{"id":{"repo_id":"harvard","oai_identifier":"oai:dash.harvard.edu:1/38811515"},"canonical_url":"https://search.dev.ndltd.org/etd/harvard/oai:dash.harvard.edu:1/38811515","repository":{"repo_id":"harvard","name":"Harvard University","base_url":"https://dash.harvard.edu/server/oai/request"},"display":{"title":"Evaluating Lineups and Complementary Play Styles in the NBA","abstract":"NBA coaches and general managers are tasked with building lineups and rosters that maximize their chances of winning. Further, basketball is a team sport where interactions between the players in a lineup can be integral to the success or failure of that lineup; managing a team is not as simple as building a lineup of the best possible players without considering the way in which their play styles interact with one another. However, the most prominent statistical methods for evaluating players make the explicit assumption that there are no interaction effects between the players on the court in producing point outcomes. This thesis introduces a model to predict the expected point differential between two five-player lineups in the NBA based on the players’ play styles and the way in which these play styles interact, where play styles are represented by a collection of statistics collected from play-by-play data to describe each player’s tendencies. This model, which is shown to give a significant predictive improvement over today’s standard adjusted plus/minus models for player evaluation, is then used to evaluate players, lineups, and teams.","abstract_html":"NBA coaches and general managers are tasked with building lineups and rosters that maximize their chances of winning. Further, basketball is a team sport where interactions between the players in a lineup can be integral to the success or failure of that lineup; managing a team is not as simple as building a lineup of the best possible players without considering the way in which their play styles interact with one another. However, the most prominent statistical methods for evaluating players make the explicit assumption that there are no interaction effects between the players on the court in producing point outcomes. This thesis introduces a model to predict the expected point differential between two five-player lineups in the NBA based on the players’ play styles and the way in which these play styles interact, where play styles are represented by a collection of statistics collected from play-by-play data to describe each player’s tendencies. This model, which is shown to give a significant predictive improvement over today’s standard adjusted plus/minus models for player evaluation, is then used to evaluate players, lineups, and teams.","abstract_has_math":false,"creators":["Goldberg, Matthew"],"institution":"Harvard College","degree_name":"AB","degree_level":"Undergraduate","degree_discipline":null,"degree_department":null,"school":null,"contributors":[],"advisors":[],"committee_chairs":[],"committee_members":[],"year":2017,"date_issued":"2017-10-13","date_published":"2017-10-13","updated_at":"2026-07-27T19:55:49Z","subjects":["Computer Science","Statistics"],"languages":["en"],"rights":[],"rights_urls":[],"identifier_entries":[]},"links":{"outbound_url":"http://nrs.harvard.edu/urn-3:HUL.InstRepos:38811515","outbound_label":"Repository record","outbound_source":"dc:identifier.uri"},"metadata_groups":[{"id":"people","label":"People","entries":[{"key":"dc:creator","label":"Author","values":["Goldberg, Matthew"]}]},{"id":"academic_context","label":"Academic Context","entries":[{"key":"dc:date.accessioned","label":"Dc Date Accessioned","values":["2019-03-26T10:57:27Z"]},{"key":"dc:date.available","label":"Dc Date Available","values":["2019-03-26T10:57:27Z"]},{"key":"dc:date.issued","label":"Date","values":["2017-10-13"]},{"key":"dc:type","label":"Dc Type","values":["Thesis or Dissertation"]},{"key":"thesis:degree_level","label":"Degree Level","values":["Undergraduate"]},{"key":"thesis:degree_name","label":"Degree Name","values":["AB"]},{"key":"thesis:institution_name","label":"Thesis Institution Name","values":["Harvard College"]}]},{"id":"subjects_keywords","label":"Subjects and Keywords","entries":[{"key":"dc:subject","label":"Dc Subject","values":["Computer Science","Statistics"]}]},{"id":"language_rights","label":"Language and Rights","entries":[{"key":"dc:language.iso","label":"Language (ISO)","values":["en"]}]},{"id":"identifiers","label":"Identifiers","entries":[{"key":"dc:identifier.uri","label":"Identifier URI","values":["http://nrs.harvard.edu/urn-3:HUL.InstRepos:38811515"]}]},{"id":"additional","label":"Additional Metadata","entries":[{"key":"dc:description.abstract","label":"Abstract","values":["NBA coaches and general managers are tasked with building lineups and rosters that maximize their chances of winning. Further, basketball is a team sport where interactions between the players in a lineup can be integral to the success or failure of that lineup; managing a team is not as simple as building a lineup of the best possible players without considering the way in which their play styles interact with one another. However, the most prominent statistical methods for evaluating players make the explicit assumption that there are no interaction effects between the players on the court in producing point outcomes. This thesis introduces a model to predict the expected point differential between two five-player lineups in the NBA based on the players’ play styles and the way in which these play styles interact, where play styles are represented by a collection of statistics collected from play-by-play data to describe each player’s tendencies. This model, which is shown to give a significant predictive improvement over today’s standard adjusted plus/minus models for player evaluation, is then used to evaluate players, lineups, and teams."]},{"key":"dc:format.mimetype","label":"Dc Format Mimetype","values":["application/pdf"]},{"key":"dc:title","label":"Title","values":["Evaluating Lineups and Complementary Play Styles in the NBA"]}]}],"canonical_facts":{"dc:creator":["Goldberg, Matthew"],"dc:date.accessioned":["2019-03-26T10:57:27Z"],"dc:date.available":["2019-03-26T10:57:27Z"],"dc:date.issued":["2017-10-13"],"dc:description.abstract":["NBA coaches and general managers are tasked with building lineups and rosters that maximize their chances of winning. Further, basketball is a team sport where interactions between the players in a lineup can be integral to the success or failure of that lineup; managing a team is not as simple as building a lineup of the best possible players without considering the way in which their play styles interact with one another. However, the most prominent statistical methods for evaluating players make the explicit assumption that there are no interaction effects between the players on the court in producing point outcomes. This thesis introduces a model to predict the expected point differential between two five-player lineups in the NBA based on the players’ play styles and the way in which these play styles interact, where play styles are represented by a collection of statistics collected from play-by-play data to describe each player’s tendencies. This model, which is shown to give a significant predictive improvement over today’s standard adjusted plus/minus models for player evaluation, is then used to evaluate players, lineups, and teams."],"dc:format.mimetype":["application/pdf"],"dc:identifier.uri":["http://nrs.harvard.edu/urn-3:HUL.InstRepos:38811515"],"dc:language.iso":["en"],"dc:subject":["Computer Science","Statistics"],"dc:title":["Evaluating Lineups and Complementary Play Styles in the NBA"],"dc:type":["Thesis or Dissertation"],"thesis:degree_level":["Undergraduate"],"thesis:degree_name":["AB"],"thesis:institution_name":["Harvard College"]},"updated_at":"2026-07-27T19:55:49Z"}