CryptoRUs’ George Tung Breaks Down Why Prediction Markets Are Beating Polls

Prediction markets are increasingly outperforming traditional polling as forecasting tools — and the reason comes down to one thing – financial conviction. When people put real money behind a prediction, they don’t lie.

The rise of platforms like Polymarket and Kalshi is challenging the dominance of traditional forecasting. Pollsters have long been the dominant voice in predicting political and economic outcomes. But a string of high-profile polling failures from the 2016 U.S. election to Brexit has opened the door to a challenger that punishes uncertainty with hard cash.

Why Money Makes Better Data

The core argument for prediction markets is behavioural. Exit polls and surveys suffer from a well-documented problem: respondents often give answers they think sound reasonable, or answers that reflect who they want to win rather than who they think will win. There’s no cost to being wrong on a survey form.

Prediction markets eliminate that gap entirely. Every probability reflected in a market price represents someone who was willing to risk actual capital on that outcome.

“It takes conviction to place a prediction or a bet,” George Tung, founder of ClashPicks and host of the widely followed CryptosRUs channel, told BeInCrypto. “You have to be pretty sure that something’s going to happen for you to actually put down real money.”

That conviction makes the data generated by prediction markets fundamentally different in quality. It isn’t sentiment, it’s skin in the game.

The numbers back this up. Independent research by data scientist Alex McCullough, published via a Dune dashboard, found that Polymarket predicts outcomes with roughly 86% accuracy one month before an event resolves, climbing to around 91% in the final four hours. The study analysed Polymarket’s historical data, excluding markets with extreme probabilities to avoid skewing results.

The Polling Problem

Traditional polling has been struggling. Despite methodological overhauls after 2016 and 2020, polls still overestimated Kamala Harris’s chances in the 2024 U.S. election and underestimated Donald Trump’s, especially within swing states.

Prediction markets, meanwhile, told a different story well ahead of election night. Tung is emphatic that this edge is skill-based, not random.

“If you’re predicting on an outcome like a presidential election or if gold is going to go up this week — it’s skill-based,” he told BeInCrypto. “There are people that do an extensive amount of research and they study things.”

Subscribe to our YouTube channel to watch leaders and journalists provide expert insights

The structural reason is speed. Polls take days to field, weight, and publish. A well-resourced prediction market reprices in minutes when new information hits.

Not Without Flaws

However, the case for prediction markets isn’t watertight. Critics point to a significant structural vulnerability: when participation is concentrated among a small, homogeneous group of traders, markets can be moved by a single large actor — producing prices that reflect individual conviction rather than genuine collective wisdom.

The demographic gap is also real. Prediction market participants skew heavily toward crypto-native, financially sophisticated users — hardly a representative sample of the broader public. Critics argue this limits how far the “wisdom of crowds” argument actually stretches when the crowd is this narrow.

Tung acknowledged the tension directly.

“I agree that as the platform gets bigger and there are more people on it, the more accurate it’s going to be,” he said. But he pushed back on the framing that demographic reach is a weakness unique to prediction markets. “What other form of data has more people predicting than prediction marketplaces combined? What data actually has a bigger demographic than this?”

It’s a fair challenge — and one the polling industry has yet to convincingly answer.

Newer platforms entering the space are betting that broadening participation is the key. ClashPicks, Tung’s own prediction market built on Solana, offers a free-to-predict model designed to lower the barrier for first-time users, with the explicit goal of pulling in participants who would never open a Polymarket account.

What Comes Next

Whether or not prediction markets fully displace polling is beside the point. They’ve already changed the conversation. Institutional investors, campaign strategists, and media organisations are now incorporating prediction market data alongside and sometimes instead of traditional polling aggregates.

The scale of institutional interest is hard to ignore: in October 2025, Intercontinental Exchange (ICE) invested $2 billion into Polymarket, valuing the company at $9 billion. That’s not a bet on a niche crypto experiment. That’s a signal that the financial mainstream is taking prediction markets seriously as a data infrastructure play.

The next test will be whether the industry can broaden its participant base without losing the skin-in-the-game quality that makes the data valuable in the first place. More participants mean more diverse information, but only if those participants are genuinely informed, not just speculating. That balance is still being worked out.

For now, prediction markets are the most honest mirror we have for what people actually believe will happen, because getting it wrong costs them something.