Della Vedova Prediction Market Indices
Weekly market-structure intelligence for prediction markets. Who trades, who profits, where the information edge sits, and how crowds price uncertainty. Built from every on-chain Polymarket trade since 2022 (640 million and counting).
“Polymarket traders moved $776 million this week, above trend, with 37,569 new participants joining the market.”
- Traders moved $776 million through Polymarket in the week of 2026-04-20, across 30.9 million individual trades, with 37,569 wallets placing their first-ever bet (below the recent trend). Algorithmic wallets handled 91% of the trade count; everyone else split the remaining 9%.
- The most-traded market was "Will the next Prime Minister of Hungary be Péter Magyar?", pulling in $6 million of volume. Politics dominated the week's attention.
- An unusual reversal: retail traders outperformed bots this week. Active retail earned $7K (+84 bps on volume) while algorithms lost $137K (-119 bps). Weeks like this are rare (bots are typically on top by the execution channel).
What happened this week
| Metric | Value | Note |
|---|---|---|
| Total USD volume | $775.8M | Dollars traded across every market this week |
| Total trades | 30.87M | One count per on-chain match (each match has a maker and a taker) |
| Active wallets | 318,595 | Distinct wallets that traded at least once this week (maker or taker) |
| New participants this week | 37,569 | Wallets that placed their first-ever Polymarket trade this week (13-week average: 49,348) |
| Flagged wallets active | 691 | Wallets flagged by the insider-trading test that actually traded this week (of 6,291 flagged in total) |
Who is participating: share by wallet type
| Type | Share of participation | Participations | What that means (examples) |
|---|---|---|---|
| Algorithmic | 91.0% | 56.2M | 50+ trades/day or 1,000+ lifetime. Market-making scripts, arbitrage bots, and other high-frequency algorithms. |
| Active Retail | 7.6% | 4.7M | 10 to 1,000 lifetime trades. Think: the engaged hobbyist who bets on elections, sports, and news across many weeks. |
| Sophisticated | 1.2% | 764K | Big volume, diversified across markets, 30+ day presence. Think: funds and professional discretionary traders. |
| Casual | 0.11% | 67K | Between 2 and 9 lifetime trades. Think: someone who bet on the Super Bowl and one election, then moved on. |
| One-Shot | 0.009% | 5K | A single lifetime trade. Think: placed one bet on a viral market (Taylor Swift engagement, election night) and left. |
Each match has a maker (the limit order) and a taker (the market order filling it); both parties count. Algorithmic wallets are usually the maker; retail and casual traders are usually the taker. Treating only the maker side would attribute ~95% to algorithms and hide the human side entirely. The five-type classification comes from the lifetime trade-history thresholds in our first paper. See the methodology page for details.
Most popular markets by volume
Top 5 resolved-trade markets by USD volume in the week of 2026-W17.
| # | Market | Category | USD volume | Trades |
|---|---|---|---|---|
| 1 | Will the next Prime Minister of Hungary be Péter Magyar? | n/a | $6.4M | 14,666 |
| 2 | Will the next Prime Minister of Hungary be Viktor Orbán? | n/a | $6.4M | 13,231 |
| 3 | Will there be no change in Fed interest rates after the April 2026 meeting? | n/a | $5.1M | 9,257 |
| 4 | Will Roberto Sánchez Palomino win the 2026 Peruvian presidential election? | n/a | $2.4M | 41,401 |
| 5 | Will Rafael López Aliaga win the 2026 Peruvian presidential election? | n/a | $2.4M | 29,550 |
Volume is the sum of usdc_amount across resolved trades for the week.
Market names come from the static polymarket_markets.csv snapshot; new markets created after that snapshot appear with IDs only.
Who profits (algorithmic wallets vs humans)
Every trade’s P&L splits cleanly into directional (did you pick the winning side?) and execution (did you get a better price than the market’s volume-weighted average?). Across 640 million resolved trades, algorithmic wallets capture almost all of the dollars; retail loses on execution even when their picks are right. Source: Della Vedova (2026), “Who Profits from Prediction Markets?”
