My best NBA projections use rankings that incorporate both scores of games and closing point spreads from the markets. These ensemble rankings rate Golden State 5.3 points better than Cleveland on a neutral court.
I then calculate a series win probability by considering this rating differential and a home court advantage of 4.5 points. This implies a 86.3% win probability for Golden State.
It’s worth breaking down some of the components of these ensemble rankings.
By team rankings that consider all games, Golden State rates 7.1 points better than Cleveland on a neutral court. However, this is a misleading number as teams rest their star players and deal with injuries.
I adjust for these player absences by only considering games in which a team has its top 2 or 3 players. These filtered team rankings make Golden State 4.0 points better than Cleveland, which corresponds to a 80.5% win probability.
I like this number since it closely corresponds with another set of my calculations.
In my market rankings, I take closing lines and adjusting for opposition with my ranking algorithm. The playoffs markets have been distorted by injuries, so I use data from the trade deadline in February until the end of the regular season.
These market rankings make Golden State 4.6 points better than Cleveland on a neutral court. This corresponds to a 83.5% series win probability for the Warriors.
Each of these components gives a win probability for Golden State higher than the 73% implied by the markets (Golden State -270 on Wednesday at Bookmaker). The markets feel like Cleveland and LeBron James can outplay their year long numbers more than Golden State.
I don’t buy it, and it comes down to defense.
In points allowed per possession, Cleveland ranked 21st this season, a poor showing for a championship caliber team. This group is capable of better, as they ranked 10th last season.
However, their defense isn’t nearly as good as that of Golden State, which ranked 2nd this season after 5th last season. It will make the difference in this series, which could be much shorter than anyone expects.