The Favorite-Longshot Bias Is Backwards On Kalshi Weather — And The Tick Size Explains It
The favorite-longshot bias is one of the oldest and most replicated findings in betting research: longshots are overpriced and favorites are underpriced, at racetracks, at sportsbooks, nearly everywhere anyone has checked since the late 1940s. So when we started measuring settlement rates on Kalshi's daily weather ladders, we expected the tails, the 2¢ and 3¢ bands for the record scorcher or the freak cool-down, to be mispriced in the usual direction: too expensive for how rarely they hit. Our running measurement keeps saying the opposite: on the cells where the quote is real, the deep tails have been settling more often than their price implies. The classic bias, inverted. A contrarian result with no explanation is a curiosity, so this piece leads with the explanation, and it is not a story about bettor psychology at all. It is a story about a one-cent tick, and by the end of it you will know where the vig on a ladder actually lives, because it has to live somewhere.
The Quick Answer
A century of racetrack and sportsbook research says longshots are overpriced; on Kalshi daily weather ladders we measure the reverse, with cheap tail bands settling more often than their prices imply. The mechanism is the 1-cent minimum tick: a market maker cannot load a proportional vig into a 2¢ contract because the smallest possible shade is half the price again, so the vig migrates to the big prices near the money and the tails trade close to honest. The precise finding, the tick-size arithmetic, and what it does and does not mean for anyone trading these ladders are all below.
What The Favorite-Longshot Bias Says
The finding was first measured at American racetracks in the late 1940s and has been replicated so many times since that economists treat it as a standing anomaly: bettors pay too much for longshots and too little for favorites. Back a long string of 50/1 horses and you lose far more than the odds alone would cost you; back heavy favorites and you lose less, sometimes almost nothing. Sportsbook moneylines and futures show the same tilt.
The usual explanations are behavioral. People like lottery tickets, so demand for the 50/1 thrill is strong and price-insensitive, and people misjudge small probabilities, reading a 1% chance as more like 3%. The bookmaker obliges: the longshot gets shaded hardest precisely because the person buying it is not doing arithmetic on the price. The bias has always been told as a story about the punters.
That framing hides an assumption so basic it almost never gets stated: that the operator can shade any price by any amount it wants. At a track or a sportsbook, odds are effectively continuous, so a fair 40/1 shot can be posted at 25/1 and the vig disappears into a number most bettors never decompose. Hold that assumption, because Kalshi breaks it, and breaking it is what flips the bias.
The Finding, Precisely Stated
Here is what we measure, stated carefully enough to survive a challenge. On Kalshi's daily temperature markets, each city's bands form a ladder that reads as a probability distribution, and each band settles yes or no against the official reading at the market's named weather station, so there is no ambiguity about what happened. Take only the rungs where the quote is a real two-sided market, where bid and ask both sit inside the cheap range rather than one stale print pretending to be a price. Group those rungs by price: the 1¢ to 5¢ deep tails, the 10¢ to 15¢ band, the 20¢ to 30¢ band. In every one of those groups, settlement has so far arrived more often than the price implies; call it an early read, not a completed proof. A band priced like a 3% event has been hitting more often than 3% of the time, not less. The cheap seats on a weather ladder have been underpriced, which is exactly backwards from what the racetrack literature says cheap seats should be.
Three honesty notes belong next to that claim. First, the sample is nowhere near statistically significant: separating a 3% event from a 4% event takes on the order of a thousand settled contracts, and weather markets have no closing line to grade against, so a claim like this can only be graded against realized settlement over a large sample, the same slow way any prediction market's accuracy gets judged. Second, the measurement is price-referenced on the conservative side of the quote, where our orders actually filled; read against the midpoint of the quote instead, the gap narrows. Third, we are not publishing the running numbers, here or anywhere, deliberately: this page is permanent, any figure printed on it would freeze and drift wrong, and our Kalshi weather markets hub carries the current honest assessment in words, not performance figures, by design. What earns this finding an article before the sample matures is not the count; the direction has a mechanism, and the mechanism was there before we measured anything.
The Mechanism: A Vig With Nowhere To Live
Kalshi prices move in one-cent steps. That sounds like trivia until you ask the question every market operator has to answer: where do I put my margin?
A sportsbook answers it anywhere it likes. An exchange is not a sportsbook, but a market maker quoting a ladder faces the same need: quote prices that, summed across the ladder, come to a bit more than 100%, so that being on the other side of the flow earns something. The overround has to be distributed across the rungs. And the one-cent tick dictates the distribution, because a shade smaller than a full cent does not exist.
| Fair Price Of The Rung | Smallest possible shade | That shade as a share of the price |
|---|---|---|
| 2¢ | 1¢ | 50% |
| 8¢ | 1¢ | 12.5% |
| 15¢ | 1¢ | 6.7% |
| 30¢ | 1¢ | 3.3% |
| 55¢ | 1¢ | 1.8% |
Illustrative arithmetic, not a live board: the shade is fixed at one tick, so its relative cost is set entirely by the price it lands on.
The 2¢ row is the whole argument. To collect any margin at all on that rung, the maker must move the price a minimum of one cent, which is a 50% markup on a 2¢ contract. A buyer who will pay 3¢ for a fair 2¢ band is rare, and a competing maker can undercut the inflated quote by a cent and take the entire queue. So the tail rung simply does not get shaded. There is nowhere to put the vig that the tick structure will accept. The honest refinement is that a maker can still earn the spread itself, resting a 1¢ bid and a 3¢ ask around a 2¢-fair band; what it cannot do is sit the whole quote above fair, because competition caps the ask while the resting bid fills below fair. At the tails the margin survives as a spread straddling fair, not a shade on top of it, which is exactly why fills out there can settle above their traded price. Down at 55¢, the same one-cent shade costs the counterparty 1.8% of the price, cheap enough to survive competition, and a couple of cents spread across the fat middle rungs funds the maker's whole operation. The vig flows downhill to where the prices are big, which is near the money, and the tails are left trading at raw probability.
Rounding then adds the last turn of the screw. A tail whose fair probability is 2.6% has no 2.6¢ price available; the tradable prices are 2¢ and 3¢, and a seller who wants a fill in any reasonable time transacts at the 2¢ bid, below fair. Near the money, a rounding error of half a cent on a 55¢ contract is noise. On the tails, the grid itself pushes traded prices under true probability. Underpriced tails are not a mystery in that structure; they are close to an accounting identity.
That mechanism also makes a prediction it could be wrong about, which is what separates an explanation from a story: if the inversion is really about tick size, it should weaken and disappear on any market with a finer tick or a wider price range, where the smallest possible shade is proportionally trivial at every rung. A ladder quoted in tenths of a cent would have somewhere to put the vig on the tails again, and the classic direction should reassert itself. That is testable, and it is the test we would want run against this finding.
The grid decides where the margin can survive. A one-cent shade is 50% of a 2¢ contract but 1.8% of a 55¢ one, so the overround parks near the money, and rounding pushes tail trades toward, and sometimes under, fair.
Same Bettors, Different Plumbing
Notice what this explanation does not need: it never mentions the psychology of the people trading. The behavioral stories about longshot lovers are presumably as true of weather traders as of racetrack crowds; nobody buys the 2¢ record-heat band without a little lottery ticket in their heart. But at the track, the structure lets the operator monetize that love by shading the longshot, and on a one-cent grid it cannot be done. Same humans, different plumbing, opposite bias. A market's biases are not laws of human nature; they are properties of its price grid. In most prediction markets, where prices range wide and a one-cent tick is proportionally small, the classic direction still holds, and our guide to what a prediction market price means rightly lists longshot bias among the reasons an extreme price can overstate its outcome. The one-cent weather grid, where the tick is half the price of a deep tail, is the structural exception, and the tick is the reason.
The sportsbook side of that contrast is worth one practical sentence, because American odds are exactly the continuous grid the weather ladder lacks. A book can quote a fair +4000 shot at +2500 and the shade vanishes into the number at any price point, with no one-cent floor to stop it; that invisible, infinitely divisible shade is the classic bias doing its work, and it is precisely the freedom the tick grid takes away from a weather market maker. The only defense against a shade that can live anywhere is seeing every book's version of the price at once, which is what our live odds screen does, and as we put it on our DraftKings vs FanDuel breakdown, certain books simply give you better odds in certain markets.
What This Finding Does Not Mean
The wrong reading of everything above is that cheap weather bands are a money machine, so this section exists to close that door, and we care about it more than any other section on the page.
Underpriced is not the same as frequently right. A 3¢ band that truly hits 4% of the time is still a contract that loses 24 times out of 25. Whether a string of those buys grinds out anything after variance is a question that takes, again, on the order of a thousand settled contracts to answer, and nobody holding a week of results knows anything. We are running exactly that experiment ourselves; it is far too young to say anything yet, and we say so plainly on the hub, where the assessment is kept current instead of frozen here.
Execution can eat the entire effect. When the whole question is worth a few cents, crossing the spread to get filled instantly can cost more than the mispricing you came for; an instant fill on a tail usually means you paid for it. And the obvious trade is the crowded trade: the cheap tail rungs are by far the most congested queues on the ladder, because everyone who has done the tick-size arithmetic in this article arrives at the same rungs, so the resting orders stacked ahead of you at a tail's ask dwarf the queue at an expensive rung. Being right about the direction of the bias does not mean you get filled at the price that carries it, and the rungs a reader would reach for after this article are exactly the ones where they cannot. Kalshi's fees are charged against a position whose whole upside is that same few cents, which tilts the arithmetic further. A finding measured at the resting bid does not survive being chased at the ask.
