Back to blog|General

The Veezoo Analytics Cup 2026: Meet the Finalists and the Winner

We handed the internet 90 years of World Cup data and a single brief. Meet the three finalists, see how they built broadcast-ready tools in days, and find out who won.

June 29, 2026
11 min read
The Veezoo Analytics Cup 2026: Meet the Finalists and the Winner

A few weeks ago we handed the internet a football problem. With the 2026 FIFA World Cup on the horizon, we loaded 90 years of World Cup data, players, matches, referees, bookings and venues, into Veezoo, turned it into a knowledge graph you can question in plain language, and set a single brief: build the tools a broadcast team would actually use to tell stories about the matches happening on any given day. Trivia, trends, head-to-head history, match briefings, the things a commentator needs at their fingertips minutes before kick-off.

The entries came in from all over the world, and they were good. So good that our judging panel had a genuinely hard time. Every submission was scored against four criteria.

35%
Insight Depth
Non-obvious, actionable findings. Good questions count as much as good answers.
25%
Agentic Usage
Creative use of conversational AI, multi-step reasoning and agents.
20%
Storytelling & Design
Clarity, visual quality and the narrative arc of the Canvases.
20%
Creativity
Unexpected angles, risk-taking, surprise factor.

The panel brought three perspectives. Andy Cotgreave hosted the competition and served as head judge, drawing on a long career in data visualization. JP Ribeiro, Veezoo's co-founder and CTO, was our technical evaluator. And we were joined by guest judge Stefan Kühn, who does sports analytics for professional football club FC St. Pauli and is CTO of a sports-technology startup. Between them, the panel covered design, engineering and the sport itself, a strong lineup for a World Cup brief.

One thing is worth saying plainly before we get to the entries. None of the finalists had used Veezoo before this competition. They signed up, explored the data, and built everything you are about to see in a matter of days, with no onboarding and no help from our team. As JP put it during the judging, that is exactly what agentic analytics is meant to do: lower the barrier so anyone can turn a dataset into something useful.

A quick caveat before we start. Each of these entries is far richer than a blog post can hold, so what follows highlights only a few standouts from each, and the screen recordings capture just part of what they built. For the full picture, the complete walkthrough interviews are linked at the end.

So here are the three finalists, starting with third place.

🥉  Third place

Stelios Chiotis

London-based product manager, BI and data-viz background, founder of his own amateur football club

Stelios built a single, tightly designed Canvas, the Match Briefing Canvas, aimed squarely at commentators and reporters: a screen that hands them on-air ready lines they can say out loud, with the option to drill deeper whenever they want more. His guiding principle was restraint. Everything fits on one screen. Minimal scrolling, low cognitive load, tabbed navigation, so a broadcaster is never hunting for information mid-broadcast.

  • Match Digest: The home tab, a one-screen match briefing: an on-air line ready to read out, the last meeting, why the game matters, a world benchmark, and quick hits like the main narrative, the player to watch and the stat of the match.
  • Lineups: Stelios got Veezoo to draw an actual football pitch with the squads laid out in formation, colour-coded by recent form. There is no built-in "4-3-3 chart" in any analytics tool. He simply described what he wanted and iterated on the design until Veezoo placed every player in the right position.
  • Head-to-head: The two teams’ full match history, with scores, venue and date, expanding to reveal the scorers only when you ask for them.
  • Team Stats: Familiar side-by-side bars in each team's colours, the same visual language you see on broadcast TV.
  • Key Players: Side-by-side player comparisons, Messi against Fontaine for example.
  • Top Facts: Four pages of broadcast-ready trivia.
  • Story Angles: Themed talking points, each with a ready-to-say line plus the supporting data behind it, filterable by pre-match, half-time or full-time.
Stelios — a few highlights from his single-screen commentator Canvas (not the full entry)
What the judges loved
JP called it exceptional work, singling out the hand-built pitch lineup. As he put it, “we organized this with World Cup data but didn't add any football-specific component. He created that lineup completely by himself.” Stefan loved how compact and readable it was, “no fluffy words, every bit of relevant information,” and was struck by the verify-this-claim feature, where a world-knowledge statement gets checked back against the actual data. Andy summed it up: “Love the Player Line Up view. Hits the brief perfectly with novel approaches. The drill-down in Story Angles is so sweetly done.

“It’s like I’ve got another work colleague with me, like I have a designer with me, able to go through that iterative process really quickly.”

Stelios Chiotis, on building with Veezoo for the first time
🥈  Second place

Dzmitryi “Dima” Yarmolin

BI consultant based in Frankfurt

Dima gave his entry a personality. The Yarmolin Wire is a fictional sports-media outlet, designed to let commentators and writers generate stories on demand for any matchup in the tournament. The standout was his use of nested agents, a set of specialist sub-agents that each do a job a real newsroom would recognize, split between agents you call on demand and desks that publish on a schedule.

