Rosternomics

TAB — Talent Above Budget

How much roster talent a team assembled above what its payroll predicted — the roster-construction half of front-office skill.

What it measures

WAB asks "did the wins show up?"; TAB asks "did you build the talent?" They differ by exactly one thing — the luck in converting talent into wins (run sequencing, one-run games, bullpen timing) — so WAB − TAB is that conversion luck, and TAB strips it out. That makes TAB the cleaner, stickier read on the executive. (It isn't fully luck-proof: an injury lowers a player's WAR, so injuries hit TAB and WAB alike — TAB only removes the conversion layer.)

The model

Same setup as WAB, but the response is the team's season-pace roster WAR instead of win%:

\[ \text{WAR}_{162} = \text{teamWAR}\times\frac{162}{G}, \qquad \text{WAR}_{162} = a_w + b_w\,\ln(\text{relpay}) \] \[ \text{TAB} = \text{WAR}_{162} - [\,a_w + b_w\,\ln(\text{relpay})\,] \]

Fitted: WAR₁₆₂ = 34.3 + 13.0·ln(relpay). Because WAR is built from hundreds of plate appearances and innings, it is far less noisy than W–L record — so TAB is a stickier skill signal (within-GM autocorrelation ~0.48 vs WAB's ~0.38).

TAB fit and TAB-vs-WAB

TAB vs WAB

In the right-hand panel, a team above the diagonal won more than its talent (good wins-luck); below, it won less. The vertical gap between a team's TAB and WAB is exactly the part a front office doesn't control. Use TAB to judge the executive, WAB to judge the season's bottom line.

Is it a skill, and does it predict?

The receipts — playoff rate by single-season TAB

Every full team-season (1985–2025) bucketed by its TAB that year:

TAB that seasonteam-seasonsmade playoffswon pennantwon WS
> +155080%24%10%
+5 to +1531155%14%7%
−5 to +544520%4%2%
−15 to −52913%1%1%
< −15612%0%0%

Same monotonic climb as WAB but a touch shallower at the extremes — because TAB strips out the wins-luck that lets a merely-good roster overshoot into the playoffs. It's measuring the talent you built, not the record it happened to produce.

Worked example — the 2002 Oakland A's

Inputs: full-roster fWAR = 47.2 (full season, so WAR₁₆₂ = 47.2). relpay 0.70 → \(\ln = -0.357\).

1. predicted WAR₁₆₂ \(= 34.3 + 13.0\times(-0.357) = 29.7\)

2. \(\text{TAB} = 47.2 - 29.7 = \mathbf{+17.5}\)

Oakland assembled ~18 WAR more talent than a bottom-tier payroll should buy. Note TAB (+17.5) sits below their WAB (+25.9): the gap is the Pythagorean/sequencing luck TAB deliberately strips out.

Fit on full team-seasons (≥100 games), 1985–2024/25. All figures are franchise-level outcomes credited to the decision-maker in the relevant year — see the GM profiles for per-executive numbers.