Health Warning Signs
In-season warning signs, each tested by role against every injured-list placement since 2024 before it’s trusted.
What it predicts
Injured-list placement (arm injuries for pitchers; any for hitters) within 30 days after the sign appears.
Inputs
- MLB transactions 2024–26 (placements, transfers, activations; MLB’s injury text mapped to body part)
- Primary-fastball velocity, arm angle and spin by appearance; hitter bat speed by game
- MLB and minor-league innings and pitches by season (minor leagues from 2025)
Method
- Signs use only appearances before the date, and the primary fastball as of that date
- Relative risk vs. appearances without the sign, and the distinct injuries behind each sign. 95% intervals: with 10+ injuries, a bootstrap that resamples pitchers (one pitcher’s outings aren’t independent); with fewer, an exact interval that counts each injury once (a bootstrap of a handful of injuries reads resamples with none as zero risk)
- Verdicts come from those numbers: “Strong” or “Real” needs 10+ injuries and an interval above 1.0; fewer injuries are “Suggestive” at most, and only at a relative risk of 1.5+
- Split by role: a starter’s typical outing so far had 25+ primary fastballs. The live warning goes only to the role whose test holds (starters); a reliever’s drop is a watch item
- The backtest uses the live rule: 8+ earlier outings, and no outings in the 45 days after a return from the IL. Pitchers back that recently are listed apart, with their raw change
- What it catches: arm placements the check could see and flagged, alarms per club-season, lead time by role; a sensitivity without each pitcher’s last outing before the IL
- Recurrence: how often an arm placement in 2024–25 was followed by another within a year, by role (pitcher as MLB lists him at the placement, or 100+ MLB pitches; hitters apart) and body part, a part with fewer than 20 placements falling back to the role’s all-arm rate (a league base rate for the plan’s medical ledger, not one player’s risk)
- Season-end test: a drop over a pitcher’s last three outings of 2024 or 2025 against an arm placement after the regular season ends, through May 31 (September placements left out). Two winter injuries: too few to tell
- IL stints: a 60-day transfer or re-filing (even written as a new placement) is one stint; days missed are regular-season days in the placement’s season
How well it works
Computed by the pipeline from the data on this site
- Fastball −1.0 mph
- 1.80× (1.23–2.51) 69 of 781 flagged vs. 1241 of 25,277
- Fastball −1.5 mph
- 2.36× (1.07–4.55) 18 of 153 flagged vs. 1292 of 25,905
- Arm angle ±3°
- 1.32× (0.93–1.82) 154 of 2,384 flagged vs. 1156 of 23,674
- Spin −100 rpm
- 1.32× (0.60–2.30) 21 of 318 flagged vs. 1289 of 25,740
- Bat speed −1.5 mph (hitters)
- 0.98× (0.89–1.08) 1761 of 19,165 flagged vs. 8277 of 88,289
- Season-end drop → arm IL by May 31 next season (pooled 2024–25 season ends): too few injuries to grade
- 0.34× (0.04–1.26) 2 of 37 vs. 104 of 655
- Fastball −1.0 mph, season by season (all / starters / relievers)
- 2024 1.05× · 0.89× · 1.09× (11 injuries)2025 2.61× · 5.72× · 1.78× (14 injuries)2026 2.02× · 6.03× · 1.38× (14 injuries)
- Forward: the rule set on 2024–25, scored on 2026 only
- 2.02× (0.95–3.48) · starters 6.03× · relievers 1.38× the same rule (−1.0 mph over three outings, 8+ earlier outings: round thresholds, not fit) scored on 2026 outings only. The starter/reliever split was found looking at 2024–26 together, so by role this is a check, not a clean forward test
Intervals are clustered by pitcher where published (one pitcher’s outings aren’t independent). The pooled rows use 2024–26, the same seasons the warnings are used on: an in-sample check; the forward row is the test.
Data: MLB transactions 2024–26 (injured-list placements); fastball velocity, spin and arm angle by appearance and bat speed by game, Statcast 2024–26; minor-league workloads from 2025.
What it can't see
- No medical, imaging or pain data.
- The thresholds are round numbers, but the starter/reliever split was found on the same 2024–26 data it is scored on; 2026 alone is a check, not a clean forward test.
- Pitchers with fewer than 8 earlier outings that season (early in the season or just arrived) and those within 45 days of an IL return aren’t checked; the page counts the placements each reason misses.
- Workload jumps can’t be tested yet: minor-league innings start in 2025 in our data. A season lost to the IL is shown as that, not as a percentage jump.
- Most flagged pitchers don’t get hurt; this is triage, not diagnosis.
How we'd measure impact
- Arm IL days per 1,000 innings before vs. after the check protocol
- False alarms per club-season the training staff will tolerate (about one a season on starters today)