The on-demand agents build a canvas the moment you ask:

  • The Pre-Match Pack: Pick two teams and get a seven-part briefing: a headline number, head-to-head, the last three meetings, key players, managerial history, the narrative thread, and three follow-up questions. Everything a commentator needs before kickoff, pulled in one pass.
  • The Storyline Generator: Give it a theme and it returns a three-act pitch, the setup, the data behind it, and the angle, with broadcast-ready lines and three follow-ups. It turns a vague idea into pitchable story angles, each traceable to the data.
  • The Manager Intelligence Desk: Pick any manager and get a full profile: a peer archetype, win rate measured against their own era, full tournament record, and where to look next. It places a manager among their peers rather than reducing them to a raw win total.

The scheduled desks are built to run automatically, the Wire each morning and the Anomaly Desk weekly. Scheduling was switched off for the submission, so they can also be run on demand:

  • The Touchline Daily Wire: A daily desk note in four parts: Born Today, On This Day, a Tournament-Day Echo, and a Weird Stat of the Day. A morning brief of anniversaries and on-air facts, ready before the first meeting.
  • The Anomaly Desk: Surfaces oddities in the record across teams, eras, managers, officials, venues and player longevity, each with its numbers and a query trace. A weekly lead generator that finds the strange results a writer would never think to query.

The Wire carries two standing pages of its own as well. A refereeing analysis compares referees across confederations and leans into the fairness questions that always swirl around a World Cup, the part Stefan singled out. And a Stadiums page maps the venues from past World Cups staged in Mexico and the United States, a nod to the 2026 hosts.

The look and feel was deliberate, inspired by The Athletic and The Economist, and Dima balanced descriptive, article-like agents with sharper analytical ones. His process is a nice tell of how this kind of work is changing. It starts with pen and paper, moves through Claude to structure the agent instructions, then iterates inside Veezoo.

Dima — a few highlights from The Yarmolin Wire (not the full entry)
What the judges loved
Stefan was taken with the refereeing visualizations, “these really help explore things like fairness, very compact design, very readable,” and called the whole entry “a really outstanding example.” JP highlighted how far Dima pushed the charts and visualizations compared to the other entries, along with the clever selection of agents and desks. Andy praised the “brilliant mix of data storytelling, a strong design ethic, and excellent nested agents.
🥇  Winner

Oliver Yang

Data-science master’s student at Columbia, and a die-hard football fan

Our 2026 Veezoo Analytics Cup champion is Oliver Yang, also known as Chung-Yeh Yang. His entry, The Storyline Engine, did something none of the others tried. Oliver built six distinct agents, each a “fan” you would find in a newsroom, every one interrogating the same data with its own editorial voice.

  • The Cynic: The hard, factual read.
  • The Historian: The long memory of past meetings.
  • The Tactician: The formations and the football.
  • The Romantic: Following a single player’s arc, Messi or Ronaldo.
  • The Ultra: The obsessive, knows-every-stat team superfan.
  • The Oracle: The one who surfaces the hidden gem nobody else spotted.

Those voices feed three purpose-built canvases for different moments of the broadcast day. A Morning Pitch for reporters and morning bulletins gives an opening read on the matchup, a headline from each persona, and an editorial-priority steer on which angle matters most. A deeper Match Dossier for live broadcasters is the kind of canvas that tells you that if Argentina score first, England's historical win rate drops from 80% to 56%, exactly the line you would reach for on air. And a Tournament Pulse, built for mid-tournament, steps back from any single game to ask what story this World Cup is telling. On top of that sat a Canvas Builder agent that generates the right UI on the fly, taking the user by the hand: which canvas do you want, which teams are playing tonight?

Under the hood it was genuinely ambitious. Oliver built and refined the six sub-agents over roughly three days, made each agent aware of the others so they share context, enforced visual consistency through Veezoo's query language, and even wrote up the whole system as a short academic-style paper.

Oliver — a few highlights from The Storyline Engine (not the full entry)
What the judges loved
This is where all three converged. Stefan called it “the most complete” of the entries, noting that “data doesn't speak for itself, it needs a medium to be transported,” and he loved the consensus-versus-debate framing between the personas. JP, wearing his technical hat, was drawn to the multi-agent design and the canvas builder, and appreciated the written paper. Andy pointed to the “incredible work with these six different agents, each with different perspectives.” The judges also kept noting how business-applicable the idea is. Imagine pointing a Cynic, a Romantic and an Oracle at your own company data instead of getting a room full of yes-people.

In the end, all three judges agreed. Oliver edges it. Congratulations, Oliver, winner of the Veezoo Analytics Cup 2026.

Watch the full sessions

We capped each finalist's prize with a signed copy of Andy Cotgreave's book, and the hardware to match.

🥇
First
MacBook Pro
🥈
Second
iPad
🥉
Third
AirPods Pro

If you want the deeper dive, the full interviews and the judges' deliberation are below. These are the unedited walkthroughs where each finalist talks through exactly how they built their entry.

The judging session and winner announcement
Oliver Yang, full interview
Dzmitryi “Dima” Yarmolin, full interview
Stelios Chiotis, full interview

Build your own

The most remarkable thing about these three entries is that every one was built by someone using Veezoo for the first time, in a matter of days, with no help from us. That is the whole point. Agentic analytics lowers the barrier to turning a dataset into something people can actually use.

You don't need a World Cup dataset to try it.

Point Veezoo at your own data and see what you can build.

Start a free trial

Ready to transform your analytics?

See how Veezoo can help your team get answers from data in seconds.