Skip to main content
Blaze Sports Intel
ScoresCollege BaseballIntelAgentPricing
College Baseball scores updated live every 30 seconds
HomeScoresBaseballAnalytics
ExplorerVisualsPark FactorsConferencesCompare TeamsCompare Conf.MethodologyGlossary
Home/College Baseball/Savant/Methodology

BSI SAVANT

Methodology

How BSI computes park-adjusted sabermetrics for 300+ D1 college baseball programs. Every number on the Savant Explorer traces back to a specific methodology documented here.

Data Pipeline

ESPN Box Scores → Highlightly Pro (330 teams) → BSI Ingest Worker → D1 Raw Stats → BSI Savant Compute (6h cycle) → Advanced Metrics in D1 → KV Cache → Savant API → You

Every 6 hours, the Savant Compute engine pulls raw box score data, applies Tango-methodology linear weights, regresses park factors, computes conference-adjusted metrics, and writes the results to long-term storage. The API serves cached reads with a 5-minute refresh window.

wOBA — Weighted On-Base Average

wOBA assigns run values to each offensive outcome based on how much each event contributes to scoring. A home run is worth more than a single, and wOBA captures that gap precisely.

wOBA = (0.69×BB + 0.72×HBP + 0.89×1B + 1.27×2B + 1.62×3B + 2.10×HR) / (AB + BB + SF + HBP)

BSI adaptation for college baseball: The weights above are derived from Tom Tango's linear weight methodology as applied to MLB run environments. College baseball has a higher run environment (~5.5 runs/game vs. ~4.5 in MLB), so BSI recalibrates the wOBA scale factor annually to normalize to the D1 OBP baseline.

Stabilization: ~150 plate appearances. Below this threshold, BSI displays the value with a sample-size indicator. The metric is most reliable after conference play begins (Week 6+).

wRC+ — Weighted Runs Created Plus

wRC+ answers the question: “How productive is this hitter compared to the league average, adjusted for park and conference?” 100 is average. 120 means 20% better than average. 80 means 20% worse.

Why it matters for college baseball: A .350 hitter at a bandbox in the Big South is not the same as a .320 hitter at Lindsey Nelson Stadium in the SEC. wRC+ strips away the park and conference context to give you a true comparison. It is the single best cross-conference batting metric.

Stabilization: ~200 plate appearances. Most reliable in the second half of the season.

FIP — Fielding Independent Pitching

ERA tells you what happened. FIP tells you what the pitcher actually controlled. By isolating strikeouts, walks, hit-by-pitches, and home runs — the outcomes that don't depend on fielders — FIP is a better predictor of future ERA than ERA itself.

FIP = ((13×HR + 3×(BB+HBP) - 2×K) / IP) + FIP constant

The FIP constant: BSI computes a D1-specific FIP constant each season by setting league FIP equal to league ERA. This centers the scale so that a 4.00 FIP means “average for college baseball,” not for MLB.

Stabilization: ~60 innings pitched. Below this, the K/BB ratio is a more reliable indicator of true talent.

ERA- — Adjusted ERA

ERA- adjusts a pitcher's ERA for park and conference context, then scales to 100. Below 100 is better than average. A pitcher with 80 ERA- allowed 20% fewer runs than the conference-adjusted average. This is the best cross-conference pitching comparison metric — a 3.50 ERA in the SEC is not the same as a 3.50 ERA in the SWAC, and ERA- accounts for that difference.

Park Factors

Every stadium affects offense differently. BSI computes park factors by comparing run-scoring at each venue to the league baseline, then regressing toward 1.0 to account for small sample sizes.

Regression toward 1.0: A park that shows a 1.30 runs factor over 15 games is regressed toward 1.0 more aggressively than a park with 40 games of data. This prevents early-season noise from distorting player metrics. BSI uses a sample-size weighting function: the more games, the more the raw factor is trusted.

Factors computed: Runs, hits, home runs, walks, strikeouts — each with independent regression rates. Currently tracking 195 D1 venues.

Conference Strength Index

The Conference Strength Index combines inter-conference win percentage, average run environment, aggregate wOBA, aggregate ERA, and RPI average into a single composite score. This powers the conference adjustments in wRC+ and ERA-.

Why it matters: A .320 wOBA in the SEC represents fundamentally different offensive production than a .340 wOBA in the Big South. The conference strength index quantifies that difference so that cross-conference player comparisons are meaningful, not misleading.

Data Sources

PRIMARY

Highlightly Pro — 330 D1 teams, live game data, venue metadata, win predictions. Updated every 30-60 seconds during live games.

SECONDARY

ESPN — Box scores, standings, rankings, schedules. Fallback data source when Highlightly is unavailable.

COMPUTE

BSI Savant Compute — Proprietary engine running on Cloudflare. Ingests raw box scores, applies Tango linear weights, computes park-adjusted metrics, and writes to D1 every 6 hours.

AI-Accessible via MCP

Every metric on this platform is available programmatically through the BSI College Baseball Sabermetrics MCP server — the only AI-accessible college baseball analytics interface in existence.

AI assistants like Claude can query live scores, standings, rankings, team sabermetrics, player stats, conference strength indices, and national leaderboards directly. Nine tools, 330 D1 programs, updated every 30-60 seconds during live games.

sabermetrics.blazesportsintel.com

Known Limitations

  • No Statcast data. College baseball does not have exit velocity, launch angle, or spin rate tracking at scale. BSI metrics are computed from traditional box score data, not TrackMan/Hawk-Eye data.
  • BABIP instability. BABIP requires ~400 plate appearances to stabilize. Most college hitters never reach this threshold in a single season (~200-250 PA). BSI displays BABIP but flags it as high-variance.
  • Park factor regression. Early-season park factors are heavily regressed toward 1.0 due to small sample sizes. Mid-season factors (30+ games) are more reliable.
  • 6-hour refresh cycle. Advanced metrics update every 6 hours. Live box scores update in real-time, but derived metrics (wOBA, wRC+, FIP) reflect the last compute cycle.

Questions about the methodology? Reach out at austin@blazesportsintel.com