Season workbench

Sabr-Jev

Season stats lie by omission. The verdict below says act, review, or set aside — with full probabilities, never hype.

Showing 25 of 50 — empty combinations are not offered.

Robbie Ray · 2021 · pitcher

FIP: 3.734
ERA: 2.840
K-BB%: 0.254

IP	193.333
HR/9	1.536
BB/9	2.421
K/9	11.545

No previous season in frozen catalog for evidence comparison.

Sample: 193.333 IP (minimum 50) · qualified

No data warnings.

Model read

regression risk provisional thresholds

The supplied metrics contain a meaningful regression-risk signal.

Confidence 46.0% · 28.0% signal · moderate

The metrics provide a meaningful but not decisive signal.

Signal: moderate (28.0%)
Read confidence: 46.0%

Confidence is mixed — the full probabilities below are the verdict, not a single call.

Verdict reflects the model's read of this season alone. Outcome in the Oracle pane is historical fact — they are independent and may not align.

Shows its work — audit trail

Model: jev-1.13.0
Recorded: 2026-09-21T18:57:27Z
State hash: sha256:f2470c2a63e7…
Questions hash: sha256:6635e0fc7d7a…
Card artifact: priv/data/cards.json

confidence_in_signal: escalate
fip_rise_ge_0_50_next: review
profile: act
season_read: review

Under review: full probabilities below, no single bold recommendation.

confidence_in_signal

weak	10.0%
moderate	52.0%
strong	38.0%

fip_rise_ge_0_50_next

P(true)	28.0%
Confidence max(p, 1-p)	72.0%

profile

strikeout	96.0%
command	2.0%
contact_management	1.0%
balanced	1.0%
other	0.0%

season_read

breakout	14.0%
stable	16.0%
regression_risk	57.0%
small_sample	12.0%
other	1.0%

Oracle — retrospective eval-only pane, never sent to Jev

Metric	2021	2022	Change (actual outcome)
FIP	3.73	4.10	▲ rose +0.37 (target: rise ≥ 0.50)

Target: fip_rise_ge_0_50_next · Label: false · Season: 2022

Outcome ≠ recommendation. Change flag shows what historically happened; the read above shows what the model recommended at the time. No correlation is implied.

At the time, the model projected 28.0% chance of noul change. Actual outcome is in the table above.

Evaluation status

Prospective evaluation: accuracy pending. No accuracy claim is made.

2024 → 2025 joins are retrospective plumbing only. The local ledger is hash chained but cannot prove external capture time.