I built a system that tracks every Polymarket market, records price snapshots every few hours, and stores it all in a SQLite database. After 24 days of collection (March 18 - April 11, 2026), I have 8,050 markets and 6.6 million price records covering $4.98 billion in total volume.
Here's what the data says about crashes, recoveries, and where the real edge is.
The Dataset
| Metric | Value |
|---|---|
| Markets tracked | 8,050 |
| Price records collected | 6,619,038 |
| Total volume tracked | $4.98B |
| Avg liquidity per active market | $192,609 |
| Categories covered | 9 (sports, crypto, politics, geopolitics, economics, science/tech, weather, entertainment, other) |
| Collection period | March 18 - April 11, 2026 |
The most actively tracked markets have 3,300+ price snapshots each -- that's a price update roughly every 10 minutes for the busiest ones. Markets like "Will Iran hold a presidential election by June 30?" and "Ukraine signs peace deal with Russia before 2027?" sit at the top with 3,326 snapshots each.
How Markets Move: The Distribution
I bucketed every market by its single-day price change. Out of 3,145 markets with measurable day-over-day data:
| Movement Bucket | Count | % of Total |
|---|---|---|
| Slight dip (0% to -5%) | 1,009 | 32.1% |
| Slight up (0% to +5%) | 815 | 25.9% |
| Large rally (+30%+) | 508 | 16.2% |
| Crash -30% to -50% | 258 | 8.2% |
| Crash -50%+ | 196 | 6.2% |
| Crash -15% to -30% | 108 | 3.4% |
| Moderate up (+5% to +15%) | 97 | 3.1% |
| Moderate drop (-5% to -15%) | 86 | 2.7% |
| Strong up (+15% to +30%) | 68 | 2.2% |
The fat tail is real: 17.8% of markets experienced a crash of 15% or more in a single day. That's 562 markets. But the upside tail is fatter -- 16.2% rallied 30%+ in a day. Prediction markets are volatile by nature because they resolve to 0 or 1.
What Happens After a Crash
This is the question that matters for traders. I looked at all 562 markets that crashed 15%+ and checked where they ended up.
| Post-Crash Price Level | Count | % of Crashed Markets |
|---|---|---|
| Dead (1c or less) | 512 | 91.1% |
| Low (2c - 10c) | 36 | 6.4% |
| Partial (11c - 30c) | 3 | 0.5% |
| Moderate (31c - 60c) | 6 | 1.1% |
| Strong (61c - 90c) | 2 | 0.4% |
| Full recovery (90c+) | 3 | 0.5% |
91.1% of markets that crash 15%+ never recover. They go to 1 cent and stay there. The crash was right -- the event didn't happen or the team lost.
But 8.9% show some life after the crash. And 2.5% recover to 10 cents or above. The 3 that fully recovered to 99c represent genuine mispriced crashes where the market overreacted and corrected.
The Recovery Stories
The markets that crash and recover are the interesting ones. These are the highest-volume markets that dropped 15%+ and bounced back above 50 cents:
| Market | Crash | Current Price | Volume |
|---|---|---|---|
| CDU wins most seats in Rhineland-Palatinate elections | -15.5% | 99.9c | $642K |
| Thunder vs. Lakers O/U 222.5 | -51.0% | 63.9c | $268K |
| OpenAI IPO market cap above $1.2T | -16.0% | 65.0c | $240K |
| Magic vs. Mavericks O/U 239.5 | -49.5% | 99.9c | $230K |
| 76ers spread (-4.5) | -52.5% | 99.9c | $128K |
Pattern: sports over/unders and spreads crash mid-game when momentum shifts, then recover when the final score lands. Political markets crash on headlines, then correct when fundamentals reassert. Crypto price markets are the most permanent crashes -- when Bitcoin misses a target, it's done.
Crash Rate by Category
Not all categories crash equally:
| Category | Crashes (15%+) | Crash Rate |
|---|---|---|
| Other (sports, esports) | 457 | 10.7% |
| Weather | 8 | 7.8% |
| Science/Tech | 9 | 6.0% |
| Crypto | 35 | 3.0% |
| Sports | 33 | 3.1% |
| Geopolitics | 6 | 1.7% |
| Economics | 2 | 1.6% |
| Politics | 11 | 1.4% |
| Entertainment | 1 | 1.3% |
Sports and esports markets (categorized as "other") crash the most -- 10.7% of them see a 15%+ single-day drop. These are live-game markets where score swings cause rapid price movement. Weather markets come next at 7.8%. Politics and economics are the most stable, crashing under 2% of the time.
The Crash Monitor Strategy
Knowing that 91.1% of crashes are permanent isn't the whole story. The question is: can you filter for the 8.9% that recover?
I built a crash monitor that applies filters before buying the dip:
- Skip resolved/expiring markets -- if it's settling in hours, the crash is final
- Volume threshold -- low-volume crashes are illiquid traps
- Category awareness -- crypto price target misses don't recover; sports lines do
- Price level checks -- buying at 2c has different risk than buying at 40c
Paper trading this selective approach: 86.5% win rate across 84 trades, $74.35 paper P&L. The strategy doesn't buy every crash -- it filters for the pattern of temporary overreaction in liquid markets.
The Worst Single-Day Crashes
For context, the biggest single-day drops in the dataset:
| Market | Drop | Volume |
|---|---|---|
| Al Hilal Saudi Club win (Apr 4) | -81.5% | $123K |
| Warriors vs. Kings | -81.5% | $2.85M |
| Benfica win (Apr 6) | -79.5% | $246K |
| Weibo Gaming vs NiP (LoL) | -77.5% | $944K |
| ETH above $2,100 on April 7 | -77.0% | $153K |
These are resolution crashes. The game ended, the price target was missed, the event didn't happen. No amount of "buying the dip" saves you here. The crash monitor's job is to distinguish these permanent crashes from the temporary ones.
What This Means for Traders
- Most crashes are correct. 91% of 15%+ drops go to zero. Don't buy every dip.
- Category matters. Sports line markets (spreads, over/unders) have the highest crash rate but also the highest recovery rate -- the crash happens mid-game, not post-resolution.
- Volume is signal. High-volume crashes in non-resolved markets are more likely to be overreactions. Low-volume crashes in niche markets tend to be permanent.
- Speed matters. The recovery window is narrow. If a market is going to bounce, it usually happens within hours, not days.
- Selective strategies work. Buying every crash loses money. Filtering for specific patterns yields 86.5% win rate in paper trading.
Explore the Data
All of this analysis comes from a live database that updates every few hours. I built PolyScope to make this data browsable -- you can see every market, its price history, volume, and category breakdown.
Explore the data yourself on PolyScope -- free. No signup required.
Data collected from Polymarket's public API. 8,050 markets, 6.6M price records, March 18 - April 11 2026. Analysis performed with SQLite queries against the raw dataset. Paper trading results are not indicative of future performance.
Top comments (0)