We Found Another Edge in Prediction Markets

There’s apparently serious money in hoping the President says “tariff” and not “tariffs.”

In our last post on prediction markets, we took a look at Polymarket and couldn’t wrap our heads around how so many accounts had equity curves that were virtually straight lines up.

We cracked the books, asked around, and did everything else we could to find out.

In sum, it wasn’t predicting.

We found that the accounts were almost all trading the 5-minute crypto up/down markets, running a fair value calculation against a live BTC feed, and picking off stale retail orders on a slow CLOB.

Pretty unsexy, but extremely profitable.

Naturally, this sparked a fire in us to find out who else was making a killing on prediction markets, and exactly how.

It wasn’t long before we found out where.

Today, we’ll be taking another deep dive into prediction markets, but in a corner that not many people know about, with one of the highest ceilings for edge we've seen.

By the end of this, you’ll either be extremely motivated to go after some of these yourself, or intimidated by how seriously people build operations around this stuff.

So, without further ado, let’s get right into it.

What’d They Say Again?

If you’ve spent any time on Polymarket or Kalshi during a major political speech, you've probably seen these and immediately written them off:

Mention Markets

As the name implies, these are essentially markets that resolve based on whether or not a person says a specific phrase during some pre-defined event.

Presidential addresses, earnings calls, livestreams, anything:

Naturally, our first thought was a normal one:

“How could anyone systematically know what someone is going to say?”

Sure, maybe you have insider edge and spoke to the person in question, maybe you built a historical dataset of past talks from that person to get a fair value; but either way, how can you really know something like that?

Well, as we’ve learned time and time again, it all comes down to speed and understanding niche market mechanisms.

Be First, Be Smarter, or Cheat

While we wouldn’t say the average counterparty here is as under-developed as those in the BTC up or down markets, they’re often under-tooled.

A meaningful share of the volume in these contracts is essentially narrative-based. People watch the speech, hear the political coverage, decide Trump is definitely going to mention tariffs because it’s been in the news all week, and place a directional bet hours before the event.

A different share is moved by what can be dubbed as “vibes positioning”. The contract for “tariff” is priced at 95 cents because everyone knows tariffs are the topic. Someone notices that “border” is only at 30 cents and decides that’s a better bet, without any particular reason to think so.

Some popular accounts also run livestreams of their activity and the typical infrastructure looks like this:

In the above demonstration, the trader was making a bet on whether or not the speaker would say “autism”. While watching the livestream, he placed a “yes” bet when the context of the conversation steered in that direction, which was ultimately followed by the speaker saying the target word.

It's not the most advanced setup, but taking the trade around the immediate context was actually pretty smart and the average price of the trade was ~12 cents, resulting in roughly a 9x return on capital, so, not too shabby.

However, there are a few structural problems with this approach.

The biggest problem is the inherent lag of streaming through a third-party service. For most live broadcasts, especially politically sensitive ones, there’s often a deliberate delay before any video reaches the base CDN server.

It’s often a safety mechanism for if something happens that shouldn’t be broadcast (e.g., an act of violence).

Now, mention markets tend to be rather illiquid, so realistically, most participants in that market are using the same livestream to base trades on, making the above approach not totally flawed.

However, not all mention markets are as illiquid. When there’s a major event like a Fed conference that draws $1m+ in volume, retail gets wiped and things get serious.

To see how, we have to introduce you to a little something known as the Bloomberg Squawk Box:

When there’s a market-moving event being broadcast live, the order of who gets the feed first is well established:

1. A person at the venue with a microphone-grade audio feed hears it first.

2. A Bloomberg terminal squawk gets it next, usually within a second.

3. The TV broadcast is anywhere from 4 to 12 seconds behind. Closed captioning on that broadcast adds another 6 to 15 seconds.

4. A major newswire transcript landing in someone’s inbox might be 30 to 60 seconds later.

5. The retail trader refreshing CNN’s website is somewhere in the back of all of this.

For small mention markets like this one where you can maybe put down $500 max, the competition is fairer. When it’s actually market moving, you can be certain that anyone at home with a Bloomberg terminal is doing exactly this.

It isn't a total game over though; a terminal can technically be purchased by anyone for $30,000/yr and there's still edge in the smaller markets where everyone's on the same feed and you can be smart about context. For the larger markets however, you want your feed to be genuinely the fastest possible.

Fortunately, if you aren’t interested in shelling out a down payment for a niche market that might not even exist tomorrow, there is still another way.

Just Read The Rules