And the other side of these trades carries the risk shape this entire series repeats on purpose: selling an unlikely outcome collects a small premium and risks most of a dollar, so one loss erases the premiums from a long stretch of wins, and a red day that wipes out a green run is that trade working as designed, not breaking. We say that with our own money where our mouth is, because Stokastic trades these ladders and we are the seller. Read that disclosure twice: the finding in this article, if it holds, runs against the side we trade. We are publishing evidence that the premiums we collect on the tails are thinner than they look, which we would rather do honestly than discover expensively.
How To Use It If You Trade These Ladders
What survives all those caveats is a reading habit, not a pick:
- Start from how these markets work and read the whole ladder, not one band in isolation; the middle rungs are where the vig is parked, so the shape of the middle is what tells you whether a tail price is structure or opinion.
- Before treating a cheap rung as mispriced, check that its quote is two-sided with real depth. A lone 2¢ resting print is the stale-print case this article excludes, not the mispricing it describes.
- Remember that afternoon highs and overnight lows are two different games that resolve at opposite ends of the day, so a tail on one book is not evidence about the other.
- When your own honest distribution says a band is worth 2.5¢, understand that the grid will quote it at 2¢ or 3¢, and which side of that rounding you transact on is most of the trade. Rest orders instead of crossing spreads: on a tail, the spread you cross is the same one-cent gap this whole article is about.
- Size every position as if the miss lands tonight, because the tick structure that makes tails cheap does nothing to make them frequent.
And if you want to watch probability-first thinking applied with real prices attached before putting a dollar anywhere, our free expert picks cost nothing and show the habit in practice daily.
Favorite-Longshot Bias FAQ
What is the favorite-longshot bias? The favorite-longshot bias is the long-documented pattern, first measured at racetracks in the late 1940s, that longshots return far less than their odds imply while favorites return close to fair. In plain terms: the betting public historically overpays for low-probability outcomes, and operators shade those prices hardest because demand for lottery-style payoffs is price-insensitive.
Why would the bias run backwards on Kalshi weather markets? Because of the 1-cent minimum tick. A market maker cannot load a proportional margin into a 2¢ contract, since the smallest possible shade is a full cent, a 50% markup that competition immediately undercuts. The vig therefore concentrates on the big prices near the money, and tail prices sit at, or get rounded below, honest probability.
Does that mean buying cheap weather bands is profitable? Not on any evidence we would defend. Even a tail priced like a 3% event that truly hits 4% still loses about 24 times in 25, spreads and fees can eat the entire gap, and proving the gap exists at all takes on the order of a thousand settled contracts. We trade these markets ourselves and keep the current, deliberately number-free assessment on the hub, because a claim like this deserves to be graded against realized settlement over a large sample, not asserted off a hot streak.
Where does the vig live on a prediction market ladder? Near the money. On a one-cent grid, a one-tick shade costs 1.8% of a 55¢ contract but 50% of a 2¢ contract, so the middle rungs are where a market maker's margin can survive competition. Summed across a ladder, the overround exists, but it is carried almost entirely by the big prices rather than spread evenly.
A century of bias research quietly assumed the operator could always put the vig wherever the punters were softest, and for a century of continuous odds that was true. Shrink the price grid to one cent and the assumption dies on the cheap rungs: the 50% markup in that table's top row is the sound of a vig with nowhere to live. That is the whole inversion, and it is why we trust the direction of this finding even while the sample is too young to trust the size of it. The grid rearranged where the margin can hide, the toll left the tails, and the still-unproven attempt to find out what that is worth continues one temperature at a time, with our current read of it on the Kalshi weather markets hub.
Disclosure and fine print. Stokastic trades Kalshi weather markets and holds positions in them; where this series describes a settled position, we were the seller, which means the finding in this article, if it matures, runs against the side we trade. Kalshi event contracts are CFTC-regulated event derivatives traded on a designated contract market, not sportsbook wagers, and a position can lose its full value. Selling an unlikely outcome collects a small premium and risks most of a dollar, so one loss can erase the premiums from a long stretch of wins; size accordingly. 18+; Kalshi offers broad, state-specific availability under federal oversight, and the risk of loss is real and, on the side we trade, individually large. This series is an open research log of a strategy we have not proven; its sample is still far too small to judge, and no figure on this page is a live price. Nothing here is trading advice, and nothing on this page is a pick or a recommendation.
