BSI Coach Intelligence — Methodology
How the Coach Intelligence Score is computed.
Transparent weights, peer-reviewed citations, and the limits BSI publishes alongside the rank.
The Formula
CIS is a multi-criteria decision analysis composite, not a regression-validated prediction. No peer-reviewed paper has isolated a clean single-coach causal effect at the MLB level — the famous "Mazzone Effect" lived in popular press, not in a peer-reviewed journal, and failed to replicate in Baltimore.
CIS = Development Outcomes (w1)
+ Methodology / Communication (w4)
+ Reputation (w3)
+ Longevity (w2)
+ Playing Résumé Modifier (w5)
+ Level Difficulty (w6)
− Safety / Overtraining Risk (w7)
− Uncertainty PenaltyTier-varying Weights
Methodology and communication carry the most weight at youth (40% pitching, 42% hitting) and the least at pro (20% pitching, 22% hitting). Outcomes invert the curve.
| Pitching Weight | Youth | College | Pro |
|---|---|---|---|
| w1 Development outcomes | 0.25 | 0.35 | 0.45 |
| w2 Longevity at level | 0.10 | 0.10 | 0.10 |
| w3 Peer reputation | 0.05 | 0.10 | 0.15 |
| w4 Methodology & communication | 0.40 | 0.30 | 0.20 |
| w5 Playing-résumé modifier | 0.05 | 0.05 | 0.05 |
| w6 Level difficulty | 0.15 | 0.10 | 0.05 |
| w7 Injury-risk penalty (cap) | −0.20 | −0.10 | −0.10 |
Standings vs Power Rankings
Standings = CurrentScore (clean monthly rank) PowerRank = 0.75 · CurrentScore + 0.25 · Trend90
A coach can be #2 in Standings and #1 in PowerRank when their last 90 days of verified athlete outcomes are trending sharply upward. The UI shows both.
Hierarchical Bayesian Shrinkage + 95% Bootstrap CI
Per-pitcher outcome variance is large. BSI applies hierarchical Bayesian shrinkage to all outcome metrics — small-sample deltas pull toward the population mean. Final scores are bootstrapped (n = 5,000 resamples) to produce a 95% credibility interval. Coaches with overlapping CIs are displayed as ties, not ranked off each other.
Sensor Noise Floor
Below these thresholds, the model does not credit a coach with a delta:
- ~1.5 mph velocity (pitching) or bat speed (hitting)
- ~1° release-point or attack-angle change
- ~3 percentage points strike rate
Source: Kusafuka et al. 2022 (Sensors, PMC8879135). Blast-class sensors have ICC ~0.78 for speed, ~0.58 for angle. Below floor → marked unverified, not credited.
Honest Gaps
- Causal attribution is permanently underdetermined. No peer-reviewed regression has isolated a clean MLB pitching-coach effect. CIs displayed prominently.
- Spanish-language Caribbean and Latin American pipelines are largely invisible. Estimated 30–50% of South Florida elite youth instruction is invisible to current ingestion. Phase 3 priority.
- NPB / KBO / Taiwanese pitching development is off-radar. Phase 4 International bracket addresses.
- Travel-ball middle-layer dilution. Rate the individual coach, not the franchise.
- Injury data on private youth coaches is essentially absent. w7 invokable only for coaches with identifiable high-risk methodology. Unknown rates → w7 = 0, visibly flagged.
- The Mazzone Effect is partially fictional. Bradbury's 0.64 ERA staff effect was popular-press, not peer-reviewed; failed to replicate in Baltimore.
Legal Posture
BSI ranking output is first-party editorial speech with no Section 230 protection. Defenses include opinion doctrine (Milkovich v. Lorain Journal), this transparent disclosed methodology, factual accuracy of inputs via the verification stack, and a 5-business-day correction-and-takedown SLA. Youth/HS coaches treated as private figures (Minnesota v. McGuire 2019) — verification is non-negotiable. SafeSport Centralized Disciplinary Database cross-check is mandatory before publication of any youth-tier coach. No exceptions.