A slot machine in Vegas pays you back about 93 cents for every dollar you put in. That’s considered one of the worst deals in gambling.
On Kalshi — the legal, U.S.-regulated prediction market — there are bets where people get back 43 cents on the dollar.
Worse than a slot machine. By a lot. And people line up to take these bets every single day.
A researcher named Jonathan Becker looked at 72.1 million trades on Kalshi, worth a total of $18.26 billion, to figure out what’s really going on. What he found explains a lot about why most people lose money on prediction markets — and who’s on the other side of those losses.
Here’s the simple version.
The “longshot trap”
Kalshi works like this: every bet is a contract that pays $1 if it wins or $0 if it loses. The price tells you what the market thinks the odds are. A 5-cent contract should win 5% of the time. A 95-cent contract should win 95% of the time.
Should.
In real life, the cheap bets lose more than they should, and the expensive bets win more than they should.
-
Bets priced at 5 cents only win 4.18% of the time, not 5%.
-
Bets priced at 95 cents win 95.83% of the time, not 95%.
That doesn’t sound like much. But it means people who chase the cheap, exciting longshots are quietly getting fleeced. Every. Single. Time.
This isn’t a Kalshi problem — it’s been seen at horse tracks since 1949. People love betting on the underdog because the payout is huge if it hits. They overpay for the dream. The market figured this out a long time ago.
So who’s getting paid?
Here’s where it gets interesting. Kalshi (like every market) has two kinds of people trading:
-
Takers — people who hit “buy now” and take whatever price is on the screen. The impulsive ones.
-
Makers — people who place an order at a price they want and wait for someone to come to them. The patient ones.
Becker split the data by these two groups. The result is brutal:
WhoAverage return per tradeTakers (the impulsive buyers)–1.12%Makers (the patient sellers)+1.12%
The takers lose. The makers win. Every penny one side loses, the other side keeps. It’s that clean.
And it gets worse at the extremes. On 1-cent longshot bets, takers only win 0.43% of the time — less than half of what the price says they should. Makers selling those same bets win 1.57% of the time. The market is wrong, and one side eats the whole error.
The “I want it to happen” tax
Here’s the part that really stings. Takers don’t just lose because they’re impatient. They lose because they have a habit, and the habit is always betting YES.
People want to bet on things happening. “Yes, my team will win.” “Yes, Bitcoin will hit $200k.” “Yes, this celebrity will win the award.” Betting NO feels boring. Betting NO feels like rooting against your own life.
But mathematically, betting NO on a cheap longshot is way better than betting YES on it. Look at this:
-
A 1-cent YES bet pays back an average of –41%. You lose almost half your money.
-
A 1-cent NO bet pays back an average of +23%. You make money.
Same price. Wildly different outcome. The market is charging people extra for the privilege of feeling optimistic. Becker calls this the “Optimism Tax.”
In the cheap longshot range, takers make up 41–47% of all the YES buying. They’re piling into the side that loses. Patient makers are happy to sell it to them.
It depends what you’re betting on
Not every Kalshi market is a meat grinder. Some are pretty fair. The gap between takers and makers depends on the category:
What you bet onGap (how badly takers lose)Finance (stocks, GDP, rates)0.17% — basically fairPolitics1.02%Sports2.23%Crypto2.69%Entertainment (award shows, etc.)4.79%Media7.28%World events7.32% — brutal
The pattern is simple. Boring, technical questions = fair market. Fun, emotional questions = rip-off.
Finance bets attract people who think in math. Award show bets attract anyone who watches TV and “has a feeling.” Guess which group loses more money.
Sports is the biggest category by far — about 72% of all the money traded on Kalshi. The gap there isn’t huge per trade, but on $6.1 billion of taker volume, even a 2% loss is a ton of money flowing in one direction.
This is new, by the way
The wild part is that this rip-off didn’t always exist. Back in 2021–2023, the situation was actually flipped — takers were winning and makers were losing. Kalshi was small, mostly hobbyists, and the people placing patient orders were just as clueless as the people hitting “buy.”
Two things changed:
-
In October 2024, Kalshi won a court case against the CFTC and got the right to list political bets.
-
The 2024 election went bananas. Volume jumped from $30 million in Q3 2024 to $820 million in Q4 2024.
That kind of money attracted real professionals. Pro trading firms with algorithms moved in, parked themselves on the maker side, and started quietly hoovering up the spread from every impulsive sports fan and crypto degen hitting buy.
Before the election: takers were ahead by 2.9%. After: makers ahead by 2.5%. A swing of more than 5 percentage points, almost overnight.
The amateurs didn’t change. The pros showed up.
What it all means
If you bet on Kalshi, the math is straightforward:
-
Cheap longshots are a trap. They lose more often than the price suggests.
-
Betting YES is more expensive than betting NO at the same price. You’re paying for the feeling.
-
Hitting “buy now” is the loser’s move. Patience pays.
-
Boring markets are fair. Fun markets are not. Finance bets are nearly fair odds. Sports, crypto, and entertainment are where the money quietly leaves your account.
-
Pros are here now. Kalshi used to be a fair fight. It isn’t anymore. The other side of your trade is increasingly an algorithm built to take your money.
None of this means prediction markets are broken. They’re actually surprisingly accurate overall. It just means there’s a clear winning side and a clear losing side — and most people, without realizing it, are on the losing one.
The next time you see a 5-cent contract on something exciting and think “even if it only hits once in a while, I’ll win big” — remember that 72 million trades say you almost certainly won’t.