These Are the Sharps Actually Making Money on Prediction Markets
Hosted by Joe Weisenthal and Tracy Alloway
- 48m
- Transcript
Summary
Written by a model from the transcript; not yet reviewed.
This episode of Odd Lots features the reporter Adam, who wrote a New York Times article on prediction market traders consistently beating the market, along with two members of the 'MAGA Kiwi Club' Discord group: Brian Golden, who specializes in inflation contracts, and Daniel Reichman (Carnitas Taco), who focuses on elections and politics. The conversation explores how these 'sharps' build their edges through original research, community information sharing, and a willingness to change their minds based on data.
Brian describes rebuilding the BLS inflation formula in Excel over three months, noting his average absolute error over two years beats the Bloomberg consensus despite having only an undergraduate degree in drama. Daniel walks through the New Jersey governor's race, where their group's napkin model predicted a 12-14 point margin while polls showed a close race. The group also discusses the Romanian election as a 'dark day' when local Romanian bettors correctly identified the frontrunner's collapse.
The episode touches on market efficiency, the reliability of prediction market prices versus polls, insider trading (with a specific example of Jacob Elordi's Critics' Choice Award going from 1 to 40 cents on massive volume), and why AI tools like ChatGPT are unreliable for trading edges because they confirm whatever the user wants to hear.
- Brian rebuilt the BLS inflation formula in Excel over three months and his average absolute error over two years beats the Bloomberg consensus, suggesting institutional forecasters are not trying hard enough on data that has already occurred. 20:22
- The comment sections on prediction markets are a reliable contrarian signal because sharps keep quiet while squares advocate for their side, making the more-commented side typically the wrong one. 30:34
- LLMs like ChatGPT will confirm whatever the user wants to hear, making them unreliable for trading edges; the real edge comes from original data gathering, on-the-ground polling, and human judgment. 45:24
- The Romanian election showed that even experienced sharps can be wrong when local knowledge (Romanians betting against the frontrunner who became a laughingstock) beats international analysis built on historical patterns. 33:50
- Prediction markets are zero-sum and most participants lose; the sharps in the MAGA Kiwi Club Discord group consistently beat the market through specialized work, community information sharing, and an Avengers-style division of expertise. 7:20
Highlights
- 6:20
How Adam Found the Sharpsstory
Adam explains he was losing money in prediction markets and wondered who he was losing to. He started asking around, and one person led to the next until he found a Discord full of people consistently crushing it.
- 12:40
Siloed Media Distorts Election Pricesinsight
Brian argues elections will always be the most mispriced markets because people live in siloed media environments. He cites Spencer Pratt's LA mayoral price spiking to 27% despite LA being a Democrat plus 42 city.
- 20:22
Rebuilding the BLS Formula in Excelstrategy
Brian, a theater kid from the Midwest with an undergraduate degree in drama, spent three months reverse-engineering the BLS inflation formula in Excel. He says it says more about the softness of institutional forecasters than about his own skill.
- 27:59
New Jersey Governor's Race Modeldata
Daniel describes how a simple napkin model in their Discord predicted a 12-14 point margin for the New Jersey governor's race three months before polls came in showing a close race. The prediction market prices were higher than polls but still below the actual result.
- 30:34
Bet Against the Comment Sectioninsight
Brian states as a guiding principle: always bet against the comment section. Sharps keep their mouths shut except when talking to each other, so the side with more advocacy is typically the wrong side.
- 33:50
The Romanian Election Dark Daystory
Daniel recounts how the group put significant money on Simeon winning the Romanian runoff based on historical patterns, but he left the country, skipped debates, and became a laughingstock in Romanian media they had not picked up on. Local Romanians bet against them and were right.
- 38:55
Jacob Elordi Insider Trading Examplestory
Brian describes a specific market where Jacob Elordi's Critics' Choice Award odds went from 1 cent to 40 cents on massive volume the night before, which he is certain was an insider trade. He lost that bet and says the price movement pattern is recognizable.
- 45:24
AI Is Square Moneyinsight
Daniel says people who just ask LLMs questions and think that gives them an edge are some of the squarest money out there. Brian adds that ChatGPT will confirm whatever you want to hear, shifting its answer each time you push back.
About this episode
Here's a couple things about prediction markets. A lot of it is pure gambling and speculation, much of it on things with very little economic relevance. Another fact is that in all likelihood, if you yourself started trading right now, you'd probably lose your shirt. But there is money being made by some dedicated traders, really focused on areas like politics and economics. On this episode, we speak with Brian Golden and Daniel Reichman, who are part of a private Discord called Maga Kiwi Club, where serious prediction markets traders swap ideas and make real money. We discuss the remarkable efforts they go to in order to spot opportunities, the systematic biases among traders, how they feel about insider trading, and other major issues that surround the space. Alongside Brian and Daniel, we also speak with NYC-based journalist and producer Adam Iscoe, who recently profiled these traders https://www.nytimes.com/2026/05/26/magazine/polymarket-prediction-wall-street.html for The New York Times Magazine. Only Bloomberg.com subscribers can get the Odd Lots newsletter in their inbox, plus unlimited access to the site and app. Sign up at bloomberg.com/subscriptions/oddlots https://www.bloomberg.com/subscriptions/oddlots?in_source=oddlotspodcast See omnystudio.com/listener https://omnystudio.com/listener for privacy information.
Transcript
9,685 words · 8 speakers
Speaker names marked are inferred from speaking time and the show’s known hosts — they are not voice-verified.
Topics
- Prediction Markets
- Market Efficiency
- Inflation Forecasting
- Election Trading
- Insider Trading
- AI Limitations
- Behavioral Finance
Page updated · transcript and summary are machine-generated