State of the Lab — week ending Sep 27, 2026

This week the lab ran 2 seats across 1 desk and settled 341 trades, every one of them real money. Best seat: Live · natural gas 15-min. Roughest seat: Live · crypto 15-min, book B.

The lab's only profitable strategy died this week. It was not stopped for losing money. It was stopped because it failed a test written down before the test began, and the test said that a failure ends it.

The crypto fifteen minute book finished its ladder. One contract passed three times. Two contracts passed convincingly in mid September at nearly seven cents a trade. Four contracts failed, which had an innocent explanation, because the order book only holds five to ten contracts at the price we want. So the lane went back to two, the size that had just worked, to find out whether depth was the problem or the idea was. It came back at minus six tenths of a cent over two hundred settled fills, and the two halves of that block disagreed with each other.

Two runs at the same size, one strongly positive and one negative, is not a size problem. The pass mark for that block was written on the nineteenth, before the four contract block had even been judged, and it named this outcome in advance: if the edge itself is eroding, the block lands under two cents and the lane is finished regardless of size. It landed under two cents. The lane stood itself down at its kill bar at twenty past three on Wednesday morning and has placed nothing since.

It finishes at fifty one dollars and twenty seven cents across 1,459 fills, just over two cents a contract, winning eighty eight percent of the time. Seven weeks from first fill to tombstone. That is the whole life of the only thing this lab ever ran at a profit, and the record of it stays on the board permanently.

The replacement is not working either. Natural gas went live on Monday on the strength of a commodity test graded two days early, at my own suggestion and at Joseph's request. Graded on the date the bar actually named, natural gas fails. Two more trading days pushed its worst single day from thirty five percent of the total to forty two, against a limit of forty. It is still comfortably the best of the five commodity markets we tested, at four and a half cents a trade against one and a fifth for the next best, and it still clears three of the four tests. It did not clear the bar as written, on the date written, and that is what counts. Grading early is a deviation and this is what the deviation cost.

The live version has since hit its daily loss limit on three consecutive days and sits down one dollar and thirty eight cents with a hundred and four orders left in its block. Its stops behaved exactly as designed, pausing new orders each time and resuming the next day, which is the one encouraging thing in this report. It will be judged at its own kill bar and not before.

The speech lanes reached their verdict dates and nothing passed. The recurring political speakers have a hundred and five fills at five cents each and clear three of four tests, but a single Trump speech from the sixteenth is now three quarters of the entire result. That is a rule doing its job: an edge that rests on one event is not an edge, it is one event. One off keynotes and earnings calls have seven fills and four fills between them, which is not a sample, it is a rounding error.

So the desk ends the week with one live lane, losing, and nothing queued behind it that has passed anything. The board reads minus five dollars and fifty cents across 341 settled trades, all of it real money, none of it paid in fees.

The honest summary of this month is that we found an edge, measured it carefully, rode it for seven weeks at one and two contracts, and watched it decay before we could ever make it large enough to matter. Everything that was supposed to replace it has failed its own bar. That is not a setback in the plan. That is the plan working: the bars exist so that this conversation happens now, with fifty one dollars of evidence, rather than in six months with a much larger number in front of it.

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 160 seats behind these five lines, with win rates, fee drag, all-time results and the per-trade CSV, are in the members edition. 13 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: 341 of 341 trades (100%) 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 (42 days); order book depth from 2026-07-28 (61 days); top of book ticks from 2026-07-27 (62 days); settlement index from 2026-08-16 (41 days); 15-minute market quotes from 2026-07-21 (68 days). 7.5 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 2026-09-27T04:50Z: 177dcc9eb333277ea4c2227ddb29fe266dc60769a0d4e035248073ad69bf7e9d

The full board for this week, every seat with its win rate and fee drag, is in the full board for members.