Market vs Reality
← Basketball reports

NBA Prediction Market Report: November 10–16, 2026

Introduction and Method

Market vs Reality monitors how well pregame prediction-market probabilities correspond to actual NBA game outcomes. Using historical snapshots taken from the one-hour pregame snapshot bucket—approximately an hour before tipoff—we analyze probabilities before games begin. Probabilities are normalized so that the home and away win chances sum to 100%. We do not use updated or in-game odds for this report.

The main measurement used is the Brier score. This score quantifies squared error between predicted win probabilities and actual results. Lower Brier scores indicate better prediction accuracy. A perfect forecast scores 0.0000, while a naive constant 50/50 prediction scores 0.2500 per game.

Our dataset covers all 51 NBA games played from November 10 through November 16, 2026.

Results Summary

Between November 10 and 16, 51 games were analyzed using one-hour pregame market snapshots. No games were excluded, and no favorites were tied in probability at snapshot time.

The average Brier score across all games was 0.0960. This is 0.1540 points lower than the 0.2500 baseline score assigned to a constant 50/50 prediction, indicating that the market probabilities were generally closer to observed results than this baseline.

Favorites won 44 of the 51 games during this period, yielding a favorite win percentage of 86.3%. This reflects a strong market consensus predictive power for favorites during the week.

The best prediction occurred on November 10 in the game between the Los Angeles Clippers (LAC) and the Phoenix Suns (PHX). The market gave PHX a 99.5% chance to win approximately an hour before the game, and Phoenix won convincingly with a 133-110 score. The resulting Brier score for this game was 0.0000 when rounded, representing one of the closest possible matches between probability and outcome.

The largest prediction miss came on November 13 in the Oklahoma City Thunder (OKC) at Los Angeles Clippers (LAC) game. The market strongly favored OKC with a 75.5% win probability, but LAC won by a narrow margin of 113-112. This produced a Brier score of 0.5700, a substantial difference from the typical scores seen during the week.

There were multiple notable upsets where the underdog won against the favored team:

No games were excluded due to tied favorites, so all scheduled contests for the week are included in this analysis.

What We Learned This Week

This week’s data shows that market-implied probabilities taken from approximately one hour before tipoff generally outperform a naïve 50/50 baseline when predicting NBA game outcomes. The average Brier score of 0.0960 is well below the baseline of 0.2500, and favorite teams prevailed in the majority of matchups (86.3% of games).

Nonetheless, sizable misses still occur, as shown by the largest Brier score of 0.5700 on one upset by LAC over OKC. Several other games featured notable upsets where underdogs prevailed despite being given substantially lower winning probabilities.

Continuous tracking of these pregame snapshots can help reveal how closely market probabilities align with actual NBA results on a weekly basis. For this week, prediction markets demonstrated strong accuracy during this reporting period, although occasional significant misses still occurred.