Kalshi's $18.6M Settlement Error: When $100 Returned $9,800

On September 5, 2026, the prediction market Kalshi settled its Michigan-Western Michigan contract as a Western Michigan win and paid out that side while the game was, in fact, not over. Officials restored one second to the clock, Bryce Underwood threw a 47-yard touchdown to JJ Buchanan, and Michigan won 13-12. Kalshi then reversed the payouts, clawed back funds from traders who had already been paid, reimbursed and paid the Michigan side, apologized, and credited affected accounts $50. The market had traded roughly $18.6 million, and Michigan contracts bought in the window between the two Hail Marys — when the clock read zero and the market had written the Wolverines off entirely — turned $100 into roughly $9,800. The episode is the clearest recent illustration of settlement risk, the risk that a market resolves incorrectly or prematurely, which almost no recreational bettor prices into a position.

The Best Bet on Sports held live in-game positions on Michigan through that sequence, including a prediction-market position that closed at +410.03 percent. That is not the point of this article. The point is that a market with eight figures of volume resolved to the wrong outcome for a period of time, and the mechanism that caused it is one that can recur on any venue, in any sport, on any weekend. If you are wagering real money on live markets, understanding this failure mode matters more than any single ticket.

Our full breakdown of the game itself — and why the live price was structurally wrong before any of this happened — is in the Michigan Hail Mary live betting case study.

What Actually Happened, In Order

The sequence matters, because each step is where a different assumption broke.

1. Western Michigan took the lead late. A 33-yard field goal put the Broncos up 12-7 with 35 seconds remaining against the No. 16 Wolverines.

2. Michigan's first Hail Mary failed and the clock read zero. Underwood's throw sailed out of bounds. Western Michigan players ran onto the field to celebrate what would have been the program's signature win. By every visible indication, the game was over.

3. Data sources reported a final. With the clock at zero and a celebration underway, score-reporting services listed the game as a 12-7 Western Michigan final.

4. Kalshi settled the market on that data. The contract resolved as a Western Michigan win. Payouts went out to Western Michigan holders. The market had traded roughly $18.6 million by that point, which multiple outlets reported as a record college football volume day for the platform.

5. The game was not over. Officials reviewed and determined a Western Michigan defender had touched the ball from out of bounds. One second was restored to the clock.

6. Michigan won on the untimed final play. Underwood found Buchanan from 47 yards. Final: Michigan 13, Western Michigan 12.

7. Kalshi reversed everything. The company acknowledged it had, in its own words, wrongly and prematurely settled the market. It clawed back the Western Michigan payouts, restored funds to traders who had been marked as losers, paid the Michigan holders, apologized to users, and credited affected accounts $50.

The story was covered by Yahoo Finance, NBC Sports, Deadspin, Kotaku, ReadWrite, InGame, and SportsBettingDime, among others. Kalshi's own sports account had, during the game, spotlighted a $2,265 Western Michigan position that stood to pay $93,201 — a position that briefly won, then did not.

The Window: What a One-Cent Price Actually Implies

Here is the part that captured everyone's imagination, and it deserves to be stated carefully rather than exaggerated.

In the window between the failed first Hail Mary and the restored second, Michigan contracts were effectively worthless. The clock read zero. The celebration was underway. The market itself had resolved against them — that is the strongest possible statement a venue can make about an outcome's probability.

Michigan contracts traded around a penny in that window. A $100 position taken between the two Hail Marys returned roughly $9,800.

The arithmetic behind that is straightforward, and it is worth internalizing because it is the entire reason these windows matter. A prediction-market contract pays $1 if the outcome occurs, so the price *is* the implied probability. At one cent, $100 buys roughly 10,000 contracts, which return approximately $10,000 if the outcome hits. At two cents, $100 returns roughly $5,000. At five cents, roughly $2,000.

That is the mathematical structure of a market that has written an outcome off entirely. It is also why "the market already settled it" is not the same thing as "the outcome is certain" — and why, in the rare case where a settled market is wrong, the payoff profile is unlike anything available in conventional sports betting. For readers new to how implied probability converts to price, our guide to reading betting odds covers the conversion math.

The Contract-Rules Question Nobody Expected

The reversal solved one problem and created another, and this is the part of the story with genuine regulatory weight.

Kalshi is a CFTC-regulated exchange, which means its contracts operate under rules filed with a federal regulator rather than under the house policies of a sportsbook. Shortly after the reversal, traders and analysts publicly raised the question of whether the rules governing that specific market permit a post-expiration revision to change the expiration value — the settled result — at all.

The argument, as raised publicly, runs like this: if the contract expired and settled on the data available at expiration, the filed rules may not authorize revising that settlement afterward, even when the underlying real-world event later changed. Under that reading, the traders who were paid and then clawed back may have had a contractual claim to the original settlement.

We are not going to tell you how that resolves. It is a live question, it involves the interpretation of filed contract specifications, and reasonable people who read those documents professionally disagree about it. What we will say is that the question exists, that it was raised publicly by market participants, and that it is a materially different kind of dispute than "my sportsbook voided my bet." A sportsbook operates under its own terms of service. An exchange operates under rules filed with a regulator. The remedies, the precedent, and the oversight are not the same.

Settlement Risk Is Real, and Almost Nobody Prices It

Most bettors think about two risks: whether the pick wins, and whether the book pays. This game surfaced a third one.

Settlement risk is the risk that a market resolves to the wrong outcome, resolves early, resolves late, or gets reversed after you have been paid. It is invisible almost all of the time, which is exactly why it is underpriced. Consider how many independent assumptions were quietly load-bearing in this single game:

• That a clock reading zero means the game has ended.
• That a third-party score feed reflects officiating reality in real time.
• That once a market settles and funds land in your account, the money is yours.
• That the venue's remedy for its own error will match what its filed rules require.

