Nickel Sloth

About

Nickel Sloth is the only agent here with no opinion about sport at all. It does not read a team, a venue, an injury or a form line. It reads two prices on the same market and the gap between them. Handed a weather market or an election market shaped the same way, it would trade it identically, which is the point: prediction markets are a venue before they are a sports product, and this is the rule that treats them that way.

What it acts on is the oldest documented pattern in betting markets. Cheap outcomes are priced richer than they turn out to deserve and dear ones cheaper, and the effect has been reported at racetracks since 1949 and on prediction-market exchanges since. This agent takes the far side of the cheap outcome, which on a two-sided game market means buying the expensive side and holding it to settlement. It fires only between 94 and 97 cents.

Both edges of that band are arithmetic. The bottom is the first cent Picky Badger refuses: that rule's band ends at 93 cents and its published line is never at 94 or above, so starting here means no contract on any board can ever be priced by both agents and the two records stay independent. The top is where the unavoidable friction, Kalshi's fee plus half of the tightest spread the exchange can quote, passes a quarter of the whole payoff: 23.5 percent at 97 cents, 31.9 percent at 98. Past that the trade is a fee wearing a costume.

The honest cost is the shape of the position. At 95 cents a loser is worth about nineteen winners, and the record is therefore very slow to say anything. Telling a real 2-point edge from no edge at that price takes roughly 930 settled picks; a 1-point edge takes about 3,700. Across every board and every settled week this site holds, 2 of 104 quoted sides sat inside the band. At that rate this record will not be readable for years, and quite possibly never. It is written here at the start rather than discovered at the end.

Published 23 September 2026 with an empty record and a stated retirement condition: if it has fewer than 50 settled picks two full seasons from now, the band is too thin to measure on this board and the agent should be retired rather than left standing as decoration.

The rule

Read both sides of a two-sided win market and take the dearer one. Buy it only when its ask sits between 94¢ and 97¢, the bid-ask spread is 1¢, at least 100 contracts rest at that ask and the contract has traded at least once. The cheap side has to carry its own traded two-sided quote: if 100¢ minus the cheap side's bid comes out below the dear side's ask, the same claim is cheaper as NO on the longshot than as YES on the favourite and there is no pick. Otherwise buy 100 contracts at the ask, no earlier than 24 hours before the start and no later than 30 minutes before it, and hold to settlement. No sell target. Anything outside the band, or on a wider spread, is no pick.

How it works

• Runs on the scheduled job like the other armed agents, twice an hour, across six game series: WNBA, NFL, MLB, NHL, NBA and NCAAF. Nothing about it is operated by hand.
• Reads the whole event, not one side of it. An event has to present exactly two quoted sides, because a lone expensive quote is not a longshot being faded, it is just a high number.
• Sorts the two sides by ask and takes the dearer. It never checks which one is at home, because at 94 cents and up only one side of a two-way market can be in the band anyway.
• Runs the house gate on that side: 100 or more contracts resting at the ask, a contract that has traded at least once, and then a spread test of 1 cent rather than the house 3. The house figure is a ceiling, and this agent binds itself under it, never over it.
• Checks the other route into the same claim. A Kalshi market's NO ask is 100 minus its YES bid, so the longshot not winning can be bought as NO on the longshot as well as YES on the favourite. When the NO route prices lower, the agent refuses outright: the board publishes resting size at the ask only, so the cheaper route cannot be put through the depth gate, and paying the dearer of two visible prices is not something worth stamping on a public ledger.
• Enters at the ask, 100 contracts flat, no sell target, held to settlement. The row carries the bid, the ask, the resting size, the name and price of the longshot it faded, the NO-route price it compared against and the fee-plus-spread hurdle in probability points, so the whole arithmetic can be recomputed later from the row alone.

It refuses far more often than it acts, and the refusals are the ordinary case. Anything under 94 cents belongs to another rule; anything over 97 has less payoff than friction; a 2-cent spread is a refusal even though the house would allow 3; a market with one quoted side has no longshot to fade; and a board where NO on the longshot is cheaper than YES on the favourite is left alone entirely. Whole weeks with no pick at all are the expected behaviour, not a fault.

What would prove it wrong

If the dear sides it buys between 94 and 97 cents settle at about the share those prices implied, there is no favourite-longshot premium left at the top of this board once the fee and the spread are paid, and the rule is only buying the market's own opinion at a small markup. The quieter failure is that the band never fills: a rule that cannot accumulate settlements cannot be falsified either, and an unfalsifiable agent has no business on a roster that grades itself.

Why this rule

Tests the favourite-longshot bias at the one stretch of the book nothing else on the roster reads. The bias is documented as a shape rather than a sport: cheap contracts settle NO more often than their price implies and dear ones settle YES more often than theirs does. Picky Badger measures 80¢ to 93¢ in one league; this reads 94¢ to 97¢ across six and holds no view on any of them. The band is set from the published literature, not from this site's own settled history, because this site does not have enough of it. At 95¢ Kalshi's fee is 0.33 points of probability and crossing a 1¢ spread is another 0.50, so the position has to be underpriced by 0.83 points to break even, which is 16.7 percent of everything it has to win. On the 3¢ house spread gate the same position needs 1.83 points, or 36.7 percent of the payoff, which is more than the documented effect is usually reported at. That is why this agent binds itself tighter than the house and refuses rather than widening.