NEW: UC Berkeley study accidentally reveals prediction market whales are better traders...

In partnership with

_{___________________________________________________________________}

July 3, 2026

A UC Berkeley study dropped this week has been getting passed around that "whales" are actually less accurate than smaller participants across Polymarket and Kalshi markets.

They concluded: big money is "unsophisticated," driven by ideology instead of information.

From the paper's own abstract, verbatim:

"Rather than acting as sharp informed players, these large actors likely trade on ideological conviction, structurally overpaying for specific narratives and suffering from adverse selection against smaller participants."

And again in Section 3.4's Results and Analysis, discussing Mention Markets specifically:

"Unlike traditional financial markets where capital dominance correlates with institutional sophistication, prediction market Whales in cultural or political markets are frequently highly capitalized individuals trading on ideological conviction rather than objective probability. These participants disproportionately drive the structural 'Yes Bias' observed earlier in our analysis, willingly overpaying for large positions in their preferred outcomes."

Maybe they're right…but, ideological conviction?

Please.

I read the paper.

There’s some issues.

First, the sample. Is very limited.

The study looks at three market categories: 15-minute crypto markets, NBA games, and Mention Markets, markets built around whether a politician says a specific word during a live TV interview.

Total observations after their own filters: nearly 7 million trades. Sounds comprehensive.

But you have to ask, why only three market categories? Out of thousands.

No justification. No explanation of why geopolitical markets, election markets, macro markets, or long-duration contracts were excluded. No discussion of selection bias. The only constraint they mention is timing, markets that started after a certain date and resolved by a certain date.

It looks like they picked these three because they were easy to collect and backfill programmatically.

Short-duration, high-frequency, structurally simple. The paper even admits in its own limitations section that the dataset "captures only a specific slice of market behavior."

In short: they didn't study the markets where serious money actually lives. They studied the ones that were easy to scrape.

The category driving almost the entire "whales are dumb" result, the only one where the whale effect clears statistical significance, is Mention Markets. After the researchers apply their own liquidity filters, Mention Markets contribute 1,726 observations across 87 markets. That's an average of 20 trades per market.

Twenty trades per market. Two zero.

In a sample that sparse, "whale" doesn't mean big capital. It means whoever placed a modestly larger bet than the two or three other people trading that specific contract at that moment.

A whale in a $2,000 Mention Market and a whale in a $1 million whether the Strait of Hormuz remains closed or the Fed cuts rates are not the same economic entity.

The researchers at Berkeley never make that distinction.

They call them both whales and draw the same conclusion about both.

And the finding doesn't even hold across their own three categories.

In NBA markets, the specific test that would prove whales underperform comes back statistically indistinguishable from noise. In 15M Crypto, the size effect is real but the whale effect isn't.

Therefore, the headline rests almost entirely on 20 trades per market in the weakest dataset in the study.

For comparison, Prediction Market Whales catalogues 36,000 to 60,000 trades a day, every day, at a $1,000+ floor.

Their entire "whales are dumb" conclusion rests on fewer trades than we process before lunch. Which, sidebar, they also say they’re driven by ideology, which, good luck measuring that…

Second, the mechanics.

If you thought that was bad, check this out.

Section 3.4, buried after the headline conclusion, documents that Takers show positive edge across all three market categories. Their own words: "A positive pre-fee edge indicates adverse selection against market makers."

Here's what that means.

Every trade has two sides, a Maker and a Taker. The Maker posts a resting limit order and waits. The Taker crosses the spread to execute immediately. What the Taker makes, the Maker loses. They're the two sides of the same transaction.

The paper proves Takers win. Consistently. Across all three categories.

Now here's the problem: large capital in nearly every financial market on earth prefers to be a Maker. Not because they're unsophisticated, because it's cheaper. If you're moving serious size, you don't cross the spread and move the market against yourself. Most real whales, the major players, the ones who run money or direct the largest capital flows, post a limit order and let the market come to them. That's why in the PMW live feed you'll often see simultaneous buy and sell orders hitting the tape in real time. That's basic execution.

I don't think the researchers at Berkeley understand that.

Because here's the contradiction they never address: their own Section 3.4 proves that Takers beat Makers. Their headline section proves that Whales lose. But they never connect those two findings. If Takers beat Makers, and Whales are mostly Makers, which is how large capital behaves in every market on earth, then "Whales lose" isn't a discovery. It's just arithmetic. The paper proved its own conclusion was meaningless and didn't notice.

It gets more absurd. They cite a separate Kalshi paper in their own bibliography, Bürgi, Deng and Whelan, finding that Makers earn the higher return on Kalshi. The exact opposite conclusion.

Read that again: inside the same paper that concludes whales are unsophisticated losers, they cite research showing the opposite is true on a comparable platform. They put the contradiction in their own footnotes and never addressed it. Not once.

There's also no separation between a directional conviction bet and a flat, hedged, two-sided book. A market maker running near-zero net exposure, buying YES from one taker, selling YES to another, accumulates massive cumulative size and lands in the whale bucket, despite making no directional forecast at all. The paper scores them on directional edge per leg.

In other words: a sophisticated participant who takes both sides of a trade to capture the spread gets classified as a whale who "lost" on direction. The paper mistakes a profitable liquidity strategy for a losing conviction trade. Those are opposite things.

