Why I got interested in this
I've been running a prediction market called Liquor Bets for the last year, specifically focused on spirits industry questions. The setup: participants make probability-weighted predictions on specific outcomes in the spirits category. Who gets acquired. Which regions show the strongest growth. Whether a specific emerging category (mezcal, shochu, domestic rum) maintains trajectory or plateaus.
What I found consistently is that the crowd is better calibrated than individual experts, including me, which is humbling and also interesting.
How prediction markets work
The basic mechanic: participants are given a binary or range question with a defined resolution date. They stake virtual (or real) currency on their estimated probability. The market price reflects the aggregate probability weighted by participant conviction.
What makes this more informative than a survey: participants have to put skin in the game. A survey respondent can say "I think domestic rum will be the breakout category in 2026" without any cost if they're wrong. In a prediction market, conviction is expensive. That filters out the people who have opinions but no real information or reasoning behind them.
The markets that have been running in financial contexts for decades consistently outperform expert panels on forecasting accuracy. The reason is structural: experts tend to anchor on their existing view and update slowly. Markets update continuously as new information enters.
What Liquor Bets has shown
A few things that have come out of running markets on spirits-specific questions.
Acquisition timing is harder to forecast than acquisition targets. Participants with industry knowledge identify likely acquisition targets reasonably well. When the acquisition happens is much harder to predict, because that depends on deal mechanics that aren't visible to market participants.
Craft spirits growth is more regional than the national narrative suggests. National trade coverage makes it sound like craft spirits growth is uniform. The market participants, who skew toward people with on-premise and retail knowledge in specific markets, consistently show that the growth is concentrated in a handful of urban markets and that secondary markets are much more price-sensitive and traditional in their purchasing.
Bartender adoption leads broader consumer adoption by 18-24 months in premium categories. This is something the industry broadly knows, but the prediction market data makes the lag quantifiable in a way that trade surveys don't.
The wisdom of crowds has limits
I want to be honest about where prediction markets fail, because treating them as infallible is as wrong as ignoring them.
Thin markets are noisy. A prediction market with 15 participants tells you almost nothing reliable. The signal-to-noise ratio improves significantly after around 50 active participants, and gets genuinely useful at 200+. Liquor Bets is in the building phase. Some of the results I have confidence in. Others are preliminary.
Participants self-select. The people who join a spirits industry prediction market are more interested and more informed than average, which means the market doesn't represent broader consumer behavior. It represents the views of enthusiasts and industry insiders. That's a useful signal for some questions (trade and on-premise trends) and a noisy signal for others (mainstream consumer adoption).
Manipulation risk. Small markets with real-money stakes can be moved by a single large participant with bad faith motivation. This is a known problem in prediction market design.
Why this matters for the industry
The spirits industry makes enormous allocation decisions on the basis of trend forecasts that are mostly vibes. Which categories to invest in, which markets to push distribution in, which formats (bottle size, price tier, packaging) will resonate, all of these are currently decided by a mix of sales data, trade publication coverage, and the strongly-held opinions of senior brand managers.
The sales data is backward-looking. Trade publication coverage is driven by PR relationships. Senior brand manager opinions are shaped by their career incentives, not just market reality.
A well-run prediction market with good participant diversity gives you a forward-looking signal that aggregates information across a range of knowledge types. It won't replace market research. It will catch things that market research misses, specifically the distributed, informal knowledge that lives in the heads of buyers, bartenders, and enthusiasts who are closer to the consumer than any brand team.
That signal is worth something. The industry has been slow to treat it seriously, which is probably an opportunity.
