Prediction platforms are emerging as a leading indicator for key economic data, according to a report by market maker Keyrock.
In collaboration with @Dune, we’ve released our prediction markets report.
It’s a data-backed dive into how they are becoming core information and hedging infrastructure across crypto, media, and institutional finance, with insights from over 10 contributors.
Find out more 👇 pic.twitter.com/k2Lhwiha8V
— Keyrock 🔑🪨 (@keyrock) December 17, 2025
Analysts compared Kalshi’s “Inflation in 2025” market with the Nowcast index from the FRB Cleveland. The bank’s tool lags: it updates only after official statistics are released.
Prediction platforms move first, continuously absorbing market information—from analysis to rumours.
For example, in April, ahead of the CPI release, Kalshi predicted rising inflation (the forecast rose from 3.05% to 3.58%). On FedNow the figure stayed around ~3.9% only after the official publication.
In July the pattern repeated. While FedNow reacted with a delay of several days, Kalshi’s readings remained stable—the market had already priced in the shift.
“Rather than a lagging statistical update, Kalshi represents a real-time consensus about future inflation, giving traders an informational edge to anticipate macro turns before they are reflected in official forecasts,” the specialists said.
Solana-project Worm co-founder Nass Dib noted that hedge funds already use prediction platforms. These tools help them move ahead of the curve by tracking shifts in investor sentiment.
Forecast accuracy
The authors cited an analysis by data specialist Alex McCallum showing that Polymarket consistently outperforms traditional polls. The platform’s accuracy ranges between 90% and 95%:
- A month before an event, the market correctly assesses the outcome in 90.4% of cases;
- as the date approaches, the figure dips to 88.5%;
- in the decisive final hours before close, accuracy jumps to 94.1%.
Simple “did it happen or not” tallies are a blunt instrument. They equate forecasts with 90% and 51% probabilities, even though market confidence in them differs.
Keyrock’s experts proposed using the Brier index. It measures not a “hit” but the degree of match: how close the forecast probability is to the realised outcome. In this model, the ideal score is 0: the lower the score, the closer the forecast is to reality.
The average score for prediction platforms is 0.09—noticeably better than traditional industries:
- weather forecasting — 0.14–0.15;
- sports betting — 0.18–0.22.
However, much depends on liquidity. Binary yes/no markets work well with high trading volumes; in thin markets, the story is different.
“Traditional models assume continuous price movement, not discrete ‘yes or no’ outcomes. Combined with low liquidity, that leads to wider spreads, sharp volatility and more complex margin dynamics,” explained KPMG managing director Duncan Hennes.
Engagement
Keyrock also found that Polymarket outperforms most crypto projects on retention. Its results are higher than 85% of DeFi projects, wallets and centralised exchanges.
User-activity analysis identified three stages in the evolution of platform loyalty:
- Era of pioneers. The core consisted of crypto-enthusiasts driven by the novelty of the technology. Their deep engagement delivered record retention (up to six months) despite modest volumes.
- Phase of growth and stress. Expanding the audience naturally reduced average retention. The dispute with the CFTC in 2022 exacerbated matters, cutting the platform off from its key market — the United States.
- Habit formation. Broadening the roster of markets and rising liquidity reversed the negative trend. The chief sign of recovery is higher two-month retention and a slower long-term churn. The platform is gradually becoming an everyday tool for users.
The analysts stressed that Polymarket’s metrics are organic—rooted in the nature of the product.
In their view, the key to success is the event cycle. Elections, macro releases and sports seasons create a natural rhythm that regularly brings users back.
In early December, Binance founder Changpeng Zhao unveiled the prediction platform predict.fun.
Two weeks later, the decentralised crypto exchange PancakeSwap announced a prediction market launch.