Michael Kelly
5 IP · sample confidence ×0.2 · full player page
1 · What changed (2026 vs 2025)
Moving against him
ERA+13.020
wOBA against+.255
xERA+4.250
xwOBA against+.098
In his favor / neutral
Fastball spin−88 rpm
2 · Why — ranked hypotheses
1. ERA overstates the problem — results behind the process
12%Causal path: Expected outcomes → Actual outcomes
- • ERA +7.120 vs xERA(ERA at least half a run above xERA)
- • Fastball velocity +0.1 mph vs last season
Against / absent: d_whiff_pct (no data) · d_hard_hit_against (no data)
2. Spin down more than velocity — grip/feel or ball-in-hand change
8%Causal path: Pitch stuff (velo/spin/shape) → Contact rate → Expected outcomes
- • Fastball spin −88 rpm vs last season
- • Fastball velocity +0.1 mph vs last season(velocity held while spin fell)
Against / absent: d_ivb_fb (no data)
Unexplained / unobserved causes: 80% — confidence is deliberately capped; public data cannot see every layer.
3 · What the coach does tomorrow
No swing/pitch changes — hold the current process
The process metrics are intact and expected stats say results will regress upward. The costliest move is fixing what is not broken.
Explicitly protect the player from over-adjustment
Communicate that the underlying numbers are strong; luck deficits invite harmful tinkering.
Actions follow the top hypothesis; if video review rejects it, fall to the next.
Data gaps
- • Biomechanics (Hawkeye), TrackMan, and Blast Motion are team-internal — the mechanics layer is inferred from its downstream fingerprint, never observed.