{"689023":{"#nid":"689023","#data":{"type":"news","title":"Bracketology Driven by Data ","body":[{"value":"\u003Cdiv\u003E\u003Cp\u003ETens of millions of brackets have been filled out ahead of the NCAA men\u2019s and women\u2019s basketball tournaments. Some fans will choose winners based on the higher seed, others will try to predict shocking upsets, and some may choose who advances based on which mascot would win a fight, but a Georgia Tech professor has his bracket down to a (data) science. \u0026nbsp;\u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cp\u003ESince 2004, Joel Sokol, director of the Master of Science in Analytics program and the Harold E. Smalley Professor in the H. Milton Stewart School of Industrial and Systems Engineering, has used a pair of analytic methods \u2014 logistic regression and Markov chains (LRMC) \u2014 to determine the best teams in college basketball. This year, \u003Ca href=\u0022https:\/\/www2.isye.gatech.edu\/~jsokol\/lrmcclassic\/\u0022 rel=\u0022noreferrer noopener\u0022 target=\u0022_blank\u0022\u003ESokol\u2019s LRMC rankings\u003C\/a\u003E project the \u003Ca href=\u0022https:\/\/www2.isye.gatech.edu\/~jsokol\/profspicks\/profspicks26-c.pdf\u0022 rel=\u0022noreferrer noopener\u0022 target=\u0022_blank\u0022\u003EMichigan Wolverines to cut down the nets\u003C\/a\u003E at the end of the men\u2019s tournament and the \u003Ca href=\u0022https:\/\/www2.isye.gatech.edu\/~jsokol\/profspicksW\/profspicks26w-c.pdf\u0022 rel=\u0022noreferrer noopener\u0022 target=\u0022_blank\u0022\u003EConnecticut Huskies as the last team standing in the women\u2019s field\u003C\/a\u003E. \u0026nbsp;\u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cp\u003EThe algorithm compares all 350-plus Division I basketball teams against each other simultaneously during the regular season and calculates probabilities based on simple data points \u2014 who won each game, by how much, and where it was played. When the madness of March begins, Sokol\u2019s bracket forgoes the seeds assigned to teams and fills out his bracket based on the LRMC rankings.\u0026nbsp; \u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cp\u003EModels used by the tournament selection committee \u2014 \u003Ca href=\u0022https:\/\/www.ncaa.com\/news\/basketball-men\/article\/2022-12-05\/college-basketballs-net-rankings-explained\u0022 rel=\u0022noreferrer noopener\u0022 target=\u0022_blank\u0022\u003ENET\u003C\/a\u003E, \u003Ca href=\u0022https:\/\/www.ncaa.com\/news\/basketball-men\/article\/2022-02-09\/mens-college-basketball-rankings-what-kpi\u0022 rel=\u0022noreferrer noopener\u0022 target=\u0022_blank\u0022\u003EKPI\u003C\/a\u003E, \u003Ca href=\u0022https:\/\/kenpom.com\/\u0022 rel=\u0022noreferrer noopener\u0022 target=\u0022_blank\u0022\u003EKenPom\u003C\/a\u003E \u2014 measure advanced metrics like strength of schedule, possession-by-possession efficiency, opponent quality, and more, but Sokol, with expertise in sports analytics and data science, says the LRMC shows the value of simple data and a large sample size.\u0026nbsp; \u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cp\u003E\u201cThe LRMC can hold its own against those models that are based on much more advanced metrics than just scoreboard data. They may look at all kinds of information, from efficiencies down to individual player performance, but the message really is that if you have a good set of simple data, that\u2019s enough if you know how to interpret it.\u201d\u0026nbsp;\u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cp\u003ESokol compares his algorithm to nearly 100 other ranking systems and says the LRMC is often among the top performers, with the higher-ranked teams (in the LRMC rankings) winning approximately 75% of the time \u2014 a statistic that holds true in the NCAA Tournament. Sokol says that 25% of tournament games result in an upset. \u0026nbsp;\u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cp\u003EFor 2026, Sokol\u2019s projections predict that all eight No. 1 seeds \u2014 four in both the men\u2019s and women\u2019s tournaments \u2014 will reach the Final Four, but it\u2019s not always a guarantee that the highest seeds make it out of their respective regions. The inaugural LRMC rankings accurately predicted the No. 3-seeded Yellow Jackets\u2019 Final Four run in 2004 \u2014 one of the only predictive models to do so.