Model cards Positional Depth

Positional Depth

Every club’s controlled major leaguers and prospects placed in 21 slots for 2027–31, each season graded against the expected WAR a typical regular brings there.

Supports: Where the organization is deep enough to trade from, where it will be bare and must buy, draft or develop, and which clubs fit as trade partners at each spot. Refresh: Every build; bars each offseason Open Positional Depth

What it predicts

For each club, position and season: the expected WAR the club controls at the spot against a regular’s bar (need), and the regular-caliber players left without a spot (surplus). Tested on whether spots it grades short were filled from outside the organization the next season.

Inputs

Contract control by season and status (RosterResource, as in Contract Studio), with the club-option rule and option probabilities
The Winter Plan’s depth-chart engine: value at each eligible position (Fielding Lab) times that role’s playing time; the contract projections for later seasons (a two-way player’s hitting and pitching each aged on its own curve, since the contract projection sums them into one figure)
FanGraphs’ 2026 prospect board (grade, position, ETA, level) and FanGraphs’ July 2026 prospect-value table (Ben Clemens)
Every club’s regulars and their pre-season projections, 2024–26, for the bars; 2025 opening-day 40-man rosters for the backtest
Pittsburgh: the Winter Plan’s moves (decisions feed); every other club: its roster as it stands

Method

Major leaguers: guaranteed, arbitration and pre-arbitration seasons count in full; a club option only when Contract Studio’s rule (value at least the cost at the site’s $/WAR) says exercise; player options and opt-outs at Φ((salary − value) ÷ σ), σ from how much projections move year to year; mutual options at most 0.25; vesting options 0.5. A 40-man player with no contract row gets the least control his debut allows, marked as an estimate
Prospects: FanGraphs’ ETA, never earlier than the level allows (Triple-A and Double-A 2027, High-A and Single-A 2028, complex 2029, DSL 2030). Expected WAR a season = the FV table’s team-control WAR ÷ 6, bust-weighted; the chance of a regular = that ÷ a regular’s WAR at the grade, capped at 0.9. The first two seasons are ramped by the measured on-time arrival share and debut-year playing time. Positions from the board plus the moves the fielding model allows (a shortstop can play second or third)
2026 draftees (no ETA yet): 2026 + the board’s median signing-to-ETA lag for 2024–25 draftees from the same kind of school (Stats API draft: high school or college) at the grade his trade value sits at (FV 55+, 47–50, 42–45, below), never before High-A’s floor (2028) for a college pick or the complex’s (2029) for a high-school one; his post-draft stint at the complex doesn’t set his floor
Slots: C through DH, five starters, three high-leverage and four middle relievers, filled each season by a deterministic optimal assignment (Hungarian) with a replacement-level filler, so a slot can stay open
Bars: the median pre-season projection of every club’s regular at the spot, 2024–26, valued exactly like the supply (actual WAR runs higher: the players who kept the job were partly the ones having good years); published beside the actual medians
Need = 1 − assigned expected WAR ÷ bar, clipped to 0–1 (open = 1; the rotation and bullpen take their weakest slot). Covered at or above the bar, below at half or more, thin under half, open with nobody above replacement
Surplus: each regular-caliber player (at least half the bar at his best spot) with no slot, or playing elsewhere while someone at least as good, at his own best spot, holds his, adds min(1, WAR ÷ bar) scaled by what the move costs him; capped at 1. A player moved to fill a hole in a chain of moves (with no third baseman: the shortstop to third, the center fielder to short) is a flex note, not a surplus. The best unslotted catcher is the required backup, never a surplus
Backtest: rebuilt as of the 2024–25 offseason (2025 opening-day rosters, projections from 2022–24, bars from 2022–24) and scored on who held each spot in 2026, with 80% Wilson intervals; prospect ETAs checked on 2026 debuts. The page’s method note quotes every share from the build

How well it works

Computed by the pipeline from the data on this site

Backtest: each 2026 spot graded as of the 2024–25 offseason (2025 opening-day 40-man rosters; projections from 2022–24 data; control for 2026 as known then), then who actually held it. Share filled from outside the organization

open 62% (80% range 50%–73%, n = 29)thin 36% (80% range 29%–44%, n = 69)below the bar 38% (80% range 31%–46%, n = 63)covered 24% (80% range 19%–29%, n = 109) Open spots were filled from outside far more often than covered ones; thin and below-the-bar spots weren’t told apart.

Does a higher need score mean an outside fill? (chance a spot filled from outside had the higher score)

63% (80% range 58%–67%; 93 outside fills, 177 inside; 50% would be no signal)

The model’s starter was the actual 2026 regular

covered 56% (80% range 50%–62%, n = 109)below the bar 22% (80% range 16%–30%, n = 63)thin 10% (80% range 6%–16%, n = 69)

Rotations: starter need against outside starters in the 2026 top five

rank correlation 0.30 (the needier half used 2.07 outside starters, the other half 1.73)

Surplus players: did they change clubs, and were they blocked?

Changed clubs 31% (80% range 22%–41%, n = 39) against 27% (80% range 24%–30%, n = 359) for other regular-caliber players A 2026 regular at their best spot 18% (80% range 11%–27%, n = 39) against 49% (80% range 46%–53%, n = 359) Surplus marks blocked depth a club could trade; it doesn’t predict who is traded.

Prospect arrival: FanGraphs’ 2026 ETAs against 2026 debuts

ETA 2026 arrived in 2026 53% (80% range 49%–56%, n = 263): Triple-A 69% (80% range 64%–74%, n = 139), Double-A 38% (80% range 31%–44%, n = 96) ETA 2027 arrived a year early 15% (80% range 13%–18%, n = 270); from High-A, the next season 3% (80% range 2%–5%, n = 198) One season of outcomes, and the arrival ramp is fit on it: a calibration, not an out-of-sample test.

Agreement with the Winter Plan’s 2027 depth chart

9 of 9 lineup starters at the plan’s positions; the plan’s five starters in the rotation

Not validated

The arrival ramp's on-time share and debut-year playing time are measured on 2026 (the only season with a board), so the ETA calibration above is the ramp's own data, not an out-of-sample test of it.
Prospect arrivals beyond one season: the 2026 board is the only list in our data, so ETA is checked on 2026 debuts only.
2026 draftees' arrival: the board's signing-to-ETA lag for 2024–25 draftees by school and grade band is FanGraphs' forecast for those classes, not their measured arrivals (one or two seasons after signing is too soon to see when a class arrives), and a draftee's grade is the one his trade value sits at, not a scouting grade; the High-A floor for a college pick is Keystone's reading.
The FV table's WAR and the chance of a regular are FanGraphs' and McDaniel's published figures, not re-estimated or calibrated on our data.
Player-option, mutual-option and vesting probabilities are model assumptions (σ from projection updates; vesting 0.5), not checked against past option decisions.
Surplus and need in 2028–31 depend on contract projections and aging that the backtest (2026 only) doesn't reach.

Data: contract control by season (the control mart Contract Studio reads: RosterResource contracts, 40-man rosters) and Keystone’s contract projections; the Winter Plan’s depth-chart engine; FanGraphs’ 2026 prospect board and Clemens’ July 2026 FV table; the Stats API 2026 draft and 2026 debuts; every club’s regulars 2024–26 with their pre-season projections (56–90 club-seasons a slot; 131 regulars with no MLB history left out); 2025 opening-day 40-man rosters for the backtest.

What it can't see

How we'd measure impact

Share of spots graded open or thin that the club fills from outside, season by season
Whether surplus players Pittsburgh shops bring more value than the package rule alone would ask
ETA calibration on each new prospect board