News · Announcements
Introducing StatGost
Motivation and problem formulation
Sports organisations sit on dense statistical history. The failure mode is not “no data” — it is slow, brittle translation from tables into professional judgment under window pressure. We formulate StatGost as flagship stats-first reasoning: given structured sports statistics plus language-model reasoning, produce answers that remain grounded in the numbers that matter for scouting desks, performance staff, commercial teams, and analysts.
Architecture schematic (qualitative).
System architecture
Announced capability pillars: Player Stats; League Stats; Finance Stats; Sponsorship Stats; Live Scores. Positioning emphasizes stronger reasoning, more capable scouts, and sharper professional work versus prior tiers. StatGost is the flagship complement to mid-tier agentic Koba: when the work is numbers-first and professional-grade, StatGost owns that lane.
At announcement there was no live product UI or docs yet, and “scouts” remains a capability pillar without a published scout-report schema. We therefore describe evaluation design rather than fabricated scout accuracy.
Illustrative evaluation profile — not a published benchmark.
Data and methods
StatGost is designed to reason over structured sports statistics with LLM synthesis. Live scores provide situational awareness when games are in flight. Sponsorship stats connect on-pitch and commercial contexts — complementary to Sponsorlytix’s CV measurement rather than a replacement for logo detection.
For market context (not StatGost model scores), published Sponsorlytix case materials show why finance/sponsorship stats matter: Chelsea 2025/26 combined sponsor brand value of $751M — a Sponsorlytix media-value estimate, not contract value or cash return (the same distinction we apply to AVE-style proxies); FIFA World Cup 2026 case scoping 104 matches with multi-billion revenue context charts. Those figures are case-study measurement context, not StatGost benchmarks.
Illustrative evaluation profile — not a published benchmark.
Illustrative evaluation profile — not a published benchmark.
Comparison vs prior tier / siblings
| Dimension | StatGost (flagship) | Koba (mid-tier agentic) | Sponsorlytix |
|---|---|---|---|
| Core job | Stats-grounded professional reasoning | Search / research / advisory agent | CV sponsorship measurement |
| Live layer | Live scores | Live dashboard threads | Live sponsorship dashboards |
| Scout language | Yes (marketing; schema unpublished) | Not primary | Not primary |
| Finance / sponsor stats | First-class pillars | Via tools / advisory | Brand-value measurement |
Evaluation table
| Metric | Definition | Unit | Protocol | Result |
|---|---|---|---|---|
| Numeric faithfulness | Figures match source tables | % | Table-grounded QA set | Internal eval — contact sales |
| Comparison validity | Like-for-like bases in player/league comps | % | Rubric on held tasks | Internal eval — contact sales |
| Caveat coverage | Required uncertainty notes present | % | Checklist on scout-like briefs | — |
| Live score sync lag | Delay vs official feed | s | Operational sampling | — |
| Professional usefulness | Blind preference vs prior tier | win rate | Desk rater study | Internal eval — contact sales |
Limitations and failure modes
- Schema drift across leagues and seasons breaks naive comparisons.
- Small-sample overfit in early-season form narratives.
- Unpublished scout-report schema — do not treat “scouts” as a defined export format yet.
- Parked domain / no public UI at announcement — access via Eagle Eye sales / suite paths.
- Sponsorship stats ≠ CV detection — StatGost does not replace Sponsorlytix measurement.
Safety and responsible use
Stats briefs can influence careers and markets. We encourage explicit sample-size caveats, separation of measured history from projection, and care with medical or disciplinary inferences. Live scores should not be used as sole trading signals without human controls.
What’s next / availability
StatGost is announced as the flagship stats tier. Visit positioning via statgost.com as surfaces mature, and the Eagle Eye products section for suite context.
Appendix — glossary
Numeric faithfulness. Generated figures that match underlying tables.
Scout-style brief. Professional summary for recruitment/performance desks — schema not publicly standardized here.
Live scores. In-game situational layer alongside historical stats.