Market vs Reality
← Basketball reports

Market vs Reality: NBA Weekly Prediction Report (November 4–10, 2026)

Market vs Reality compares prediction-market probabilities with what actually happens on the court. This report looks at 53 completed NBA games from November 4 through November 10, 2026, using market probabilities recorded approximately one hour before tipoff. This is an analysis of predictions, not a record of real-money wagers.

The probabilities for each game are normalized so the two teams' chances add up to 100%. We measure accuracy using the Brier score, which compares a predicted probability with the actual outcome. Lower is better: 0 is perfect, while a constant 50/50 prediction scores 0.2500 on every game.

Overall Performance

Across the 53 games, the average Brier score was 0.1057, or 0.1443 points lower than the 0.2500 baseline. For this particular week, the market probabilities predicted outcomes substantially better than a simple 50/50 forecast.

Favorites won 42 of 53 games, a 79.2% win rate. That is a strong result, although 11 favorites still lost.

Best Prediction

The most accurate prediction came in Chicago at New York on November 4. The Knicks had a 99.5% predicted chance of winning and beat the Bulls 135–105. The game's Brier score was 0.0000 when rounded to four decimal places (the unrounded score was 0.000025).

Biggest Miss

The largest error came in Indiana at Memphis on November 4. Memphis was favored with a 75.5% chance to win, but Indiana won 107–103. The Brier score was 0.5700, the highest of the week.

Notable Upsets

Several other games showed why even strong favorites are not guaranteed to win:

What We Learned This Week

The prediction markets performed well over these 53 games. Their average Brier score was comfortably below the 50/50 baseline, and favorites won nearly four out of five games.

Still, the Indiana–Memphis result and other upsets show the difference between a likely outcome and a certain one. A team with a 75% chance to win can still lose; that possibility is part of what the probability is supposed to express.

One week is not enough to establish how accurate or well-calibrated these markets are over time. As more results accumulate, Market vs Reality will be able to compare weekly scores and see whether the performance holds up.

Method: The analysis uses normalized pregame probabilities from the approximately one-hour snapshot bucket. No post-tipoff observations were used in the reported calculations.