shoes22
Hall-of-Famer
- Joined
- Dec 4, 2012
- Messages
- 28,551
- Reaction score
- 39,205
- Points
- 148
http://projects.fivethirtyeight.com/2016-nba-picks/
Team-By-Team Forecast
Forecast from Record Conf Point Diff Conf Semis Conf Finals Finals Championship
17671773
+32
Thunder11-4West+7.0✓✓84%54%
17051715
-3
Cavaliers11-2East+6.3✓✓87%32%
17671772
-32
Warriors9-5West+9.9✓✓16%12%
16101618
+3
Raptors10-9East+2.9✓✓13%3%
Id be interested to see how 538 got to that total because it is SOOoo far from what Vegas has.
If anyone is interested in how 538 does their forecasts/predictions, it is an incredibly simple ELO model. You can design your own approximation at home in minutes. Literally the only thing they use to calculate their ELO ratings are overall win-loss records of a team, margin of victory in games, and whether it was a home or away game. That is literally it.
The Thunder have such a high ELO rating for 4 wins in 6 tries against the Spurs, and 3 wins in 4 tries against the Warriors...two teams that have had stronger ELO ratings than them. So by beating them (especially any time they've won on the road) they've main big gains in their rating, while their opponents' rating decrease by a similar amount.
From there they just use the mathematical odds they've calculated for individual games (exactly the same as I do for my model) to calculate the percentage odds of winning a series and a championship.
As @David. and @gourimoko noted, a system like that is a good estimation for the regular season, but next to meaningless in the playoffs...especially the later rounds where there is less of a talent disparity between teams. It does not factor individual match-ups at all. It really doesn't factor anything...just "did team A or team B win?" and by how much.
Last edited: