State of the Lab — week ending Aug 2, 2026

This week the lab ran 109 seats across 4 desks and settled 19,865 paper trades. Best seat: Market-making · XRP 15-min. Roughest seat: Esports · longshots.

The lab lost about $401 this week across 109 seats and roughly 20,000 settled paper trades. That is the number. It is not the interesting part.

The interesting part is that only about $82 of it came from being wrong. The rest — near enough $319 — went to the exchange in fees. Four fifths of the loss was not a bad read on a tennis match or a temperature or a Bitcoin strike. It was the toll charged for having an opinion at all.

Look at the two ends of the board and the point makes itself. One seat buys heavy favourites: anything priced at ninety cents or above with two minutes left on the clock. It won 93% of its trades this week and still finished red, because winning a dime costs a fee whether you were right or not. At the other end sits a seat that buys underdogs into volume spikes; it won 8%. Opposite instincts, opposite hit rates, same column. Across every seat the lab ran, the aggregate win rate was 49.4% — a coin flip, priced honestly, bled dry by friction.

Then there is the one desk that made money. The market-making seats finished up about $34, and their fee column reads zero. Not low — zero. They quote both sides and wait to be traded against rather than paying to cross the spread, which on this exchange is the difference between renting a position and being paid to hold one.

I want to be careful there, because it is the one figure on this page I trust least. A simulated market-making fill cannot model why someone took the other side of your quote, and the honest answer is usually that they knew something you did not. Read that $34 as an upper bound, not an estimate. But the direction of it is hard to ignore when everything paying a fee is underwater and the one thing that isn't, isn't.

The weather desk did something else worth noticing: nothing. It opened no positions all week. Its entry filter wants a book depth that has not appeared since Tuesday, so 289 positions sat open and the seats declined to trade. Sitting on its hands is the best thing that book has done in a month — it is down about $99 since inception.

One more thing, and it is the reason a lab is worth running at all. Every trade is a bet at a price, and that price is the market's probability claim. With nearly twenty thousand settled trades spread across the whole book this week, the claim can be checked — and Kalshi comes out of it well at the top and badly at the bottom. Everything bought at 90–99¢ paid out 93.7% of the time against an implied 93.3%: honest pricing. Everything bought under 30¢ paid out about 1.5 percentage points less often than its price implied, in every bucket, on more than five thousand trades.

That is the favourite–longshot bias, the oldest finding in betting markets, alive on Kalshi in 2026 and measurable to a decimal. It also explains the board better than any individual seat does: the lab's cheap-side seats are not unlucky, they are buying something priced above what it is worth, and then paying a fee for the privilege. The full calibration table is in the members edition.

None of this is advice, and none of it moved real money. It is a machine placing paper bets at prices the exchange was actually showing, paying the fees it would actually have paid, published in full so you can check it. Next week it will probably lose again. The question worth answering is whether it can lose by less than the toll.

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.

Every seat behind these five lines — 109 of them, with win rates, fee drag, all-time results and the per-trade CSV — is in the members edition.

Fill verification: 1,890 of 21,548 trades (9%) 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.

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 — check this one against last week's issue.

Chain head as of 2026-08-02T04:12Z: b7e3b48594fc96b14c1ff33e3dba0cfce98f3cdbb14dd0f0ae925892cd2461c2