Loading field map
Sabermetrics is not simply the use of statistics in baseball. It is a field defined by a series of competing quantitative frameworks, each proposing a different answer to the same question: what measurable factors most reliably produce wins, and how should those measurements guide decisions? From the earliest attempts to model run scoring to today's player-tracking data, the history of sabermetrics is a story of frameworks that replaced, absorbed, or coexisted with one another, driven by the limitations and tensions of their predecessors.
The first systematic quantitative framework emerged from the work of Earnshaw Cook and George Lindsey in the 1960s. The Run Expectancy Framework modeled the expected number of runs a team would score from a given game state—inning, outs, base runners—and used those probabilities to evaluate tactical decisions such as when to bunt, steal, or sacrifice. Its contribution was to replace intuition with a probabilistic baseline. Yet the framework was narrow: it addressed only in-game tactics, not player evaluation or roster construction. It coexisted with the traditional scouting and managerial wisdom of the era, often dismissed as academic curiosity. The Run Expectancy Framework laid the mathematical foundation for later work, but its limited scope left a gap that the next framework would exploit.
Bill James transformed sabermetrics by expanding its scope from tactical decisions to player evaluation. Beginning with his self-published Baseball Abstracts in the late 1970s, James developed metrics such as Runs Created, which estimated a player's total offensive contribution from raw batting data. His framework was not merely a set of new stats; it was a coherent research program that used historical data to test conventional wisdom—questioning, for example, the value of batting average or the sacrifice bunt. Jamesian Sabermetrics coexisted with the Run Expectancy Framework (which remained useful for in-game strategy) but differed fundamentally in ambition: it aimed to measure player value holistically, not just situational run probability. Despite its influence among fans and a few front offices, James's work remained marginal within Major League Baseball for most of this period. The tension between his empirical findings and the industry's reliance on scouting and tradition set the stage for the next framework's more aggressive application.
Moneyball Analytics emerged when front offices began applying Jamesian insights to roster construction, most famously by the Oakland Athletics under general manager Billy Beane. This framework narrowed the focus from general research to market exploitation: identifying undervalued skills—especially on-base percentage—that the market had not yet priced correctly. Moneyball Analytics absorbed the metrics of Bill Jamesian Sabermetrics but added a competitive strategy layer: it treated player evaluation as an arbitrage opportunity. It coexisted with the older frameworks, but its success forced other teams to adopt similar methods, gradually eroding the inefficiencies it exploited. By the late 2000s, the insights of Moneyball had become common knowledge, and the framework's advantage faded. This absorption into the mainstream created pressure for a more sophisticated approach to player valuation.
Advanced Metric Sabermetrics arose in response to the limitations of Moneyball's simple metrics. Starting around 2000, analysts developed composite statistics such as Wins Above Replacement (WAR), Fielding Independent Pitching (FIP), and Weighted On-Base Average (wOBA). These metrics aimed to capture total player value in a single number, adjusting for park, league, and positional effects. This framework transformed sabermetrics from a hobbyist pursuit into an institutionalized practice: by the 2010s, every MLB team employed analytics departments, and front offices routinely used WAR and FIP in contract negotiations and roster decisions. Advanced Metric Sabermetrics did not replace Bill Jamesian Sabermetrics or Moneyball Analytics so much as absorb and systematize them. It preserved James's empirical spirit and Moneyball's focus on market efficiency, but it demanded more rigorous statistical methods and larger datasets. Today it remains the dominant framework for player valuation, though it now coexists with a newer, more granular approach.
Statcast Analytics represents a fundamental shift from event-based measurement to process-based measurement. Beginning in 2015, MLB installed optical tracking systems in every ballpark, capturing the trajectory of every pitch and batted ball in real time. This data allowed analysts to measure exit velocity, launch angle, sprint speed, and route efficiency—physical processes rather than just outcomes. New metrics such as expected batting average (xBA) and expected slugging percentage (xSLG) use these process data to separate skill from luck. Statcast Analytics coexists with Advanced Metric Sabermetrics, but the two frameworks have different commitments. Advanced Metric relies on composite stats derived from discrete events (hits, outs, walks); Statcast focuses on continuous physical measurements. This creates a living disagreement: should player evaluation prioritize what happened (WAR) or how it happened (exit velocity, launch angle)? Statcast has not replaced Advanced Metric; rather, it has added a new layer of analysis that challenges the older framework's assumptions about what data are most informative.
Today, the leading frameworks are Advanced Metric Sabermetrics and Statcast Analytics. They agree on the fundamental principle that data-driven decision-making outperforms pure intuition. Both frameworks reject the idea that traditional statistics like batting average or ERA are sufficient. Yet they disagree on the most useful level of analysis. Advanced Metric advocates argue that composite metrics like WAR provide a stable, interpretable summary of player value, while Statcast proponents contend that process data reveal underlying skill more accurately and allow for better predictions. This tension has spurred hybrid approaches that combine event-based and process-based data using machine learning. The field is no longer a single orthodoxy but a pluralistic landscape where different frameworks answer different questions—and where the next framework will likely emerge from the unresolved debate between what happened and how it happened.