In the first 16 days of August, Kalshi made $25 million in parlay fees alone.
To quickly bring you up to speed, a parlay is essentially just a combined bet where in order to win, every individual bet in that basket must come true.
The reason anyone bothers to do this at all is that the theoretical payoff tends to be multiplicative, so you only need a relatively small bet size to generate “worth it” level returns.
For instance, combining 5, 50% likely bets into 1 parlay results in a 32x payoff.
Naturally, it’s an extremely low-probability endeavor and is why it makes up the majority of sportsbooks’ profit.
I was late to the game on this one, but after seeing the above headline, I instantly remembered that on prediction markets, there is no “house” that creates and facilitates these bets.
So if these markets were truly P2P, any trader would ask the same question I did:
“Lots of people are making these wagers that won’t pay off, couldn’t I just take the other side?”
After all, in contrast to selling low-probability out-of-the-money options, there would be a virtually infinite amount of combinations users can trade, which should help to mitigate the “pennies in front of a steamroller” effect through diversification.
So, with that clear question in mind, I immediately began pulling some data to see how it would go.
It’s no fun being the house
Starting simple, I wanted to model out the easiest version of this trade:
Of every available combo (the term for parlays), find the ones where “YES” was quoted at <= 2c
In other words, finding every parlay with a less than 2% of chance of paying off
From there, we’d simply buy the “NO” side and ideally collect a 2% return
A combo’s NO side pays out if any leg fails, so the idea was to get paid 2% for holding something that only loses when a 6-leg parlay actually pays out.
With that in mind, I pulled the data for these and immediately found that it made no sense: