John Locher / AP

On Monday night, Lester Holt questioned Donald Trump on the billionaire’s long-held belief that Barack Obama was secretly born in Kenya. I’d bet anything that a bunch of folks lept out of their chairs and cheered, positive they had hit the jackpot.

For days, internet gamblers had debated whether Holt would bring up the birther issue, placing wagers on PredictIt, a popular political betting exchange. The market settled on odds of 40 percent for “Yes”; when Holt began his line of questioning, a lot of people expected to make a tidy profit.

But they didn’t. When the debate ended, PredictIt ruled the result a “No,” meaning Holt hadn’t brought up the issue. A closer look at the actual betting language explains why:

During the presidential debate held at Hofstra University on September 26, 2016, a moderator or sanctioned questioner shall utter the word “birther”, “birthers”, or “birtherism.”

See, Holt never actually used the term “birther.” He said “birth certificate.” He mentioned that Obama was “born in the United States.”  But he didn’t say any of the three magic words that would have fulfilled the bet’s conditions. Yes, the spirit of the wager—that Trump’s fanatical pursuit of Obama’s birth certificate would become an issue in the debate—was met. But the letter of the wager wasn’t.

Some would-be-winners got pretty heated. One commenter’s take:

Resolving this to “no” is completely antithetical to the purpose and spirit of the market. The purpose of the market is to accurately assess the likelihood of a particular event, not to describe that [event] in excruciating detail. The intent of this market clearly was whether the moderator introduced the concept of “birtherism.” The fact that the rules allow PredictIt the latitude of interpretation implicitly support this.

Another bettor fought back:

The rules were quite clear. The moderator must say the words “birther”, “birthers”, or “birtherism”. I bet based on those rules. The moderator did not say any of those words.

The “spirit” of the market was expressed very clearly in the rules. I'm sorry for your loss.

When the debate concluded, PredictIt had two choices, both bad. It could stick with the contract language and deliver a “No” ruling, appearing pedantic and pissing off bettors who felt they were cheated out of a win. Or it can acknowledge the intent of the question and switch its ruling to a “Yes.” But its reputation could suffer. After all, if the exchange shows it is willing to fiddle around with one contract after the fact, what’s to stop it from invalidating a hundred others later?

PredictIt opted for “No” for that very reason. “Why would you put your money into a market where the rules could change at any time?” asked spokesperson Brandi Travis. “That’s not the kind of market we want to run… You just can’t go against the rules. That’s why they’re written. That’s why they’re very black-and-white.”

This has come up before. In 2015, one PredictIt wager asked whether New Jersey Governor Chris Christie would enter the race for president by the end of June. On the very last day of the June, Christie announced his intention to run—but he didn’t file his Federal Elections Commission paperwork until the next day. The gamblers who bet on his June entrance lost their money.

Similarly, in 2012, another exchange awarded payouts to gamblers who bet Mitt Romney would win the Iowa Republican caucuses. A later recount would confirm a tie between Romney and Rick Santorum, but the exchange held firm, arguing the original results fulfilled the rules of the betting contract.

So much rests on how the bets are worded; they must be precise enough to eliminate ambiguity, but broad enough to encompass every possibility. “It is really, really hard to write questions,” said David Rothschild, an economist at Microsoft who studies prediction markets. “I cannot stress enough how many times professional markets, well-done markets have had controversies over this. There are just so many different things that could happen.”

To most, this probably seems like minor kerfuffle. Gambling in political betting markets is a niche fascination at best; most people don’t get a thrill out of guessing whether Obama’s approval rating will be 51 percent or higher between October 4 and October 6.

That said, as regular folks grow increasingly interested in the numbers behind elections, media outlets are paying more attention to how each candidate fares on prediction markets. Sharp movement in one direction or another could spark a story; it might also jog one of the many live election prediction models that have become so popular. If prediction markets are going to influence an election, it is imperative that they’re run well. PredictIt will take plenty of heat for their decision to stick by the rules. But it’s for the best that they did.

We want to hear what you think about this article. Submit a letter to the editor or write to letters@theatlantic.com.