NBA Season Preview 2018-19: Optimizing the Celtics roster

The 2018-19 Celtics have a problem; they have too much talent. Most people wouldn’t consider that a problem but I do. An abundance of talent is not a bad thing, but getting everyone the playing time they deserve will be a challenge for Brad Stevens. I’m not sure any coach in the NBA feels sorry for coach Stevens and his playing-time predicament but it will be interesting to see how it works out.

Let’s analyze this problem from the mindset of a computer scientist. I’m not a computer scientist but I did read a book by Tom Griffiths, Algorithms to Live By. In the book and in this TED talk Griffiths trips about how computer scientists solve problems and how we can use these ideas to make better decisions. The Celtics have an optimization problem and Griffiths offers two methods that could help optimize the Celtics’ lineups.

The first rule is called the 37 percent rule and states that in order to get the optimal choice we need to try an experiment on at least 37 percent of the opportunities and then go with the next choice that is as good as or better than the rest. For an NBA season, it would make sense to experiment for the first 37 percent of the season or the first 30 games. Combining the 37 percent rule with another rule of Griffiths, the Celtics can maximize their optimal choice.

The second rule is called the explore vs exploit paradox and deals with the trade-offs of finding new options versus exploiting known options. The paradox states that if you plan on being in the situation for a long time that you should spend time exploring. Alternatively, if you plan on being in the situation for a short time, you should go ahead and exploit whatever advantage you have.  In terms of this Celtics team, we can assume they will be around for a long time. They have great young players on favorable contracts and should be set up for years. Exploring lineup will pay dividends not only at the end of this regular season and playoffs but also in the years to come.

Let’s try to optimize the Celtics lineup in three important areas including starting lineup, second unit, and closers. Taking some advice from fellow Nylon Calculus contributor I will be using pbpstats to look at lineup performance from last season. The main problem with this approach is that the Celtics were devastated by injuries last year. Gordon Hayward played in only five minutes before breaking his leg and Kyrie Irving missed most the second half of the season and all of the playoffs.

With the starters, I am assuming that Kyrie Irving and Al Horford are locked into the starting lineup. The rest of the three spots are up for grabs with Hayward, Marcus Smart, Jayson Tatum, Jaylen Brown, and even Baynes and Marcus Morris available. To investigate this, I examined the lineups in the first six minutes of the game with Kyrie and Horford on the floor. By far the most used and one of the most successful starting lineups included Brown, Tatum, Horford, Irving, and Baynes. Surprisingly Baynes was a part of this lineup and will be interesting to see if this lineup is used this year with a player like Hayward available. But imagine a second group that involves the scoring ability of Hayward with the toughness of Morris and Smart and energy of Rozier. Whether it is Hayward or one of either Brown or Tatum coming off the bench they could be an asset and do some real damage against other second teams.

Speaking of second teams I also looked at lineups in the second half of the first quarter and the start of the second. These second teams, while not the most important time of the game, can play a valuable role in keeping leads or limiting deficits. The second quarter is the lowest plus-minus of any quarter for the Celtics last year and could be one of the biggest areas of improvement. Their best second team from last year is made up of all returning players of Rozier, Ojeleye, Theis, Morris, and Smart. This could be a quality lineup used in the second quarter but Celtics will also have Hayward back. Whether it is Hayward, Brown, or Tatum playing the role of the sixth man; we could expect one of these lineups to include the new sixth man instead of some of the departed pieces from last year.

Finally, the end of the game the most important time of the game and the best players need to be playing.  The closers were limited to the last six minutes of the fourth quarter with the score within 10 points. A popular closing lineup was also a popular starting lineup with Brown, Tatum, Horford, Irving, and Smart. However, this lineup had a -5.3 differential per 100 possessions.  Swapping out Brown for Rozier lead to +16.9 differential over 27 minutes. Also, the clutch playoff fourth quarter was evident with Rozier, Brown, Tatum, Horford, and Morris who were + 24.1 per 100 possessions. The plot below summarizes the lineups by +/- per 100 possessions (PM) and eFG with an emphasis on the most successful lineups with the most possessions.

Expecting to have Hayward and Irving for most of the season, coach Brad Stevens will have a multitude of lineup configurations. As with most lineup analysis, a lot of the lineups do not have enough of a sample. With most of the lineups having about 20 minutes of total play together, we cannot assume that the lineups will perform at the same level in the future.

The Celtics have a lot of talent and finding their best lineups will play an important role in their success. Exploring lineup performance and exploiting their lineup advantage will be an interesting part of watching the Celtics this season.