As you might know, we held a prediction contest for the World Cup. Just over 70 people participated in our contest and were asked to predict the winners for each round once the matchups were determined. Another question we asked all participants was to pick the eventual winner of the tournament. Participants could switch their pick at any time, but you would receive fewer points the closer we got to the final game. In June, at the start of the tournament, the most popular pick was Brazil. Twelve (12) percent picked Spain to win, and the same number picked the runner-up Argentina. Our contest was just for fun, and no money was involved. At the same time, the two big prediction markets (Kalshi and Polymarket) were breaking records with around $40 billion being bet on the World Cup games during the tournament and roughly $6 billion on the final alone. These markets are viewed by some as a more accurate predictor than surveys or other market research tools, since people placing bets on these markets have "skin in the game" and therefore are making more accurate predictions. Is this theory correct? We certainly need a lot more data to know the answer, but in this one small sample test, placing bets did not make for a more accurate set of predictions throughout the tournament.
Key Takeaways:
➡️ / 📈 From the beginning: As mentioned above, 12% of participants in our contest had predicted that Spain would win on or before June 29th. On that same day, 11% of Polymarket bets had Spain winning, and 10.5% of bets on Kalshi had the same. The prediction markets were much more bullish on France and Argentina than our contestants, with 24% of bets predicting France would win and about 22% were on Argentina.
🔢🔢🔢Round of 32/16/8: In the Round of 32, the Trendency panelists did a better job of predicting the winner in 7 of the matches. Prediction market users were better in 2 games, and the remaining 7 were nearly identical between the two. In the following round, the contest participants had an edge in 3 of the matches, while the prediction markets did better in 1, and 4 were roughly the same. In the quarter-finals, contest participants were more accurate in 2 matches, while the prediction markets did better in 1, and 1 match was nearly identical.
🏁🏆🏆 The Finals: In both the semi-finals and the finals, both the contestants and the prediction markets were split nearly 50-50. In the France vs. Spain matchup, both groups thought France was going to win (they did not), but our contest participants were much more sure of France's chances. In the other semi-final, England was given the edge with about 52% of both cohorts feeling England would win. In the finals, 52% of prediction market bets were on Spain, while 52% of our contestants picked Argentina to win.
This is a small sample size test, but the results point to a big question mark on whether or not the need for people putting money behind their predictions makes for a more accurate outcome.
