Franchise Cornerstone — Duncan Index

What We Measure Over a Career

The seasonal composite measures one “Duncan-shaped” year. The career index measures the longitudinal signal the seasonal number cannot carry on its own: who actually was Duncan-shaped year after year, by minutes- weighted average, for season after consistent season. The top of the all-time list is itself a tight cluster — Jokić clearly the inheritor (and perhaps elevator) of the Duncan mantle, LeBron the longest career on record, Garnett, Duncan himself, all within a few points of each another.¹

An important note: the Institute’s data archive begins with the 1996–97 season. Stockton, Pippen, Hakeem, Malone, and a generation of pre-1997 Duncan- archetype anchors are impacted (or completely omitted) by this constraint. The research committee has determined that presenting the data with all the nuance available supersedes the desire to make it fully cross-generational.²

What We Are Measuring

• impact
  Combined offensive + defensive RAPTOR (538's published historical figures, 1996–present), z-scored against league baselines. A two-way multiplier rewards the rare both-ends-elite shape that defines the archetype. A floor- and ceiling-raising signal — drawn from existing HWCC fields — credits players whose teams are built around their habits, even when on/off swings are muted by team depth.

  off_z   = max(0, raptor_offense / 3.0)
  def_z   = max(0, raptor_defense / 2.5)
  total_z = max(0, (raptor_offense + raptor_defense) / 4.0)
  
  two_way   = 1 + 0.40 × min(off_z, def_z, 1.0)
  both_ends = min(floor_raising, ceiling_raising)
  fr_cr_z   = clamp((both_ends − 50) / 30, 0, 1.5)
  
  impact = clamp(total_z × two_way + 0.40 × fr_cr_z, 0, 3)

• consistency
  Coefficient of variation in scoring across the regular season, plus the rate of single-digit games — the “every tip you knew what you’d get” signal. A separate teammate-sensitivity component credits players who play well across many lineups.

  cv_z         = clamp((0.45 − ppg_cv)              / 0.15, 0, 1)
  under10_z    = clamp((0.10 − games_under_10_rate) / 0.10, 0, 1)
  stat_consist = 0.60 × cv_z + 0.40 × under10_z
  
  lineup_consist = clamp(1 − teammate_sensitivity / 75, 0, 1)
  
  consistency = max(0.30, 0.55 × stat_consist + 0.45 × lineup_consist)

• minutes-played-scale
  A soft multiplicative ceiling: low-minutes seasons can't top the leaderboard. Floor at 0.50 so elite per-minute players still earn a real score; no hard gate.

  mpg_scale = clamp((mpg − (μ_mpg − 0.3 × σ_mpg)) / σ_mpg, 0.50, 1.0)
              anchored on 2015-16:  μ_mpg = 23.96,  σ_mpg = 7.10

• usage discipline
  The Duncan archetype is high-impact at moderate-to-high usage, not Harden-style ball-dominance. The dock starts above the league mean by a full standard deviation and a half — efficient post-up volume is not penalized.

  usg_excess     = max(0, (usg_pct − (μ_usg + 1.5 × σ_usg)) / σ_usg)
  usg_multiplier = max(0.50, 1 − 0.30 × usg_excess)
                   anchored on 2015-16:  μ_usg = 20.05,  σ_usg = 4.99

• team success
  The index assumes the team was good. A graduated dampener scales the impact signal by team win percentage; great-stat-stuffer-on-bad-team seasons can't ride to the top of an index named for the man whose entire identity was elevating his team into champions.

  team_dampener = clamp(0.40 + 0.75 × team_win_pct, 0.65, 1.0)
                  ( 80% wins → 1.00,  50% → 0.78,  30% → 0.65 floor )
  
  impact ← impact × team_dampener

• playoff lift
  Per-season playoff Win Shares, additive bonus. The Duncan archetype always performs in the postseason; the formula credits showing up when it mattered.

  playoff_bonus = clamp(playoff_ws × 1.5, 0, 10)
                  ( 0 ws → 0,  3 ws → 4.5,  6 ws → 9, capped 10 )

• drama tax
  Small graduated penalty for ball-pounding signatures (isolation, foul-hunting, fadeaway-seeking). Not a hard gate — a smoothed dock that places Harden, Westbrook, and similar archetypes in the right tier.

  drama_tax = iso_pct                       × 22
            + max(0, pull_up_rate    − 0.15) × 12
            + max(0, fadeaway_rate   − 0.04) × 10
            + max(0, ft_drawing_rate − 0.25) × 15
              ( capped at 22 )

core = impact × consistency × mpg_scale × usg_multiplier

durability  = clamp((games_played − 65) / 3.4, −4, 5)
usage_bonus = (impact > 0.5)  ?  min(5, max(0, μ_usg − usg_pct) × 0.4)  :  0

raw_score = core × 35
          + durability + usage_bonus + playoff_bonus
          − drama_tax

duncan_index = clamp(raw_score, 0, 100)

duncan_career = Σ ( duncan_season × minutes_season )
              / Σ minutes_season