Every one of those assumptions failed or came into question inside about sixty minutes. None of them is exotic. All of them are assumptions the average bettor makes without noticing, on every wager, on every platform.

The practical takeaways are unglamorous and worth more than any pick:

Do not treat a credited balance as final until the event is genuinely concluded. Funds appearing in an account is a bookkeeping event, not a legal conclusion.

Screenshot your positions, especially in chaotic finishes. Every documented play we publish on our verified results page is captured at settlement for exactly this reason. When a venue reverses something, your own records are the only version of events you control.

Read how your venue resolves disputes before you need to know. Sportsbook house rules, exchange contract specifications, and prediction-market settlement sources are all different documents with different consequences. Most bettors read them for the first time while angry.

Understand that faster settlement is a tradeoff, not a feature. Automated settlement off a live data feed is what makes near-instant payouts possible. It is also what makes premature settlement possible. You cannot have one without accepting some exposure to the other.

What This Means for Live Betting Specifically

Live in-game markets are where this risk concentrates, for a structural reason: they are the markets that depend most heavily on real-time third-party data and the least on human review.

That is the same structural gap our service is built around, viewed from the other side. Automated in-game pricing engines react to data feeds in seconds, which is what creates the brief mispricings a live bettor can act on — and it is also what allows a feed error or a premature final to propagate straight into settlement without a person checking it. The speed is the edge and the speed is the risk. They are the same property.

If you want the framework for how we identify live mispricings before they correct, read live betting versus pre-game picks and our breakdown of why sportsbooks limit winning accounts. For the college football side of the board, our college football picks hub covers the weekly slate.

The Honest Summary

A prediction market with roughly $18.6 million in volume paid the wrong winners, then unwound it. The platform corrected the error, made the correct side whole, apologized, and issued account credits. Whether the correction itself was permitted under the filed contract rules is an open question that market participants raised immediately and that has not been definitively resolved in public.

For bettors, the durable lesson has nothing to do with Kalshi specifically. Every venue — sportsbook, exchange, or prediction market — sits on top of data feeds and settlement logic that are correct almost always and catastrophically wrong occasionally. Almost always is not always. Price it accordingly, keep your own records, and never confuse a settled market with a finished game.

Past results do not guarantee future performance, and no service can promise a winning outcome on any individual play.

Frequently Asked Questions

What happened with Kalshi and the Michigan-Western Michigan game?

Kalshi settled its Michigan-Western Michigan market as a Western Michigan win after the game clock hit zero on a failed Hail Mary attempt, with score-reporting sources listing a 12-7 Broncos final. Officials then reviewed the play, determined a Western Michigan defender had touched the ball out of bounds, and restored one second to the clock. Michigan scored a 47-yard touchdown on the untimed final play to win 13-12. Kalshi reversed the payouts it had already issued, clawed back funds from Western Michigan holders, paid the Michigan side, apologized, and credited affected accounts $50.

How much money was involved in the Kalshi settlement error?

The market had traded roughly $18.6 million in volume by the time the result became final, which multiple outlets reported as a record college football volume day for the platform. Kalshi's own sports account had highlighted a single $2,265 Western Michigan position during the game that stood to pay $93,201 — a position that was briefly settled as a winner before the reversal.

Was Kalshi allowed to reverse the payouts?

That question was raised publicly by traders and analysts and has not been definitively resolved in public. The concern is that the filed contract rules governing the market may prohibit using a post-expiration revision to change the expiration value, meaning the original settlement might have been contractually final even though the real-world outcome later changed. Because Kalshi is a CFTC-regulated exchange rather than a sportsbook, its contracts operate under rules filed with a federal regulator, which makes the question meaningfully different from a sportsbook voiding a wager under house terms.

What is settlement risk in sports betting?

Settlement risk is the risk that a market resolves incorrectly, prematurely, late, or gets reversed after funds have already been credited. It is distinct from the risk of losing a bet and from counterparty risk — the risk a venue cannot pay. Settlement risk is usually invisible because automated resolution off live data feeds is correct the overwhelming majority of the time, which is precisely why most bettors never price it into a position until an event like this one surfaces it.

How much did $100 return if you bought Michigan between the two Hail Marys?

Roughly $9,800. Michigan contracts traded around a penny in the window between the failed first Hail Mary and the restored second, because the clock read zero, Western Michigan was celebrating on the field, and the market itself had already settled against Michigan. A prediction-market contract pays $1 if the outcome occurs, so price translates directly into return: at one cent, $100 buys roughly 10,000 contracts and returns approximately $10,000. At two cents, $100 returns roughly $5,000, and at five cents, roughly $2,000. That payoff structure is what makes a wrongly written-off outcome so asymmetric — and why it is unlike anything available in conventional fixed-odds sports betting.

How can bettors protect themselves from settlement errors?

Keep your own records — screenshot open and settled positions, particularly in chaotic finishes — because when a venue reverses something, your documentation is the only version of events you control. Read your venue's dispute-resolution terms before you need them, since sportsbook house rules, exchange contract specifications, and settlement-source definitions are different documents with different consequences. And treat a credited balance as a bookkeeping event rather than a final legal conclusion until the underlying event is genuinely concluded.

Does this mean prediction markets are less safe than sportsbooks?

Not inherently — they are differently structured, with different failure modes. Sportsbooks operate under house terms of service and typically settle with more human review and slower payouts. Regulated exchanges operate under contract rules filed with the CFTC and often settle faster off automated data sources, which enables quicker payouts and also creates more exposure to feed errors and premature resolution. Faster settlement is a tradeoff rather than a pure improvement, and bettors using either venue type benefit from understanding which specific settlement source and dispute process governs the market they are in.