State of the Lab — Nov 8, 2026

No. 15 · 8 Nov 2026 · All ponds

State of the Lab — week ending Nov 8, 2026

This week the lab ran 21 seats across 4 desks and settled 26,265 trades, nearly all simulated, 54 of them real money. Best seat: Speech markets. Roughest seat: Early leaning favourite.

Real money this week. The last minute bot settled 30 real trades and won 17, for minus $2.93 after fees. Since it went live on 14 October it stands at minus $1.65 on 92 contracts, every fill from Kalshi's own records. It is in its second block of 60 contracts, judged on the same bar as the first, which it failed. The new cheap side lane, the same signal but only under 55 cents, had its first week: 16 trades, 3 won, minus $1.22. Cheap fills win less often by design, and one week of a lane that needs about ten weeks says nothing yet.

The paper twin of the last minute rule had a worse week than real money: 31 trades at minus 14 cents a contract. Last minute looked weak almost everywhere, which is exactly what this week's main new piece of work is about. We built a market adapter: a switch that sits above our strategies and only lets a trade through in conditions where that strategy has worked before. We wrote down four candidate conditions before any new data arrived and grade each one only on trades made after that. When the first days pointed the wrong way, we added the four opposites, also judged only on trades from that moment on, so the days that prompted them do not count. Nothing trades on any of this. After four days no condition passes, and several are keeping the worse trades and skipping the better ones. Verdicts come in late November.

Three more results came in, all failures, so all free. The night against day test failed and was stopped on 3 November: the night trades lost 1.36 cents a contract over 1,078 fills. Resting a limit order inside the spread, instead of paying the offer, failed: the order filled most of the time, saved about 7.5 cents of price, and still lost 10.67 cents a contract, because the fills it got were the trades turning against it. And a popular stock breakout method, tested on eight years of US stocks including the ones that later delisted, averaged a small gain per trade but lost to simply holding the index: minus 11.8 percent a year against plus 16.9.

The crypto desk settled 25,077 fifteen minute trades for minus $356.78, after $381.45 of fees. Before fees it made $24.67, and the toll took all of it and more. The Bitcoin hedge collected $1.07 in funding this week and $4.18 since 18 October. The whole package, a Kalshi short with the same amount of real Bitcoin so price moves cancel, stands at minus $0.55 including about $5 of one time setup fees. Kalshi's first 7 percent payment is due by about 13 November.

Published this week: a viewer's 90 cent ladder, tested on 44,000 entries, which wins 95 percent counted the way paper counts and loses money counted honestly, and the limit order test above. A second viewer idea, a bot that looks for the side about to flip in the last few minutes, was tested this weekend on 22,482 trades and failed: the cheap side won 21.6 percent of the time at an average price of 20 cents, almost exactly what it cost, and the fee took the rest, leaving plus 0.09 cents a contract. The hourly version of the last minute rule gets its verdict on 11 November. We are still hunting, and every one of these results narrows the search.

Members watch all of this move in the control room, rebuilt every hour, and run the strategies themselves in Ping, live on paper. Your first month is $5.

Desk Settled Net ¢/trade Fees
Market making 0 $0.00 — $0.00
live money 54 −$4.15 −7.69¢ $0.52
Paper strategies 1,134 −$34.89 −3.08¢ n/a
Crypto 15-min rules 25,077 −$356.78 −1.42¢ $381.45
all desks 26,265 −$395.82 −1.51¢ $381.97

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 188 seats behind these 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: 26,307 of 26,308 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 (84 days); order book depth from 2026-07-28 (103 days); top of book ticks from 2026-07-27 (104 days); settlement index from 2026-08-16 (84 days); 15-minute market quotes from 2026-07-21 (110 days). 15.0 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 2026-11-08T05:22Z: 029eb4443208052fc32f200d95b67fe16ca885e8497b049331d1265409e6a39e

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