How Roster Construction Changed When Nobody Anymore Talks About Batting Average
I spent years watching front offices make decisions based on things like OPS+ and wOBA without really understanding why those metrics mattered more than the old ones. Then I started working with players who actually had to perform under the systems these metrics predicted. It is a weird experience when you realize the spreadsheet version of a player and the human version of a player are often different people entirely. The Dodgers became one of the most prominent examples of what happens when an organization commits fully to analytics-driven roster construction. They did not invent the approach, but they executed it at a scale most teams simply cannot match. Their front office treats every roster spot as a resource allocation problem, and Fangraphs and Baseball Savant provide the data infrastructure that makes that possible. Here is how it actually works in practice, not the simplified version you see in articles written by people who have never opened a query builder.
The Data Layer Behind Roster Decisions
Fangraphs aggregates and normalizes tracking data across millions of plate appearances and defensive plays. Baseball Savant pulls directly from Statcast's laser and camera systems to give you exit velocity, launch angle, sprint speed, and spin rate for every batted ball and pitch thrown in the majors. When a roster decision needs to be made about a 40th man on the margin, those two sources together can answer questions that used to require months of subjective evaluation. I once spent three days trying to figure out whether a specific reliever was actually getting better swings and misses or just benefiting from weak contact around the zone. Fangraphs' strikeout rate and whiff rate metrics pointed one direction, but Baseball Savant's pitch-level data told a different story. The workaround was pulling the actual pitch movement values and cross-referencing them with batting average on balls in play against each pitch type. What looked like a groundball pitcher on Fangraphs was actually a flyball pitcher whose results were inflated by elite defense behind him. That distinction changes everything about how you plan to use him.
What This Looks Like For a Front Office
The Dodgers use these tools continuously throughout the year, not just during the winter meetings. Here is the practical workflow most competitive organizations follow now. First, they define their roster requirements using market value analysis. Fangraphs' WAR calculations, specifically fWAR versus bWAR, give you a standardized currency for comparing players across positions and league contexts. A player worth 3.0 fWAR is treated as approximately equivalent in value regardless of whether he is a catcher in Atlanta or an outfielder in Toronto. This standardization is what allows teams to trade and acquire players without getting bogged down in positional debates. Second, they identify undersold skills. Baseball Savant's tracking data reveals abilities that traditional statistics miss entirely. Sprint speed matters for base running value. Launch angle distribution matters for home run probability. Pitch spin rates matter for predicting which pitchers will regress. Teams that notice these patterns early gain an edge before the rest of the league catches up. The Dodgers identified this approach years ago and have been compounding that advantage ever since.
Get the Full Details
Third, they stress-test every assumption. This is where most amateur analyses fail. A player might look like a bargain because his ERA is low, but his strand rate is unsustainable, or his peripheral metrics suggest upcoming regression. Fangraphs provides regression models and expected statistics that account for these factors. Baseball Savant provides the raw tracking data to verify whether the regression signals are real or statistical noise.
The Pitfalls Everyone Ignores
I need to be blunt about the limitations because people who only read the success stories give you a completely distorted picture of how this works. The biggest problem is sample size. A player's Statcast numbers over forty games mean almost nothing. The league-wide averages stabilize around two hundred plate appearances or eighty pitches for most metrics, but individual player evaluations require substantially more data. I have seen multiple teams make significant roster commitments based on small-sample Statcast outliers that regressed to mediocrity within three weeks. The data is not wrong. The interpretation was wrong. Another issue is context stripping. Advanced metrics evaluate individual performance in isolation from game state, ballpark factors, and roster construction constraints. A player might have excellent underlying numbers but fit poorly into a team's specific strategic needs. The Dodgers understand this because they have the roster depth to be selective. Smaller market teams sometimes overpay for statistical value without accounting for how that value translates into actual wins in their particular context.
There is also the replication problem. Fangraphs and Baseball Savant both publish their methodologies, but neither one gives you direct access to their underlying code or raw database queries. If you want to run your own models or modify their calculations, you are working with cleaned and aggregated data rather than the original events. This limitation matters more than most people realize when you are trying to validate a specific finding or adapt a model to a different league context.
What You Can Actually Do With This Information
If you are evaluating players or building a roster model, start with Fangraphs for the big-picture projections and historical context. Use their depth charts, minor league databases, and draft tracker to identify trends across the system. Then use Baseball Savant to drill into the tracking data for any player who looks interesting. The combination of these two sources covers approximately ninety percent of what a front office uses for daily decision-making. For the Dodgers specifically, their roster construction philosophy has evolved from aggressive spending to more nuanced resource management. The shift became noticeable around 2022 and 2023 when they began prioritizing internal development and cost-controlled talent over free-agent acquisitions. Fangraphs' positional value charts and payroll efficiency metrics reflect this transition clearly. The underlying approach has not changed. They are still optimizing for wins, but they have gotten more sophisticated about where that optimization happens within their resource constraints. The modern game is not reshaped by any single metric or platform. It is reshaped by the institutional habit of questioning every assumption with data before making a decision. Fangraphs and Baseball Savant are simply the tools that make that habit executable at scale. The teams that adopt the habit systematically outperform the teams that treat analytics as decoration.
