Strategies
Every built-in comes with the lab's own result, so a template is never mistaken for an edge. Modify one (the original stays as it is), or start a new one. Settings and the rule's code are edited together.
Bet against a sudden price spike, expecting it to come back. Popular in backtests; the lab's paper grade lost on all 12 coins it tried.
Lab record · paper
Paper, honest fills: lost on all 12 coins. Spikes on Kalshi mostly carried real information.
Jul to Oct 2026
Buy the side the market already favours, late in the window. The lab's most tested idea: it wins often and still loses money, because the price you pay already says it wins often.
Lab record · real money
Won about 9 trades in 10 and lost money anyway: one loss at 88c wipes out about eight wins. Real-money record.
Apr to Sep 2026
Copy the side of the first $1,000+ trade in a market. The lab tracked 1,855 of them, about $4 million: the big money lost 4.9%, and copying it faster did not help.
Lab record · paper
Public tape, Aug to Sep 2026: -2.92c a contract whether you copied 0.25 s, 2 s or 10 s later.
Aug to Sep 2026
When Bitcoin jumps, buy the side the jump favours before the Kalshi price catches up. The lab measured Kalshi repricing within about a quarter of a second.
Lab record · real money
Kalshi already moved about 4.3c within 0.25 s of the jump. Chasing it paid the new price and the fee.
Jul to Sep 2026
Rest a bid on whichever side is priced 5 cents or less. Pays about a dollar when it hits, and the lab's finding is that it does not hit often enough. Shipped so you can watch that for yourself instead of taking our word.
Lab record · paper
Bought the long shot 38,565 times on paper across twelve 15-minute crypto markets, paying the ask: 2.8 cents a trade short, on every one of the twelve. This copy rests a maker bid instead of paying the ask, the one change the study said might matter. Expect long runs of misses between hits.
study, Oct 2026
Rest bids below the market and let other people come to you, on series where makers pay no fee. The only family in the lab's record that finished ahead with real money, before it failed its own test at size.
Lab record · real money
+$51.27 over 1,459 real fills at 1 to 2 lots. Failed its pre-registered bar when sized up to 4 lots; retired.
Aug to Sep 2026
Buy contracts priced 97c to 99c in the last minutes: a near-certain cent or two, many times a day. The lab's 'picky' version is on a forward test now.
Lab record · paper
52,323 contracts graded from the tape. Pays a cent most of the time and loses 97 the rest; the test is whether the rare loss comes often enough to eat the pennies.
forward test from Oct 2026
Trade the arithmetic, not the chart. In the last minute of a 15 minute Bitcoin market, part of the settlement average is already decided; this rule projects it and buys the side it favours when the projection sits far enough from the line. You set the window, the distance and the price cap. The lab's live rule in this family runs with settings that are held until its verdict; these defaults are neutral.
Lab record · paper
Not a lab result: the lab's own live rule in this family uses held settings. The public measurement behind the idea, over 635 real settlements: with 10 seconds left the projected average missed the final value by about 47 cents, the spot price by about 4 dollars; in close calls ($5 to $20 from the line) the projection picked the right side 98 times in 98, spot 92 in 99. Whether the side it picks is worth what the market charges is the whole question.
measured Oct 2026
Sam's Wednesday rule from the videos: a long shot priced at a few cents that pays about a dollar when it hits. One contract. Here so you can watch why one good Wednesday does not make a good week.
Lab record · real money
Up $1.18 by Wednesday, -16c for the week. One 88c favourite loss undid eight wins.
one week, Oct 2026