Third, and again, they never actually measured ideology.

The only tool in the entire study that could plausibly capture conviction or belief is a separate NLI sentiment analysis on trader comments. The authors' own verdict on that analysis: "the resulting figures were non-interpretable... we did not proceed with downstream modeling."

Their attempt to measure conviction came back inconclusive, by their own admission. And yet "ideological conviction" still shows up in the abstract and conclusion as the confident explanation for why whales underperform.

The mechanism they lead with isn't something they measured. They assumed it. They retrofitted a story onto a regression result that, as established above, may just be reflecting maker/taker mechanics. "Ideological conviction" isn't a finding.

I'm open to real research, better studies, and better questions.

But this paper was horrible.

Based off of two thousand thin trades, a dropped control variable, and an unmeasured mechanism?

If they couldn’t prove, let alone define whales, does that mean whales are better traders?

Regardless, again, like we’ve said before, be careful of headlines.

Enough about what Berkeley thinks whales are doing. Here's what they're actually doing.

⛪ THE MARKET NOBODY SAW COMING

🐳 YES — Will Jesus Christ return before 2027? — $9.3M volume. Nine point three million dollars on the Second Coming resolving this year. More volume than every 2028 Republican candidate combined except Eric Trump and JD Vance. Whales say yes. Make of that what you will.

⚽ THE WORLD CUP GOLDEN BALL — EVERY NAME, EVERY POSITION

While Mbappé gets the headlines, whale money is telling a completely different story about who takes home the tournament's top individual honor.

The whale picture on the Golden Ball: Every name in the conversation is getting faded. Ronaldo, Messi, Vinícius, Fernandes, Rice — all buried. Smart money either knows something about a player not on this list, or the award stays wide open through the final.

⚽ THE MATCH MARKETS

🐳 YES — Will Brazil win on 2026-07-05? — $49.0K volume. Smart money backing Brazil in their next match.
🐳 NO — Will Norway win on 2026-07-05? — $42.8K volume. Norway getting faded on the same day.
🐳 NO — Will 8+ matches be decided by penalty shootout during the 2026 FIFA World Cup? — $20.4K volume. Whales don't see a shootout-heavy tournament.

🏀 THE MARKETS NOBODY'S COVERING

🐳 NO — Will Kingston Flemings win the 2026-27 NBA Rookie of the Year? — $272.3K volume. The next generation getting faded before the season starts.
🐳 NO — Will LeBron James play for the Miami Heat in 2026-27? — $157.5K volume. The homecoming narrative — faded.
🐳 NO — Will Arizona Cardinals win the 2027 NFL NFC Championship? — $286.8K volume.
🐳 NO — Will Minnesota Vikings win the 2027 NFL NFC Championship? — $151.3K volume.

None of this is 20 trades per market. And none of this is bots in a thin Mention Market betting on whether a politician says a word.

ALL of this is the real deal.

More prediction market market intelligence than we could dream of.

Happy 4th everyone.

________________

Don’t Wait for the OpenAI IPO

OpenAI and Anthropic could be two of the biggest AI IPOs Wall Street has seen in years.

But investors don’t have to wait for those names to hit the public markets to get exposure to the AI boom.

MarketBeat’s 7 AI Stocks to Buy Now report reveals 7 publicly traded companies already positioned to benefit as the next wave of AI investment moves beyond the private model providers.

These are the stocks investors can buy today, before the IPO crowd rushes in.

See the 7 stocks

_{___________________________________________________________________}

_DISCLAIMER:

_{The Content is}_{not intended to provide, and does not constitute}_{, financial, investment, trading, tax, legal, or any other form of professional advice. It is}_not_{a recommendation, suggestion, solicitation, or offer to buy, sell, trade, or hold any securities, event contracts, derivatives, cryptocurrencies, or other financial instruments on platforms such as Polymarket, Kalshi, or any other prediction market.}

_{Prediction Market Edge believes the Content is reliable but}_{makes no representations or warranties}_{as to its accuracy, completeness, timeliness, or suitability for any purpose. The Content is subject to change without notice, and Prediction Market Edge assumes}_{no duty or obligation}_{to update it.}

_{Trading in prediction markets involves significant risk of loss}_{, including the potential loss of your entire investment. Past performance (including any highlighted “wins” or gains) is}_{not indicative of future results}_{. Markets are volatile, influenced by news, liquidity, resolution rules, and other factors, and individual results will vary. Subscribers and readers should conduct their own independent research, consider their financial situation, risk tolerance, and objectives, and consult qualified professionals before making any trading or investment decisions.}

_{Prediction Market Edge is}_{not responsible}_{for any third-party information, market data, platform rules, or services referenced herein, including but not limited to Polymarket, Kalshi, or other exchanges. Use of the Content is at your own risk.}

_{By subscribing to or accessing this Newsletter or related materials, you agree that Prediction Market Edge and its affiliates shall not be liable for any direct, indirect, incidental, consequential, or other damages arising from your use of the Content.}

_{For important additional information, please review our full Terms of Service, Privacy Policy, and any Subscription Agreement (available on}_{predictionmarketedge.com}_).

NEW: UC Berkeley study accidentally reveals prediction market whales are better traders...

Don’t Wait for the OpenAI IPO

Reply

Keep Reading

Prediction Market Edge