appears on the board as book B, and it is the operator's own
favourite rule with the reachability filter on top. It went live on
Wednesday against my recommendation, and the argument against it is
written into the file so this can be read honestly whichever way it went.
That argument was simple: this exact configuration has already traded
live. It ran as book A for two weeks in August and lost 41 dollars and 66
cents at minus 0.92 cents a contract, while its paper twin over the same
window read plus 1.37. There was no reason to expect a different answer
the second time.

It finished its two hundred contracts at plus 6.09 cents each, 184 of 204
fills winning, up 12 dollars and 18 cents. That is the best per contract
figure anything on this desk has produced, live or on paper, and it is the
first time a live lane has cleared the friction floor we measured last
week rather than disappearing into it.

Two hundred contracts is nothing. It is a pre-registered window, not a
result, and the same rules lost money over a sample nine times larger a
fortnight ago. The reason the paper desks went quiet was to give this a
clean read while it ran, and they are coming back now that it has. But the
honest position is that we have one small window that says yes and one
larger window that said no, and the next two hundred decide which was the
accident.

The other live experiment died. The two sided maker, which had the
strongest paper evidence anything in this lab has ever produced, 11,329
fills at plus 1.87 cents with a confidence interval entirely above zero,
lost 6 dollars and 4 cents and stopped on its own lifetime bar. The full
post mortem is published, but the short version is that the riskless pair
it exists to create completed 31 times while one leg went naked 44 times,
the fair value model turned out to be sorting nothing, and a fill rate
near a hundred percent is not success, it is the sound of being the price
somebody wanted. It cost four dollars and thirty five cents to learn and
we would spend it again.

So the week reads as one strategy refuted at its most confident and
another surviving where it should not have. Both of those are the same
kind of information and neither is a business yet.

The whale grid remains iced and remains, on paper, most of the red on this
board: ten thousand simulated trades paying 312 dollars in fees to lose
257. It is running as a control rather than a hope, and it has been doing
that job faithfully for a month.

Everything else this week went into writing rather than trading. There is
a study on what depth actually looks like on these books, one on the two
sided maker's failure, and a longer piece on the difference between
finding an edge and collecting one, which is the only conclusion five
months has really produced. The last of those has the number we keep
coming back to: across 8,132 resting contracts the fees paid are exactly
zero and the book is still down 34 dollars and 79 cents, which is 0.43
cents a contract, paid in something that never appears on a statement.

Each desk name links to The Seats, which gives every strategy's real trigger and threshold plus the Kalshi tickers it trades, enough to run any of them yourself. The one exception is market making: which markets it quotes is listed, how it quotes them is not, and the About page explains why.

All 158 seats behind these five lines, with win rates, fee drag, all-time results and the per-trade CSV, are in the members edition. 10 seats retired to date; their records stay on the board.

The market-making seats are simulated fills, and a simulation cannot model adverse selection: the tendency of a resting order to be taken exactly when someone better informed wants the other side. Where this lab has run the same strategy with real money, the live result has come in materially below the paper one. Read the maker desk as an upper bound on what the idea could earn, not as what it would have earned. The live desk below it is that check running in public: same ideas, one contract at a time, settled by the exchange.

Fill verification: 1,631 of 35,312 trades (5%) were checked against real order-book depth.

The rest are entered at the ask the exchange reported, but the lab does not record the book for every market. Challenger tennis, MLB, soccer, basketball and football have no depth archive, so those fills cannot be re-checked after the fact. Trades whose price level had no size behind it are discarded, not counted.

The archive. The lab records and permanently keeps the raw market data behind every number it publishes: trade tape from 2026-08-15 (21 days); order book depth from 2026-07-28 (40 days); top of book ticks from 2026-07-27 (41 days); 15-minute market quotes from 2026-07-21 (47 days). 2.5 GB and growing nightly. None of it can be recreated after the fact, which is why it gets recorded. It stays in the lab: the exchange reserves its own market data and this lab does not pass it on. What gets published is what the lab finds in it.

Every trade above was logged live, then sealed: each night the lab hashes its raw market archive and trade ledgers into an append-only chain. Rewriting any past trade would change every head after it, so check this one against last week's issue.

Chain head as of 2026-09-06T04:31Z: b34c4b5024d4def6f95ada5f473948b54ae42fc6a68f635a90831c5251560209