Alex Egol – RSAC Writer

It is the morning after Game 1 of the NBA finals, in which the Raptors defeated the Warriors 118-109. Although we could do a whole analysis of the game, we are instead going to focus on predicting the remaining games. We will use the analytical methods of Wayne Winston and Jeff Sagarin as a model in order to do this.

Developed by Wayne Winston, adjusted plus-minus is a form of plus-minus that attempts to eliminate the standard plus minus’ inherent bias towards players who play on good lineups. It is therefore perceived as a better proxy for individual player contributions.

Once you calculate adjusted plus-minus, the next step to comparing lineups, according to the Winston Sagarin Method, would be adding up the adjusted plus-minuses of each lineup’s individual players. Comparing the two total values can give you an approximation for what the score difference will be when those two lineups are in the game. The last step is the application of a mathematical function called WINVAL which, based on historical trends, is able to minimize the score difference calculated by the addition of the plus-minuses and the score difference perceived in real games. Once you apply this function, you have established a strong comparison.

The only problem is that, for us, in 2019, the adjusted plus-minus and the WINVAL function are fairly outdated, and they are not easily accessible online. Therefore, as an alternative to adjusted plus-minus, we will use ESPN’s Real Plus-Minus statistic which is a similar statistic in terms of the range of values it typically produces. Not much has been written about the WINVAL program either, and since its sole purpose is to minimize error I believe we can account for this error in other ways.

The next thing I am going to do is establish a road map for our investigation of the NBA Finals. First, we will add the real plus-minuses of each player in each lineup for both the Warriors and Raptors. Then we will compare, for the first and second lineups, which team has a larger expected score. We will then weigh the first and second lineups according to how much each team plays each lineup, and we will have approximated the score of a game between the Raptors and Warriors. The last step will be to turn this expected score into a predictive measure.

The Raptors played Kawhi Leonard, Danny Green, Marc Gasol, Pascal Siakam and Kyle Lowry to start last night’s game. To save time, I will tell you that the total plus-minus for the lineup was 20.09. The table below illustrates this.

The Warriors played Draymond Green, Steph Curry, Klay Thompson, Andre Iguodala, and Jordan Bell to start last night’s game. However, since we’re attempting to predict the result of the future games, it would be reasonable to include Kevin Durant in this lineup, because, although he didn’t play last night, he is expected to return for Game 3. The total plus-minus for the lineup replacing Kevin Durant for Jordan Bell, and assuming Kevon Looney subsequently replaces Iguodala, is. The table below illustrates this.

Based on a comparison between these two totals, we learn that the Raptors are expected to win by approximately 1.9 points for every 100 possessions these two lineups play against each other.

Calculating the second lineup is a little bit more difficult because, assuming the same personnel choices as last night, the number of bench players for each team is different. The Warriors played 7 players off the bench (since we’re including Durant as a player who “played”) whereas the Raptors only played 4. We will therefore calculate the average real plus-minus of each team’s bench players to create a second lineup that we can compare. Between the Warriors’ seven bench players, the average real plus-minus was -1.14. Between the Raptors’ four bench players, the average was -0.62.This means the Raptors’ bench has an expected differential of +0.52 points over 100 possessions compared to the Warriors’. The tables below illustrate this.

Although the Raptors have a positive expectance either way, we will now adjust for the amount each team plays their first and second lineups. Going off of last night, the Warriors play their starters approximately 2/3 and bench approximately 1/3 of the time, and the Raptors play their starters approximately 3/4 and bench approximately 1/4 of the time. If the Warriors are expected to lose by 1.9 with their starters in and by 0.5 with their bench in, a weight tells us that, over the course of a game they play their starters 2/3 of the time, they are expected to lose by 1.43 overall. If the Raptors are expected to win by 1.9 with their starters in and by 0.5 with their bench in, a weight tells us that, over the course of a game they play their starters 3/4 of the time, they are expected to win by 1.55. Since, in a game between two opponents, the team that wins wins by the same amount of points that the team that loses loses, we can develop a proxy by calculating the mean of the absolute values of these two score values, -1.43 and 1.55. That value is 1.49.

Using this information, we can assume that the Raptors are 1.5 point favorites in every game they play against the Warriors. This does not account for home court, however. Therefore, just as Sagarin and Winston do, we will gave a +3.2 to the home team. This means the Warriors become slight favorites in their home games, approximately +1.7 point favorites, and the Raptors become +4.7 favorites at home. With 3 home games left for each team, probabilistically speaking, the Raptors appear to be favorites to win the Finals. However, in the absence of the proper tools of the Winstonian and Sagarinian method, it is impossible to guarantee that these observations have a truly significant amount of statistical importance. Anyways, I still thought it would be interesting to explore.