News · Roadmap
Bench is coming in the next 2–3 months
Motivation and problem formulation
Performance and coaching desks already consume video, GPS/load proxies, opposition reports, and medical notes — usually across tools that do not share an evaluation contract. We hypothesize Bench as a product-shaped answer to coach-facing sports intelligence with measurable workflow outcomes: reduce time-to-brief, improve traceability of recommendations to evidence, and keep humans in control of selection and load decisions.
Because Bench is Coming soon, this article publishes the proposed architecture and evaluation design we will run at launch — with pending markers until pilot results land.
Architecture schematic (qualitative) — roadmap.
Proposed system architecture
We expect Bench to sit alongside StatGost (stats reasoning), Koba (agentic research), and Sponsorlytix/Dantani (commercial/media intelligence) without duplicating those SKUs. Exact feature bullets are reserved until launch announcement. Safe statements today: Bench is an intentional suite member on a 2–3 month horizon; it is aimed at professional sports workflows rather than consumer fantasy products; evaluation will prioritize coach-desk utility and safety over vanity accuracy.
Illustrative evaluation profile — not a published benchmark.
Data and methods (proposed)
At launch we intend to evaluate Bench with: (1) time-to-first-useful-brief studies on held session packets; (2) evidence-link audits (recommendations must point to clips/tables); (3) safety probes (medical speculation, minors, discriminatory language); (4) preference tests vs prior desk tooling. Training/data specifics will be disclosed only to the level launch docs support — this roadmap post does not invent corpora sizes.
Illustrative evaluation profile — not a published benchmark.
Comparison vs suite siblings (planned positioning)
| Dimension | Bench (roadmap) | StatGost | Koba |
|---|---|---|---|
| Status | Coming in ~2–3 months | Announced flagship stats | Announced mid-tier agent |
| Primary desk | Coach / performance (hypothesized) | Scout / analyst / commercial stats | General sports professional |
| Public feature list | Not published | Five stats pillars | Five agent pillars |
| Eval posture in this post | Proposed protocols only | Pending internal results | Pending internal results |
Proposed evaluation table
| Metric | Definition | Unit | Protocol | Result |
|---|---|---|---|---|
| Time-to-brief | Minutes from session packet to usable brief | min | Lab + pilot desks | — |
| Evidence link rate | Share of recommendations with citations | % | Audit sample | — |
| Safety probe pass | Pass rate on medical/minors/bias probes | % | Red-team suite | — |
| Coach preference | Blind preference vs baseline tooling | win rate | Pilot study | Internal eval — contact sales |
| Schema stability | Breaking changes per release | count | Changelog review | — |
Limitations and failure modes (anticipated)
- Over-automation of selection decisions — Bench must not silently replace coaching judgment.
- Video rights and privacy constraints may limit what pilots can ingest.
- Feature inventory unknown publicly — treat social speculation as non-authoritative.
- Timeline risk — 2–3 months is a planning window, not a contractual SLA.
Safety and responsible use
Coach-facing systems touch athlete welfare. Proposed policy: no autonomous medical diagnosis; heightened friction around youth athletes; clear labeling of model suggestions vs staff decisions; audit logs for high-impact recommendations.
What’s next / availability
Bench is on the near-term roadmap. Watch Eagle Eye News for the launch post with concrete capabilities and the first evaluation disclosures we are willing to publish. Talk to sales for pilot interest.
Appendix — glossary
Time-to-brief. Operational metric from raw session inputs to a coach-usable summary.
Evidence link. Pointer from a recommendation to clip, table, or note.
Roadmap product. Named suite member without a full public feature card yet.