Forecast vs result
Surprise Games
The games that went furthest from what this model expected. "Surprising" here means surprising to this forecast, which is not perfectly calibrated, not necessarily to everyone watching.
Ranking
| # | Date | Matchup | Forecast favorite, chance, away–home | Final away–home | Winner's pregame chance | Margin miss | Surprise |
|---|
Most unexpected winners, all seasons
- 8%WAS beat DET126–117
- 10%SAS beat SAC142–134
- 10%CHA beat OKC124–97
- 11%DEN beat SAS128–118
- 12%ORL beat CLE108–103
- 13%PHX beat OKC135–103
- 13%SAC beat DEN128–123
- 13%MIN beat HOU117–107
- 13%UTA beat DEN118–114
- 14%ATL beat BOS117–116
- 14%SAC beat LAC118–109
- 14%IND beat OKC117–114
- 14%DET beat UTA125–116
- 15%WAS beat MEM130–122
- 16%HOU beat PHX122–121
- 16%LAL beat OKC126–99
- 17%ORL beat DET112–101
- 17%DAL beat DEN131–121
- 17%PHI beat BOS117–116
- 17%WAS beat ATL132–113
- 17%WAS beat MIA110–108
- 17%NYK beat CLE130–116
- 17%IND beat MIL139–123
- 17%CHA beat SAC125–119
- 18%SAS beat GSW114–111
How surprise is measured
- Unexpected winners
- Ranked by the pregame win probability the forecast gave the team that won, lowest first. The technical measure is −ln(p): 0.22 for an 80% winner, 0.69 for a coin flip, 2.30 for a 10% winner. Probabilities are bounded at 0.0001 inside the logarithm only, to keep it finite; displayed probabilities are the stored ones.
- Margin misses
- Ranked by the absolute difference between the forecast home margin and the final home margin, in points. A forecast of a 5-point home win followed by a 20-point home loss is a 25-point miss. This is a separate ranking: a big margin miss can still have the right winner.
- Most recent
- Games where the forecast favorite lost, newest first.
- Which games
- Every completed game with a stored forecast, one forecast per game (see Methodology). Games not yet played are never included.