State of the Lab — week ending Aug 30, 2026

This week the lab ran 100 seats across 6 desks and settled 28,683 trades, nearly all simulated and one desk of them real. Best seat: Market-making · SET. Roughest seat: Fade crypto 15-min whales @ 66–85¢.

Book A gave it all back this week. It lost 29.20 over the seven days and now
sits at minus 16.70 since it started, which means the lane that carried this
desk for a month has handed back every cent it made and then some. Eleven days
ago it was up about fifty dollars. That is what a one to two cent edge looks
like when it stops working, and it is the reason the number to trust here has
always been the long column rather than the week.

We do not yet know whether that is variance or the edge decaying. Three
thousand fills is a real sample but a small per fill figure needs a much
larger one before it can be told apart from noise. What we can say is that
the thing which has bothered us for a month is the likeliest culprit, and this
week we started testing it directly.

The problem in one sentence: every lane we have taken live harvests a spread
from somebody who chose to trade with us, and every one of them has been eaten
by the same tax. The paper twin reads about a cent and a half a fill. The live
version runs about nothing. Three cents, measured twice now on identical rules
in identical markets. Somebody is filling us precisely when they want the other
side, and often they are right.

So two new lanes went live, and they are two halves of one question.

The first rests offers well away from the market and captures no spread at all.
It earns something only when a person reaches across to take a deliberately
silly price, which is not a trade an informed counterparty makes. If the tax
comes from being picked off by people who know more, a lane that can only be
filled by mistake should not pay it.

The second does the opposite. It quotes both sides and captures a spread, but
before the match starts, when there is no game state for anyone to know more
about. If the tax is about the spread itself, this one should suffer. If it is
about information arriving while you sit there, it should not.

Between them they should tell us whether we have been mispricing our own
strategy or simply standing in the wrong place. Eighteen fills and one fill so
far, which is not a result. It is a pulse. The oil lane also went live and has
forty fills at plus a dollar twenty, and that is not a result either.

The week also cost us nine hours of crypto downtime, and the reason is worth
knowing if you run anything automated on this exchange. Kalshi is splitting its
matching engine by category, crypto moved on the twenty fourth, and orders that
do not name the new engine come back saying the market does not exist for
markets that are open and quoting. Our bot never stopped. It traded oil, which
did not move, so fills kept landing and every health check stayed green while
four fifths of the desk was dead. We have written the whole thing up
separately, including the part that cost the most time, which is that
collateral does not follow your order across engines.

The bottom line has the shape it always has. The whale grid settled thirteen
thousand simulated trades, paid 367 dollars in fees to lose 332, and is again
most of the loss and nearly all of the friction. Weather had its worst week in
a while. The maker desk on paper reads plus 174, and the live desk beside it
reads minus 30, which is the whole argument of this publication in two lines
and the reason we keep printing both.

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 156 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: 1,716 of 42,664 trades (4%) 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 (14 days); order book depth from 2026-07-28 (33 days); top of book ticks from 2026-07-27 (34 days); 15-minute market quotes from 2026-07-21 (40 days). 1.6 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 2026-08-30T04:23Z: 2091cf2f41a030dcb09a834b96109345571e8b59ed83f21ad4222748d753f596