Open problems
Eleven things nobody has figured out.
Prediction markets are the first mechanism that pays people for telling the truth about the future. Almost everything load-bearing underneath them is still unsolved.
These are the problems I keep running into. None of them are mine. As far as I can find, none of them are solved, and most can be started on a laptop this week with public data.
Each problem has three readings. Simple assumes nothing. Moderate assumes you know what a market is. Technical is the state of the art and the specific gap in it. Every page ends with where I would start and what would count as a result, because a problem you cannot begin on is just an interesting fact.
The order is roughly dependency. A market has to pay for itself before its price is worth trusting. The price has to be trustworthy before anyone can act on it. Later problems assume the earlier ones went well, which they have not.
If you pick one up, tell me. I would rather compare notes than race.
Does it pay for itself
-
01
Every market is on life support.
Every prediction market alive today is paid to be alive. Nobody has worked out the cheapest way to keep one breathing, or where the money should go when you only have a little of it.
-
02
Nobody buys the answer.
These markets produce a probability that beats most experts. There is no customer for it. The fees come from people who want action, so the accuracy is a side effect nobody pays for.
Can you trust the price
-
03
The price moves the thing it prices.
A market says the bank fails. Depositors see it and withdraw. The bank fails. The forecast was correct and it was also the cause, and calibration cannot tell those apart.
-
04
Thousands of traders, one opinion.
Agents are around a third of Polymarket activity and 14 of the top 20 wallets. They are trained on overlapping data, so they never coordinate and they still move as one.
-
05
Most of the volume is decoration.
Wash trading estimates run from 15% to 95% depending on who is counting. Every volume headline, and every study that uses volume as a control, sits on top of that.
-
06
Calibrated on average.
The standard defence of these markets is a portfolio statistic being used as though it described each market. Nobody can tell you whether the one in front of you is trustworthy.
Can you settle it
-
07
The referee holds a position.
Disputes on Polymarket are settled by token vote, and the voters can hold the outcome they are voting on. A $7M and a $237M dispute later, this is still the design.
-
08
The thermometer costs $34,000.
Someone warmed a single weather sensor and swung a Polymarket outcome. That figure is a load limit on the market, and no platform publishes theirs.
Can you act on it
The frontier
-
10
A market on how wrong the market is.
Layer one prices the event. Layer two prices whether layer one is right. Layer three prices layer two. Converge to truth, oscillate forever, or collapse into noise, and nobody has built one to see.
-
11
Nobody can price a joint distribution.
Science does not need probabilities on isolated claims. It needs "X works given Y and Z". The outcome space explodes and every known market maker goes bankrupt or intractable.
Working on one
I am reachable on X and Telegram, and I answer. If you get a result, even a negative one, it is worth publishing. Most of these have no public attempt at all, so a failed attempt with a clear writeup moves the field further than another survey.