Top 10 Truths About the Prediction Market: Only 3.14% of the 1.72 million addresses on Polymarket are "true winners"

_Original text from: _Prediction Market Accuracy: Crowd Wisdom or Informed Minority?

Compiled by | Odaily Planet Daily (@OdailyChina )

Translator | Wenser (@wenser2010 )

Editor’s note: For a long time, prediction market platforms like Polymarket and Kalshi have defined themselves as “the concentrated manifestation of crowd wisdom,” to distinguish themselves from betting platforms, and have elevated their valuation by emphasizing this narrative. But a recent paper from London Business School and Yale University, which dissects on-chain data from Polymarket, found that only less than 4% of addresses drive price changes with relatively substantial profits, while about 97% of addresses are mostly “runners-up,” with over 67% in a loss position. Considering that Polymarket’s user addresses have already exceeded 2.43 million, the research data may be somewhat lagging, but the phenomena revealed behind still warrant deep reflection.

Below is the main core content of this paper, summarized and organized by Odaily Planet Daily.

Truth One: Prediction Market Accuracy Has Nothing to Do with “Crowd Wisdom,” but Is Decided by a Small Minority of 3.14%

This is the most core conclusion of the entire paper and directly challenges the industry narrative.

Previously, many industry representatives took pride in this: Kalshi CEO Tarek Mansour said prediction markets “utilize crowd wisdom,” Polymarket CEO Shayne Coplan repeatedly promoted that “financial interests can more effectively aggregate information than experts,” and Robinhood CEO Vlad Tenev called it “the pursuit of truth by capitalism.” But the research data tell us: among 1.72 million Polymarket accounts, only about 54k accounts (3.14%) are identified as “skilled winners” (Odaily Planet Daily note: the paper summarizes these as professional players who can both average forecast and absorb information, and respond efficiently when news appears).

The main driver of price discovery in prediction markets is this minority, not the majority hiding behind “crowd wisdom.”

Truth Two: Making Money or Losing Money Can Be Pure Luck; 67% of Participants Are Essentially “Charity Givers”

In this paper, Roberto Gómez-Cram and others used a sign-randomization statistical method to categorize all traders’ accounts into four types: skillful winners (3.14%), luck-based winners (29.0%), luck-based losers (61.4%), skillful losers (6.4%).

The most counterintuitive figure is—luck-based winners account for nearly 30%, they made profits, but their trades contribute nothing to price discovery, statistically equivalent to random coin flips.

In other words, making money in prediction markets and “being able to predict the future” are two different things; while about 67% of the losers bear all the losses, essentially paying for the information advantage of the few.

Truth Three: Top Profit Players, 88% of Them Rely on Luck

Among the top 54k traders on Polymarket ranked by actual profit, only 12% are simultaneously identified by the statistical method as “skillful winners.”

In other words, the vast majority of big winners on the leaderboard, who have large profits, rely on luck from one or two big bets.

A typical case is account @majorexploiter—on a weekend in early 2026, this account invested $4.5 million across three sports events, earning over $3.6 million in profit.

Such concentrated bets are highly unsustainable; 60% of “luck-based winners” turned into losers in out-of-sample validation.

Truth Four: Skill in Prediction Markets Far Outperforms Traditional Fund Industry

The researchers split betting events into training and testing sets for out-of-sample validation.

Results show that accounts identified as “skillful players” in the training set, 44% of them remain identified as “skillful users” in the test set; by contrast, the same test on US actively managed mutual funds shows only 10% skill validity.

Conversely, the “anti-skill” (persistent losses) also remains highly consistent: 51% of “skillful losers” in the training set continue to be losers in the test set, compared to 20% for US mutual funds.

The final conclusion: Experts in prediction markets are truly skilled, while the “rookies” are truly rookies.

Truth Five: Skillful Winners’ Orders Are Highly Correlated with Final Outcomes

Based on a constructed order imbalance formula, the researchers found that for skillful winners, each 1% increase in net buying indicator (OIB) leads to about 2 basis points rise in the next period’s price, and about 8 basis points increase in the probability of the event occurring, with very high statistical significance (t-values of 12.71 and 9.51).

Meanwhile, the order flow of luck-based winners is not significant in either metric (t-values of only 1.47 and 1.49).

In other words, luck-based winners, despite making profits, do not have information content in their trades—this conclusion is very solid from the data perspective.

The phenomena observed from the study show that in markets where the settlement result is “Yes,” skillful winners are net buyers; in “No” markets, they are net sellers; they keep building positions aligned with the final outcome. Market makers tend to be net sellers in “Yes” markets and net buyers in “No” markets, consistent with their role of following directional order flow and earning spreads rather than establishing insider orders.

Truth Six: Skillful Traders Are the Only Group That Makes Prices More Accurate

Based on the premise that “some trades actually push prices toward the final outcome,” the researchers built a “price discovery contribution indicator” to measure whether the price in each time window is closer or farther from the final result.

Results show that only when the proportion of trades by skillful winners increases can the pricing error be significantly reduced (coefficient -5.00, t-value -5.54).

In contrast, trades by other three groups—luck-based winners, luck-based losers, and skillful losers—actually cause prices to deviate from the final outcome. In fact, most people are just creating noise at the trading level, and this impact grows as the market approaches settlement. During the last 20% of the betting event’s lifecycle, the contribution coefficient of skillful winners expands to -9.61.

Truth Seven: Skillful Winners Are the Only “News Traders”

To minimize errors caused by news transmission delays, the researchers selected events with clearly defined information release times, such as FOMC rate decisions and corporate earnings reports (Odaily Planet Daily note: the former is core to monetary policy expectations; the latter is key to understanding company fundamentals).

Data shows that only the order flow of skillful winners significantly shifts in the “unexpected direction” shortly after news releases.

In FOMC betting events, each 1% increase in unexpected direction corresponds to about a 5% increase in net buy volume by skillful winners (t=3.94); since the maximum unexpected move is about 6 percentage points, the reverse buying is substantial. For earnings reports, each 1% increase in unexpected direction corresponds to about 17 basis points increase in skillful winners’ net buy volume (t=2.62). In contrast, all other groups show no consistent reaction to news, some even acting in the opposite direction.

Truth Eight: Market Maker Profits Come from Spread, Not Information

Data shows that market makers on Polymarket account for only 0.1% of total accounts (about 1,660), but participate in an average of 942 betting markets per account, with an average profit of $11,832 per account.

Moreover, their order flow can predict price movements in the short term (because they keep “taking on” positions), but their influence on the final outcome is negative (from the data in Figure 3: coefficient -5.69, t=-10.30).

**This means they often take on insider traders’ sell orders in the short term, but long-term are “harvested” by insiders, mainly earning from spreads rather