\u0026nbsp; \u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cp\u003ESokol got the idea to compile the LRMC rankings one year before Tech\u2019s run to the national championship game, when the Yellow Jackets were left out of the NCAA Tournament as a bubble team, largely because of a December buzzer-beater loss to Tennessee. Since the first set of rankings, machine learning and artificial intelligence have become more accessible, yet Sokol says ChatGPT and other large language models (LLMs) aren\u2019t quite ready to handle the level of analysis required to shape the rankings.\u0026nbsp; \u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cp\u003E\u201cThese LLMs are good at sounding good, but they\u0027re not so good at doing these complex quantitative tasks,\u201d he said. \u0026nbsp;\u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E\u003Cdiv\u003E\u003Cp\u003EUltimately, though, luck is often a stubbornly unquantifiable factor when filling out a bracket, no matter the formula used to make selections, and the odds of filling out a perfect bracket are all but \u003Ca href=\u0022https:\/\/www.ncaa.com\/news\/basketball-men\/bracketiq\/2026-02-18\/perfect-ncaa-bracket-absurd-odds-march-madness-dream\u0022 rel=\u0022noreferrer noopener\u0022 target=\u0022_blank\u0022\u003Ea statistical impossibility\u003C\/a\u003E. \u0026nbsp;\u003C\/p\u003E\u003C\/div\u003E","summary":"","format":"limited_html"}],"field_subtitle":[{"value":"For two decades, a Georgia Tech professor has used simple data to track the best teams in college basketball and predict who will win the NCAA Tournament.   "}],"field_summary":[{"value":"\u003Cp\u003EFor two decades, a Georgia Tech professor has used simple data to track the best teams in college basketball and predict who will win the NCAA Tournament. \u0026nbsp;\u0026nbsp;\u003C\/p\u003E","format":"limited_html"}],"field_summary_sentence":[{"value":"For two decades, a Georgia Tech professor has used simple data to track the best teams in college basketball and predict who will win the NCAA Tournament.   "}],"uid":"36418","created_gmt":"2026-03-18 20:24:38","changed_gmt":"2026-03-27 14:20:39","author":"sgagliano3","boilerplate_text":"","field_publication":"","field_article_url":"","location":"Atlanta, GA","dateline":{"date":"2026-03-18T00:00:00-04:00","iso_date":"2026-03-18T00:00:00-04:00","tz":"America\/New_York"},"extras":[],"hg_media":{"679681":{"id":"679681","type":"image","title":"Joel Sokol","body":"\u003Cp\u003EJoel Sokol, director of the Master of Science in Analytics program and the Harold E. Smalley Professor in the H. Milton Stewart School of Industrial and Systems Engineering at Georgia Tech.\u0026nbsp;\u003C\/p\u003E","created":"1773865550","gmt_created":"2026-03-18 20:25:50","changed":"1773865550","gmt_changed":"2026-03-18 20:25:50","alt":"Joel Sokol","file":{"fid":"263871","name":"12C3046-P1-001.jpg","image_path":"\/sites\/default\/files\/2026\/03\/18\/12C3046-P1-001.jpg","image_full_path":"http:\/\/hg.gatech.edu\/\/sites\/default\/files\/2026\/03\/18\/12C3046-P1-001.jpg","mime":"image\/jpeg","size":2410903,"path_740":"http:\/\/hg.gatech.edu\/sites\/default\/files\/styles\/740xx_scale\/public\/2026\/03\/18\/12C3046-P1-001.jpg?itok=b7bFdqK3"}}},"media_ids":["679681"],"groups":[{"id":"1237","name":"College of Engineering"},{"id":"658168","name":"Experts"},{"id":"1214","name":"News Room"},{"id":"1188","name":"Research Horizons"},{"id":"1242","name":"School of Industrial and Systems Engineering (ISYE)"}],"categories":[],"keywords":[{"id":"62061","name":"March Madness"},{"id":"181299","name":"ncaa tournament"},{"id":"12204","name":"men\u0027s basketball"},{"id":"4811","name":"women\u0027s basketball"},{"id":"79951","name":"college basketball"}],"core_research_areas":[],"news_room_topics":[{"id":"71901","name":"Society and Culture"}],"event_categories":[],"invited_audience":[],"affiliations":[],"classification":[],"areas_of_expertise":[],"news_and_recent_appearances":[],"phone":[],"contact":[{"value":"\u003Cp\u003E\u003Ca href=\u0022mailto:steven.gagliano@gatech.edu\u0022\u003ESteven Gagliano\u0026nbsp;\u003C\/a\u003E\u003Cbr\u003EInstitute Communications\u003C\/p\u003E","format":"limited_html"}],"email":[],"slides":[],"orientation":[],"userdata":""}}}