Resolved trades for the week of 2026-W16 (one week behind the activity panel above; we wait for markets to resolve before booking P&L). Accuracy is the share-weighted fraction of tokens this group held on the winning side. ROI values are in basis points of USD traded.
| Trader type | Trades | Accuracy | Total P&L | Directional (bps) | Execution (bps) | Total ROI (bps) |
|---|---|---|---|---|---|---|
| Algorithmic | 575K | 53.3% | -$137K | -29 | -90 | -119 |
| Active Retail | 25K | 74.8% | $7K | +31 | +53 | +84 |
| Sophisticated | 13K | 52.3% | -$19K | -77 | -54 | -132 |
| Casual | 96 | 64.5% | $1K | +2211 | -301 | +1910 |
| One-Shot | 3 | 94.4% | $151 | +3552 | +650 | +4202 |
Directional + execution = total ROI (an algebraic identity). Casual and one-shot traders can show very large ROIs in any given week because their samples are tiny (often under 100 trades).
Algorithmic vs retail execution edge (this week)
| Metric | Value | Note |
|---|---|---|
| Algorithmic vs retail execution edge | -0.052 ▼▼ | How much better algorithmic wallets are at timing entries vs active retail (z-score vs last year: -1.38) |
Who appears to know something (insider-trading signal)
A statistical test picks out wallets whose win rate is too high, too often, to be luck. A higher flag rate means a stronger fingerprint of private information. This is a statistical signal, not a legal finding about any individual trader. Source: Della Vedova (2026), “Detecting Informed Trading in Prediction Markets.”
| Metric | Value | Note |
|---|---|---|
| Wallets flagged as likely informed | 6,291 | Out of 483,002 we can test (1.30%) |
| After strict statistical correction | 806 | Even under the harshest multiple-testing correction, this many remain |
| After moderate correction | 2,493 | False-discovery-rate correction (balanced approach) |
Flag rate by market type
We group markets by what generates the outcome. Types where somebody could plausibly know in advance (elections, committee decisions) should show higher flag rates than types driven by pure randomness (crypto prices). They do.
| Category | Wallets tested | Flagged (p < 0.01) | Flag rate |
|---|---|---|---|
| Vote | 90,797 | 2,534 | 2.79% |
| Action | 78,548 | 755 | 0.96% |
| Performance | 157,755 | 1,598 | 1.01% |
| Stochastic | 134,972 | 1,613 | 1.20% |
Rates and methodology are public. Identification of specific wallets for regulatory, litigation, or journalistic purposes is handled case-by-case; see consulting.
How crowds price uncertainty (probability weighting)
Tversky and Kahneman (1992) showed that people overweight rare events and underweight common ones. We measure the same distortion week by week using realized win rates from non-bot Polymarket trades. Source: Della Vedova and Grant (2026, in progress), “Probability Weighting from Prediction Markets.”
| Metric | Value | What it says |
|---|---|---|
| How traders weight unlikely outcomes | 0.722 ▼ | 1.0 means rational; below 1 means long shots are over-priced (today vs last year: -0.56) |
| Long-shot pricing gap | -0.022 | How much long shots win minus what the market implies (today vs last year: +0.08) |
| Calibration fit | 0.895 ▼▼ | How well one behavioral parameter explains this week’s pricing (today vs last year: -1.72) |
| Long-run probability weighting (all weeks) | 0.664 | Pooled across every resolved trade since 2022. Kahneman-Tversky experimental value is 0.65. |
Time-series charts and the calibration scatter (Prelec α = 0.664, R² = 0.987) live on the deep-dive pages: calibration, algorithmic share, PWI, execution edge, price gap, efficiency.
One email per week with the updated indices, the “this week on Polymarket” narrative, and any methodology changes. No pitches. No market recommendations. Unsubscribe in one click.
Your email is used only for the weekly update. Not shared. Not sold.
Cite this dashboard
BibTeX entries for each index are available on its deep-dive page. See the methodology, the full historical data, and the source papers.