Executive summary: stop calling them "assassination markets" and start calling them "reflexive prediction markets" because they're not bad, they just had a bad debut. We might use these markets to answer very complex questions that we otherwise wouldn't have.

Prediction Markets are now a thing

Prediction markets are now big enough that they can generally be used to predict things (market as a mirror).

But I think one of the most overlooked aspects of these types of markets, is when they can be used to coordinate and find truths about very complex systems (market as lever).

Reflexivity

George Soros calls it reflexivity: the price of a thing feeds back into the fundamentals of the thing.

Most finance pretends this doesn't happen, but once you look it's everywhere:

• A bank run is a prediction ("this bank will fail") that causes its own resolution.
• A widening CDS spread raises a company's borrowing costs, which pushes it toward the default the spread was pricing.
• And in 1992 Soros didn't forecast that the pound would fall out of the ERM, his position was large enough to be part of the cause. The trade settled itself.

Axis of Reflexivity

I think most markets fall somewhere on this axis.

• On one end: a market on tomorrow's weather, which no amount of trading can budge.
• On the other end: a market whose payout is, itself, the incentive that makes the thing happen.

We've tried this before

We've tried to make reflexive prediction markets "a thing" before, but we were so fast to jump to conclusions that we missed the big picture.

Jim Bell's Assassination Politics

In 1995, Jim Bell wrote an essay called "Assassination Politics." The mechanism was an anonymous pool where people contribute money against a public figure's name, and whoever correctly "predicts" the date of that person's death collects the pot. The scare quotes are the whole point.

The person best positioned to predict the date is the person who picks it. The market doesn't forecast the assassination. It funds it. Link to mirrored essay: https://files.omarish.com/assassination-politics.pdf

PAM - DARPA's Policy Analysis Market

Eight years later the idea's respectable cousin died in public as PAM - the Policy Analysis Market.

Robin Hanson, the economist behind futarchy and much of modern prediction market theory, architected DARPA's Policy Analysis Market, which would have let traders price geopolitical instability in the Middle East. Two senators held a press conference, called it a "terrorism futures market," and it was dead within about a day. Nobody asked whether the mechanism worked.

Bad First Impression != Bad Idea

If ARPANET (another DARPA project) had a "day-two press conference" you wouldn't be on the internet right now.

Also just wanted to call out Jim and Pam.

But what if this is something we want?

"A drug for pancreatic cancer passes Phase 3 by 2032."

"A working desalination process under $0.30 per cubic meter."

"3 new CVSS with severity ≥ 9.0 in OpenSSH by December 31, 2026"

Now walk through who shows up, because the roles are stranger and better than they first appear:

The people who can cause the outcome buy YES

• A biotech team that believes it can hit the milestone buys YES at 15 cents and earns the rest of the dollar by doing the thing.
• Notice this is insider trading, and notice that it's the entire point.

In a reflexive market, trading on your ability to cause the event isn't a crime, it's the whole point. The market pays for private information the only way that matters: by paying for the outcome itself.

The people who want the outcome buy NO

This is the counterintuitive part. If you desperately want the cure, you take the position that pays out if it doesn't happen. If the cure arrives, you happily lose your stake — and your money flows directly to whoever caused it. If it doesn't arrive, you're compensated for living in the worse world. A NO position in a good-outcome market is simultaneously a bounty and an insurance policy. Buying NO is posting the prize.

Speculators price the middle

And the live price becomes a public signal no committee can fake: the market's running estimate of whether humanity is on track to get this done. A price ticking up from 10 to 40 cents is news.

Goodhart's Law: the hard part is the finish line, not the money

Goodhart's law says the moment you define the settling event, you've created an incentive to hit the definition rather than the outcome.

Pay out on "audit published" and you'll get thin audits. Pay out on "Phase 3 passed" and you've inherited every incentive problem the FDA already polices, plus new ones.

So the entire design discipline lives in settlement: the event must be specified so that legitimately causing it is the cheapest path to the payout — cheaper than faking it, gaming the oracle, or hitting a perverse nearby target.

This is hard. It is also exactly the kind of hard that's gotten easier: verification, attestation, and adjudication are precisely what's collapsing in cost right now. The oracle problem hasn't disappeared, but it's no longer obviously the binding constraint.

Aim the mirror

Prediction markets earned their reputation as mirrors, and the mirror version is genuinely useful — I want election forecasts and Fed-decision odds as much as anyone. But the mirror framing quietly concedes the most powerful property of the instrument, the one that scared everyone off in 1995 and again in 2003. Markets move the world. They always have. Reflexivity isn't a bug to be minimized or a horror to be banned. it's a lever lying on the ground.

The question was never whether markets change the outcomes they price. The question is whether we keep pretending they don't, or start choosing the outcomes on purpose.

On-Chain Funding Mechanics

OK, this is where I need your help. When I think the actual market mechanics, how would we design this on-chain as a perps market?