Draft Value Curve
Control-years WAR by draft slot from the 2005–2016 drafts, priced against today’s slots.
What it predicts
WAR in a drafted player’s first six MLB seasons.
Inputs
- MLB Stats API draft records 2005–2016
- FanGraphs WAR 2005–2026
- 2026 slot values
Method
- Gaussian kernel smoothing on log(pick)
- High school vs. college from age on July 1 of the draft year
- Present value at Contract Studio’s $/WAR, growth and discount, starting at the observed debut lag
- Net of control-year pay: the minimum salary for three pre-arbitration seasons, then Contract Studio’s arbitration model (base by year plus pay per recent and career WAR), each weighted by the chance he reaches the majors
How well it works
Computed by the pipeline from the data on this site
- Sample
- 3,840 picks, drafts 2005–2016; draft-to-debut lag smoothed by pick, 2.8 years at #1 to 4.6 at #300
- Control WAR by round
- Picks 1–10 7.89 WAR · 88% reach MLBPicks 11–30 3.30 WAR · 74% reach MLBPicks 31–75 1.78 WAR · 56% reach MLBPicks 76–105 1.20 WAR · 46% reach MLBPicks 106–140 0.85 WAR · 40% reach MLBPicks 141–200 0.56 WAR · 35% reach MLBPicks 201–320 0.48 WAR · 20% reach MLB
- Backtest: curve fit on 2005–10 classes, scored on 2011–16 classes (control WAR per pick, predicted vs. actual; ± 2 standard errors, picks)
- Picks 1–5 10.2 vs. 7.1 (± 3.1, 30)Picks 6–10 7.7 vs. 6.5 (± 2.9, 30)Picks 11–20 3.8 vs. 4.2 (± 1.5, 60)Picks 21–35 2.5 vs. 2.7 (± 1.1, 89)Picks 36–75 1.8 vs. 1.4 (± 0.5, 240)Picks 76–140 1.0 vs. 0.9 (± 0.3, 390)Picks 141–200 0.6 vs. 0.7 (± 0.3, 360)Picks 201–320 0.6 vs. 0.3 (± 0.1, 720) Overall the later classes produced 85% of the predicted WAR. Six control seasons for 2011–16 classes include the 60-game 2020 season, and a few 2016 careers are still short, so part of any gap is not era.
Data: 3,840 picks from the 2005–2016 drafts (MLB Stats API) and FanGraphs WAR through 2026.
What it can't see
- Values slots, not individual amateurs: amateur tracking and scouting data aren’t public.
- Drafting and development have changed since 2005–2016; the bonus-pool era (2012–16) gets its own curve.
- A pick whose player didn’t sign and was drafted again later counts as zero at the earlier slot.
- Six control seasons for the later classes include the 60-game 2020 season, so the backtest runs a little low for era reasons, not only the curve’s.
- International signings aren’t covered (bonus data not public).
How we'd measure impact
- Accuracy of the curve for 2017–2020 classes as they mature (backtest on the card: fit on 2005–10, scored on 2011–16)