Justin Jacobs, PhD
The only possession-level individual impact database for the pre-play-by-play NBA era — built game by game from video, by a government-decorated research scientist with over a decade of NBA front office experience.
Justin Jacobs is a research statistician whose work spans national defense, professional basketball, and the intersection of machine learning and spatio-temporal analysis. He is a recipient of the Presidential Early Career Award in Science and Engineering — the highest honor the U.S. government bestows on early-career scientists — and holds a PhD in Statistics from the University of Maryland – Baltimore County and an MS in Mathematics from the University of Wisconsin – Milwaukee.
Before turning his research focus toward basketball, Justin spent nearly a decade as a research statistician within the Department of Defense, including a position as Principal Research Statistician at Sandia National Laboratories in Livermore, California. His defense research produced over 30 publications and a patent on pseudo-GPS methods, with work spanning spatio-temporal statistics, manifold learning, ranking analytics, streaming analysis, and recommender systems.
Since the 2012–13 NBA season, Justin has worked with multiple NBA front offices on analytic development, coaching strategy, player valuation, and performance modeling through machine learning and artificial intelligence. From January through August 2018, he joined the Orlando Magic full-time as Senior Basketball Researcher. He has also collaborated with industry research partners including Microsoft’s NBA-AI initiative, and worked with two NCAA programs between 2014 and 2016 on player movement and rotation quality metrics.
Justin is a member of the American Statistical Association and a former NCAA basketball player. Squared Statistics is his independent research platform, used to publish original basketball analytics, methodology deep-dives, and the ongoing Historical RAPM Project.
RAPM requires play-by-play stint data that the NBA did not systematically record until the late 1990s. That left an entire era — the Bad Boys, Showtime, Jordan’s ascent — analytically dark. The Historical RAPM Project is the only effort to reconstruct that data from primary sources.
Every game in the database was manually annotated from video: every substitution logged, every possession counted. The result is the only possession-level lineup dataset for the 1969–1996 era in existence. Nine seasons are currently published, with the 1985–1996 window as the primary reconstruction focus. Playoff coverage stands at approximately 70%. The project is ongoing.
Cleaned season-level player and team summaries derived from video-based lineup reconstruction. All playing time is corrected using inferred on-court presence — more accurate than raw box-score minutes. Designed for researchers, writers, and analysts who need accurate season summaries without possession-level detail.
- Player–season summaries with corrected minutes played
- Offensive and defensive possession counts
- Points scored and allowed while on court
- Season-aggregated on/off and net rating summaries
- Team-level season totals with usage context
- Quarterly updates as new games are reconstructed
Season-level inference of player rotation structure and in-game usage patterns. Characterizes not just how much players played, but when they were used — starter vs. bench roles, closing tendencies, quarter-level usage. Designed for analysts who need rotation structure without full possession-level data.
- Time-aligned player presence profiles aggregated across the season
- Quarter-level and game-clock usage summaries
- Inferred starter, bench, and closing roles
- Rotation stability and variability measures
- Team-level rotation structure indicators
The atomic record from which all summaries and impact models are derived. Full 5-on-5 possession-level stint data with complete lineup context and scoring outcomes — the only dataset of its kind for this era. Designed for professional analysts and modeling teams who want to perform their own impact estimation, matchup analysis, or simulation.
- Game-dated possession-level stints with complete offensive and defensive lineups
- Corrected playing time within each stint
- Offensive and defensive possession counts and points per stint
- Explicit substitution boundaries defining lineup changes
- Suitable for direct use in ridge regression, Bayesian models, and simulation
Pre-computed player impact estimates built directly from the Tier 3 stint dataset. Saves modeling teams from implementing the full estimation pipeline while providing validated, uncertainty-quantified RAPM estimates ready for comparison, ranking, and downstream analysis. Available only alongside or after Tier 3 purchase.
- Pre-computed RAPM coefficients (offensive, defensive, overall)
- Associated uncertainty bounds and confidence intervals
- Season-level impact estimates aligned with Tier 3 data
- Model-based summaries suitable for ranking and comparison
Bespoke analytic services for organizations that need tailored analysis, custom modeling, or expert interpretation beyond standardized data releases. Draws on over a decade of applied experience across multiple NBA organizations, including work in spatiotemporal analytics, draft and free agency modeling, and collaboration with Microsoft’s NBA-AI initiative. The $500 scoping fee covers an initial consultation to define objectives and deliverables, and is credited in full toward any subsequent project agreement.
- NBA front offices seeking historical player evaluation support
- Sports media organizations commissioning original research
- Academic institutions requiring expert collaboration
- Documentary and production teams needing analytical narrative support
All datasets are available for the 1984–85 through 1995–96 NBA seasons on a per-season basis. After purchase, email the season you wish to receive along with proof of purchase to the address on the receipt. Quarterly updates reflecting newly reconstructed games are included at no additional cost.
Questions about which tier is right for your use case? Contact via the address on your receipt or reach out through the blog.
Pingback: The Bad Boys by the Numbers: What 1988–89 Lineup Data Tells Us About Detroit’s Dynasty | Squared Statistics: Understanding Basketball Analytics