Verified facts
every number in the report is computed, not hallucinated, and traceable back to the source data. The non-negotiable floor of any sports content.
Sport
In productionNot a smarter model — one that finally carries your read of the game: which numbers matter, what they say. Applied to every game, identically.
The problem
Scores live in a stats app, performance in dashboards, audience in the CRM. And meaning — why that run swung the game, what to write and what to leave unsaid — lives in a handful of experts' heads, and never leaves them.
Put an AI on top and it inherits the fragmentation: it spits out a box score but can't interpret the way your newsroom would. Even fed the right file, it stays unstable. AI that disappoints in sport doesn't have a model problem — it has a foundation problem.
The sports substrate
Syrtis captures the raw material of sport as connected, sovereign knowledge: the raw data, the computations that turn it into facts, and above all expert interpretation — what your newsroom knows how to read in those numbers. This corpus is three things at once:
every number in the report is computed, not hallucinated, and traceable back to the source data. The non-negotiable floor of any sports content.
your experts' eye (what to analyze, how to write it, to what standard) is built into the interpreter agents. It stops living in one head; it applies to every game, identically.
connecting data, facts and expertise produces the extended analysis — runs, momentum, profiles, season trends — that only a specialist used to produce, and serves it at scale.
"A box score is data. The same box score, read with your expert's eye — every game — is knowledge."
In productionClient case · FIBA — Drazen
FIBA runs close to 3,000 games a year. Turning that raw data into reliable editorial content at that scale, with no dedicated writers: that's Drazen — the service Syrtis designed and operates for FIBA.
The real challenge isn't getting the facts right — it's faithfully translating FIBA's editorial expertise, on every game: which angle, which tone, how much depth. Co-built with FIBA's editorial team, Drazen carries the FIBA standard and reproduces it identically.
From raw data to a validated report. The match play-by-play file runs through an orchestration of components: deterministic code extracts the facts (runs, momentum, efficiency, plus/minus), interpreter agents fed with FIBA's expertise draw out the meaning, and a writer/verifier pair writes and checks. Output: a consolidated editorial report, generated with no human intervention.
In production on the FIBA Basketball World Cup 2027 European Qualifiers. Validated reports are published automatically on FIBA event sites, in a dedicated tab of each game page, between the Box Score and the Play-by-Play.
The sports expert
Sitting on the substrate, the agent reads the context it needs and writes fresh knowledge back — the Syrtis agent pattern, applied to sport.
Turns play-by-play into a complete editorial narrative, at a constant tone and standard.
Reads a player or a team across a season — strengths, trends, trajectory — the way a specialist would.
Pre-game briefing, post-game debrief: who was efficient, shooting trends, deviation from the season average.
Every claim is computed and tied back to the source data — no ungrounded takes, no hallucinations.
Your best analysts' eye, captured once, applies to every game and every team — identically.
The engine, embedded
With FIBA, Syrtis runs an own-brand service. With Baseline Basketball — an all-in-one club-management platform in production since 2024 — it does the opposite: it's the embedded AI engine inside their product, the basketball expert in the coach's pocket.
That's our underlying intent: abstract the AI away. The services that consume it shouldn't have to build it — but the engine behind it must be solid, controllable, auditable. Visible or invisible, Syrtis holds that foundation.
Baseline BasketballWhere it applies
The pattern is the same wherever an organization produces sports data and needs it interpreted to its standard —
automated game reports across the season; continuous, consistent editorial coverage.
interpreted player and team profiles, pre-game briefings, debriefs: a specialist's analysis, at a club's cost.
editorial production at volume, anchored in official data, never off-standard.
performance dashboards (runs, momentum, efficiency) read and translated into usable language.
connecting scattered stats, CRM and engagement to draw conclusions, not just charts.
compared profile synthesis over time, to decide on connected facts, not impressions.
Different sports, the same need — data that piles up, expertise that should apply everywhere but usually stays in a few heads.
Sovereignty & compliance
Game data, athlete profiles, audience: sensitive data, sometimes personal. Syrtis runs on-premises or in sovereign EU cloud — Drazen runs on a Syrtis server in the EU. Model-by-node selection keeps sensitive steps on local models. Full provenance, versioning, audit trail. GDPR and the EU AI Act respected by design.
Through ingestion nodes: APIs/webhooks, SQL databases, documents, exports. Everything enters via the indexer, with no upfront rebuild.
Yes — that's the core. Your newsroom's expertise is built into the interpreter agents and applied to every game, identically. That's exactly what a single prompt over a big file can't do.
Facts are computed by deterministic code and traced to the source; interpretation is framed and then checked by a dedicated agent. The report is validated, not guessed.
In a sovereign substrate: on-premises or EU cloud. It does not leave your perimeter.
A first service is built in hours, not months. Drazen today produces a complete report in minutes, with no human intervention.