So-Called "Bonds" in Prediction Markets

Rare events teach slowly

Disclosure: I run Kalshinomics.com, which may earn Kalshi referral fees. I may trade event contracts on Kalshi and securities on other platforms. Readers should consider this relationship when evaluating my analysis. For educational purposes only, not investment advice.

A word I hear tossed around recently in prediction market twitter is “bonding” a contract, meaning going long a contract at close to 100%, say between 80-99%, and holding through expiration. Rather than 10x your money on a long-shot the idea is you slowly win over time, just like buying a treasury or corporate bond. But these aren’t your grandfather’s savings bonds. The fundamental question when trading is “do I have edge?” And because high priced contracts leave a small margin for error, it takes a long time to measure it. The key is to have a clear reason your trade has edge in the first place, and to investigate all the ways you might be wrong.

To some, this looks like simply betting on the favorite. However there’s a subtle assumption being made by the prediction market crowd - not that these are riskless bonds, but that in general long shots are overpriced, and buying the NO contract on most events that are favorites should make money on average. I’ve seen it called the “nothing ever happens” trade. This is supported by papers like this one which showed low priced YES contracts to be systematically overpriced, and high priced contracts underpriced. And a great Dune dashboard by Alex McCullough showing a similar pattern. As a heuristic, “people like long-shot bets like parlays or lottery tickets” fits common sense.

But that doesn’t imply you should spray your bankroll across “NO” contracts, or that this back-tested edge will last forever. Past studies that long shots were overpriced doesn’t necessarily imply they are today, and adverse selection on fill size matters. Picture 10 markets you think are really ~100%, and they’re all offered at 98%. You’re not the only trader trying to do this, you might get 2000 shares on 9 of the markets, and the last one you get filled on 10k. The size remaining on a limit order book is not random.

No amount of art/craft bet sizing will help you if your engine isn’t a stream of +EV bets. The problem: as I wrote about here, even with a large edge in a 50/50 situation you need many trials to know you’re a winner. In my previous example I looked at having a 55/45 edge, in reality as you get closer to the tails of course your edge is going to shrink related to the remaining uncertainty. (It looks a little like Kalshi’s fee structure! not random either)

So let’s say you’re buying a 96% contract with true odds 98%. Chugging through the math, to get that similar false positive rate of 5% and power of 80% from my past example, we need 485 trials. The key point is not the exact math, it’s that you’re going to win again and again, and believe you’re making good bets, and when you finally lose, your lizard brain will attribute that loss to a mistake or fluke. Your win rate will feel nearly the same whether you’re buying 96% with fair odds of 98% or 94%.

The only thing that’s going to help you dial in your edge - is either a ton of data confirming your win rate, or a couple losses early on that cause you to pause. You learn very little from each incremental win but a great deal every time you lose.

An actual bond I put on

But some trades really are like bonds! I was looking at the Kalshi markets on 11/18/24, the election had already been called for Trump. But the market for who would be inaugurated as next President on Kalshi was 97 @ 98 for “Donald Trump or another representative of the Republican party” (emphasis mine). It’s important the “other Republican” bit is in there, for anyone at that age, mortality is a non-zero risk over time. This contract would expire on inauguration day (1/20/2025), so 63 days in the future.

I thought the actual fair value was closer to 100% than 98%, with the election being called it would take a major crisis to change the outcome. Not impossible, but would take a monster tail event to hit. I didn’t have a specific scenario in mind but let’s be conservative and call it 1 in 200 we hit something crazy that changes the outcome. So I’m buying 98% with my fair 99.5%.

In some sense this is like a risky bond, it’s going to return 2 cents on an investment of 98 cents over 63 days → about 12.4% annualized, plus Kalshi’s ~4% interest on the position, and has a default rate of 0.5% and recovery rate of 0.

But why was I getting this trade? The most valuable thing I could possibly get here is a glimpse into the mind of my counterparty. One hypothesis was: “a bunch of people just made $ betting on Trump, and now they want to close instead of holding for two months to get a tiny return.” But every “yes” contract creates a “no” contract, why would the folks who lost $ on Harris not want to also close at the same time, especially given they could lock in 2 cents vs a fair of near zero? Maybe winners are more likely to close early than losers who are mentally writing off the losses? That would be my ideal scenario - I’m not trading against someone who has the intent of making edge on that specific transaction, they are trading with another motivation.

In equity markets there is a somewhat similar parallel of long stock holders in a merger-arb situation selling prior to a cash takeover. They wanted to own a risky stock - now they own an asset with a bond-like return, and they are willing to give up a little edge to sell to a merger-arb trader.

The worst thing for my trade would be credible paths of election fraud or extreme political upheaval. I did not see any of these signs.

Compared to my standard menu of diversified equities and bonds - this was a very attractive yield and I allocated a % of my portfolio into it.

But the truth is: I will never know if this was a good trade. Maybe the true win rate was 95%, and I did a bad trade and got lucky. That’s an uncomfortable idea that it’s important to get comfortable with. This applies to most things in life - you don’t get enough reps on choosing a college, job, or partner to ever really know. You make the best decision you can and move on. If you were my counterparty and still like your side of the trade I’d love to hear from you.

Closing:

The trading world loves to find new ways to sell tail events. It is easy to confuse a high chance of winning with actual positive expectancy. It is very difficult to distinguish between a 95% bet and a 98% bet. This is the skill you’re implicitly taking on if you pursue this path. It’s not impossible, but you need a strong fundamental reason to believe you have edge to make the trade. Frame it more as “why is this trade available to me?”