Practice the skill nobody teaches.
Pick a resolved market. Watch the price unfold hour by hour with the original news. Back your view with simulated capital, never your real wallet.
Three steps. No jargon.
Pick a resolved case
Curated moments from the last three years: the 2024 election, the FTX collapse, the first Bitcoin ETF.
Watch it unfold, live
The price ticks forward hour by hour. The original news flows past. Take a position, and the future stays hidden until you reach it.
Read the verdict
A debrief on every replay: where you anticipated, where you chased. After ten cases, your fingerprint emerges.
Start anywhere. Each one is a lesson.
The 2024 United States Presidential Election
On the night of November 5, 2024, the market opened at 52% Trump. By 3 AM Eastern, it had moved to 95%. The night did not move in a straight line. There were three distinct repricings — the Florida call, the early Pennsylvania returns, and the AP's Wisconsin call near midnight — and the market processed each one differently. Some users bought ahead of the moves and rode the tape up. Most bought after the calls, when the information was already in the price. This case is the canonical test of a forecaster's discipline: do you anticipate, or do you chase? Both can be skills. The night will tell you which one you are.
Play this caseFifteen Days: Biden Drops Out, Harris Becomes the Nominee
Between June 27 and July 21, 2024, the Polymarket 'Biden withdraws before convention' market climbed from 14% to 96%. The catalyst was a single debate performance on the 27th — but the market did not move in a straight line. It surged the night of the debate, then drifted back as the White House denied, drifted up again on private donor pushback, drifted back on a Stephanopoulos interview that was supposed to settle it, and finally went vertical on July 21 when Biden's letter posted to X at 1:46 PM ET. This case is about reading a slow-burn information cascade: every news cycle either confirmed the trajectory or briefly looked like it broke. The skill is staying with the trend when each individual data point feels like noise.
Play this caseThe FTX Collapse
On November 2, 2022, CoinDesk published a story about Alameda Research's balance sheet that read, on the surface, like a curiosity for forensic accountants. By the time the article was a week old, the second-largest cryptocurrency exchange in the world had filed for bankruptcy and a handful of people who'd done the work in the first 48 hours had made fortunes. This case is the canonical study of how prediction markets process a slow-burning crisis. The early hours were not panic. The information was in the open — the leverage, the related-party exposure, the absurd valuations — but the price moved gradually as the implication compounded. Then Binance walked away from the acquisition, and the chart went vertical. This case teaches what 'information was in the open' actually means: it usually means a few people read the article carefully and most did not.
Play this caseSuper Bowl LVIII: Chiefs at 49ers, Overtime
Super Bowl LVIII was a study in real-time probability. The 49ers led by 10 with five minutes left in the third quarter. By the end of regulation, the score was tied. In overtime, Patrick Mahomes drove eighty yards for the winning touchdown. The Polymarket win-probability market moved through six distinct regimes that afternoon — opening favorable to San Francisco, neutral by halftime, KC-favorable in the third, SF-favorable late in the fourth, tied at overtime, then sharply KC. Each transition was an information event. This case teaches the discipline of fast-information markets: there are no slow burns here, only sharp updates as plays land. Sports replays are the cleanest training environment in Replay — the information is unambiguous, the price reaction is instant, and you find out quickly whether you read the game well or chased the action.
Play this caseThe Five Days Sam Altman Was Fired
On Friday afternoon, November 17, 2023, OpenAI's board fired Sam Altman in a four-paragraph press release. By Sunday night, ninety percent of OpenAI's employees had signed a letter threatening to resign and follow Altman to Microsoft. By Wednesday morning, Altman was back at OpenAI with a new board. The Polymarket market on 'Will Sam Altman return as OpenAI CEO by year-end?' opened Friday afternoon at 18%, dropped to 8% on Saturday as he announced he was joining Microsoft, then climbed in a single hour Sunday night to 87% when the employee letter dropped. This case is what a corporate-governance crisis looks like in real time on a prediction market — the price moved not on official statements but on signals the market believed were the actual information.
Play this caseWorld Cup Final 2022: Argentina vs. France
By the 80th minute, Argentina led France 2-0 and the prediction market had Argentina at 97%. Then Kylian Mbappé scored. Ninety seconds later he scored again. The match was tied. The market that had been 97% Argentina was 51% by the end of regulation. Extra time produced a Messi goal, then a Mbappé hat-trick from the penalty spot. Penalty kicks resolved it. The Argentine win market closed the night at 100%. This is the cleanest fast-information replay in the suite — every goal is unambiguous, the price reaction is instant, and the lesson lands hard: a 97% market is not a sure thing. Three percent is three percent.
Play this caseDid you publicly predict any of this at the time?
Replay shows you what the market did. Disagree shows you what you said. If you wrote a tweet, a Substack post, or a blog about a political moment or a crypto event, Disagree finds the quote and lets you size the conviction you already published.
The full library + debrief on every replay + signed Forecaster Report.
Pro is rolling out by waitlist. Until the payments rail lands, every case stays free.
Know what you believe. Find who agrees.
Replay shows you your calibration strengths. Pools finds people whose worldview matches yours in those categories. It stays non-custodial: each member's capital stays in their own Polymarket Safe.
Practiced the case. Now watch the real thing resolve.
Cal overlays live market probabilities on your Google Calendar events. When a market you practiced in Replay has a real resolution date coming up, it appears on your calendar, priced.
Replay sessions use simulated funds against historical Polymarket price data. No real USDC moves during a replay. Your simulated calibration is a training tool, not a prediction of how you'll do with real money. People perform worse with real money than in simulation, and that's normal. Naming it honestly is the point.