MLB Model Validation — June 2026 Research Summary

MLB Model Validation — June 2026 Research Summary

> Signal Syndicate Research Library · Model Validation · Educational intelligence · Not betting advice

This memo consolidates June 2026 MLB quant evidence for the Performance page and Research Library. Figures are historical estimates from graded research samples — not live P&L.


What the public board shows (ROUTE_B)

The Signal Board (Platform Sprint 4) displays v1 broad totals under ROUTE_B:

Rule Behavior
Model Team/Pitcher v1 broad totals scan
Edge cap Plays with \ model_edge\ > **2.5 runs** excluded from public payload
Confidence Edge **magnitude**, not win probability
P43 shadow **Not** on the board

June 2026 research sample

Metric Value
Capped plays (Route B) **105**
Win rate **50.48%**
ROI **−3.64%**
Archive context 19 June slates · 177 uncapped graded plays in archive

The cap exists because uncapped strong-edge rows showed negative ROI in shadow analysis — removing the tail is a trust fix, not a performance claim.


Shadow lane: P43 summer UNDER

Metric Value
Status **MONITOR ONLY**
OOS bets **18** (minimum for promotion review: **75**)
ROI **−4.55%**
On public board **No**

After as-of backfill expanded OOS density, ROI weakened — a textbook small-sample trap cleared by more data, not strengthened.


Shadow lane: Hits props V1 / V1.3

Forward OOS testing (730 Odds credits, 10 slates × 2025 + 2026) completed in June 2026:

Year V1 ROI @ edge 0.3 n (edge-filtered) Verdict
2024 pilot **+5.22%** 1,504 Baseline only
**2025 OOS** **−22.79%** 1,073 Sign flip — do not promote
**2026 OOS** **+9.87%** 1,284 Monitor — early-June slates only

V1.3 (85/15 Phase D blend) tied V1 on 2025/2026 because Phase D batter cache was not refreshed for new seasons. No Signal Board wiring.

Total hits lane credits: 1,120 (390 pilot + 730 OOS).


K props — RETIRED

The pitcher strikeouts props lane is RETIRED. Do not spend Odds credits on K. See: Why We Retired K Props (May–Jul).


CLV — what we track and why

Closing Line Value (CLV) measures whether a bet beat the market close — a process metric, not win rate.

Live CLV series for shadow lanes is not yet populated (insufficient closing-line pairs). We ship methodology + education until 30+ graded plays with closing lines exist.

Read: CLV Validation Study Memo.


Promotion gates (unchanged)

1. 75 OOS shadow bets minimum before promotion review
2. Walk-forward / sign-flip discipline
3. Multi-AI courtroom + founder sign-off
4. No `config.py` production changes without explicit YES


What this means for readers

Research only · estimates only · not betting advice.