The Biggest Reversal We Found in Prediction-Market History Wasn't the 2024 Election

Prediction markets are supposed to be a real-time gauge of what people actually believe, backed by money instead of opinion. The interesting moments aren't when a market confirms what everyone already expected — they're when it reverses, when the crowd changes its mind hard and fast. We pulled full price history for the biggest political and civic reversals we could find on Polymarket and ranked them by how far they swung. The biggest one surprised us.

The biggest reversals we found, ranked

An outsider's 92-point run: the biggest swing we found

Zohran Mamdani wins NYC mayor9.5% → 95.8%

Apr 23 '25

9.5%

May 24

3.5%

Jun 22

18.6%

Jul 12

72.9%

Aug 21

83.8%

Oct 20

92.3%

Nov 5

95.8%

Before he became one of the most closely watched races of 2025, Zohran Mamdani was priced as exactly what conventional wisdom said he was: a long shot. His odds bottomed at 3.5% on May 24 — practically written off — and were still just 18.6% on June 22, two days before the Democratic primary. Then the market didn't just adjust, it flipped: within the next three weeks, by July 12 (18 days after his stunning primary win over Andrew Cuomo), his odds were at 72.9%. From there it was a steady climb to a near-lock 95.8% by election night. A 92.2-point range is the widest of any market in this dataset — wider than either major party's candidate in the presidential race that same year.

The 2024 election: the market didn't just react, it predicted

Trump50% → 58.6%Biden50% → 0.1%Harris3.1% → 41.4%

Jan 5

50%

50%

3.1%

Jun 26

59.5%

33.5%

1.3%

Jul 13

59.5%

15.5%

14.5%

Jul 21

64%

7%

18.2%

Aug 5

51.5%

0.1%

44.6%

Nov 5

58.6%

0.1%

41.4%

The headline story is real: Biden opened 2024 tied with Trump at 50%, and by the day after his July 21 withdrawal his odds were essentially zero (0.1%) — a 49.9-point collapse. But the more interesting detail is in the timing. Harris's odds didn't wait for the announcement. By July 20 — the day before Biden withdrew — she was already at 18.6%, up from 1.3% before the June 27 debate that started the slide. The market had priced in a meaningful chance of a switch before it happened, then repriced sharply once it was confirmed, climbing to a peak near 54% in late August before settling into the low 40s through election day.

Trump's own swing is smaller in range (30.0 points) but has its own inflection: his odds jumped from 60.0% to 67.5% in the two days around the July 13 assassination attempt, before drifting back down as the Harris campaign consolidated support through August.

The market can be wrong right up to the finish line

Not every reversal is a mid-race prediction getting confirmed. One market — "will Kamala Harris win the popular vote" — sat at 72.7% on election day itself, November 5, 2024. A week later, it had crashed to 0.1%.

*January's 50% reflects thin early liquidity on a compound bet (Biden drops out and Harris becomes nominee and she wins the popular vote), not a real forecast.

This is the one reversal in our dataset that isn't a story about the market getting ahead of the news — it's a story about the market being confidently wrong until the votes were actually counted. Trump won not just the Electoral College in 2024 but the national popular vote too, breaking from the split-verdict pattern of 2000 and 2016. Right up to election day, bettors gave Harris better than 7-in-10 odds on the popular vote specifically — the single sharpest one-week move in this entire dataset happened after polls closed, not before.

What the world actually bets the most on

Reversals are about volatility, not size, and the two lists only partly overlap. The 2024 presidential race is both the #2 market by volume and home to two of our biggest reversals — but most of the rest of the volume leaderboard is sports markets that never needed the kind of dramatic repricing this piece is about:

Ten events, $15.30 billion combined — and the single biggest one, the World Cup, isn't in our reversal leaderboard at all. Part of that is structural: "World Cup Winner" volume is the sum of dozens of separate per-country markets (Argentina's alone, the single largest, is $174M — a sliver of the $4.33B total), not one market that either holds steady or swings. Our reversal leaderboard tracks single outcomes, so the two lists are measuring genuinely different things: total money wagered on an entire event, versus how far any one market's own price moved.

What this shows, and what it doesn't

What holds up: the specific swing magnitudes and dates above are drawn from Polymarket's own historical price feed, not estimated or reconstructed — every number is a real traded price at that timestamp. The relative ranking (Mamdani's swing exceeding either 2024 presidential candidate's) held up when we checked it against the raw data twice.

What limits it:

• This surveys a curated set of 5 markets, not an exhaustive search. We picked outcome markets tied to publicly known major-reversal stories (the 2024 presidential candidates, the 2024 popular-vote market, the 2025 NYC mayoral race) rather than systematically scanning every high-volume market for its swing. A larger reversal could exist in a market we didn't check — Polymarket alone hosts thousands of markets, and this covers 5 of them.
• Prediction-market prices reflect what traders are willing to bet, not a scientific poll. Liquidity was thin in some markets' early days (note the flagged January 2024 popular-vote figure above) — a price on low volume is a weaker signal than the same price on high volume.
• "Odds" here means the market's implied probability (the traded price of the "Yes" share), which is a market consensus, not a certified forecast — markets can and did move on rumor, momentum, and incomplete information, not just verified news.
• Polymarket is one prediction-market platform, not the entire prediction-market industry — Kalshi, Metaculus and others sometimes price the same events differently, and we did not cross-reference this specific analysis against them.

Sources

Methodology

For the volume table, we pulled Polymarket's top 15 events by all-time trading volume directly. For the reversal analysis, we did not scan every market for the biggest swing — we selected 5 specific outcome markets tied to publicly known major-reversal stories (each 2024 US presidential candidate, the 2024 popular-vote market, and the 2025 NYC mayoral race) and pulled each one's complete daily price history (fidelity: 1 day, full available range) for the relevant outcome token, then ranked the results by swing. Swing is the simple range: highest recorded price minus lowest recorded price across the market's full history, in percentage points. Checkpoint dates in the charts (debate, assassination attempt, withdrawal, primary win, election day) are drawn from public record and matched to the nearest available price point in the pulled series, not independently verified against Polymarket's own event annotations.

Want to run your own cut? Crawlora's prediction-market data endpoints cover Polymarket event lists, market detail, and token price history (plus Kalshi and Metaculus), and pricing has the free tier.