Inside Kognia's 2026 World Cup Hackathon
- 5 days ago
- 6 min read
Throughout June and July the eyes of the world were focused on the World Cup. At Kognia, we decided to use that moment to turn our own eyes, for one day, on our own data.
Earlier this month (July 2026) we paused our regular work to dedicate a full day to exploring 2026 World Cup data with no constraint other than a real football question behind each project.
The result was a set of projects that showed how Kognia's tactical data can be applied to answer questions that can truly impact how the game is viewed and analysed.
1. The space before the goal - Guillem Capellera, Research Scientist
Every goal looks inevitable in hindsight. This project asked a harder question of 32 goals analysed across 12 World Cup matches: in the ten seconds before the shot, did the scorer have more space than usual? And if so, who gave it to him?
For every square metre of the pitch, a pitch control model calculated the probability that the scorer would win a ball played there, then sums that control within a 20 metre radius around the ball. That number only means something compared to that same player's usual moments with the ball, which is why every goal gets translated into a percentile rather than a raw square metre figure. The model also separates two moments: the 10 second approach and the final two-second strike, to tell whether the space was already there or opened up right at the end. The median across the 32 goals sits at the 51st percentile, neither better nor worse than any ordinary moment of his with the ball, but the extremes tell opposite stories. Messi arrived with space that closed on him as he finished (from the 91st percentile down to the 29th), while Martinez manufactured his own in the final two seconds (from the 40th up to the 67th).
The interesting part is that the model doesn't stop at measuring space. It attributes it. By freezing each opponent at his position a few seconds earlier, it can tell whether a defender's own movement opened up the gap, or whether that defender was simply tracking a nearby run, in which case the credit goes to whoever made the run. In Porro's goal against Austria, for example, Konrad Laimer gives up 12 square metres in the final six seconds by tracking Oyarzabal's run, not because of a bad pass. The run created the space, not the ball. The single biggest concession in the dataset is even clearer: Azzedine Ounahi hands Mbappé 105 square metres in France's goal against Morocco. Of the open play goals, 16 came from a defensive movement conceding space like this, 10 already existed in the defensive structure, and 6 were finishes that simply won a control race in traffic.
2. Who makes the most of their time on the ball? - Álex Llinás, Senior Data Scientist
Alex’s approach focuses on the actual time each player spent on the pitch with the ball in play for his team (attacking phase) or against it (defensive phase). It measures how well he exploits the phase time he actually gets.
The first finding confirms the idea works: 14 of the top 20 attacking "exploiters" by this metric play for teams with under 50% possession. In attack, names like Galarza, Gyökeres, Luis Romo, and Kolašinac stand out, producing far more than their level of involvement would suggest. In defence, the pattern flips: attacking players like Bruno Fernandes, Pedri, Julián Álvarez, Cristian Romero, and Kenan Yıldız contribute far more without the ball than expected. In the combined index, the top spots go to De Bruyne, Messi, and Mbappé on the attacking side, and Pedro Vite and Paquetá on the defensive side.
3. Git Gud: moments where a team finds another gear - Antonio Rubio, Senior Computer Vision Engineer
This project draws on Kognia's tactical detections and tracking data across 100 matches of the World Cup, not a single fixture, and tells three stories about what makes a team "wake up." Hence the name, borrowed from gamer slang "git gud" (practice and get past the challenge instead of complaining about it): turns out teams do that too.
The first is "Win or go home": it compares the 72 group-stage matches against the 28 knockout matches (regulation time only) and finds that knockouts are tense early and frantic late. Only 0.43 goals per match arrive in the first 30 minutes in knockouts, versus 0.85 in the group stage (49% fewer), but 30% of their goals land in the last 15 minutes, versus 26% in the group stage. Two behaviors cross over each other here: among teams that are losing, the share of directly attacking actions grows as the match goes on, more so in the knockouts (5.8% in the 75'-90' window versus 4.5% in the group stage). Meanwhile, the team that's winning drops back. In the group stage it holds 49% of possession while ahead; in the knockouts only 41%, letting the team chasing the game reach 59%.
The second picks up exactly where the previous case left off, the hydration break, but now with the full 100 matches and a much clearer view. For each group of actions (progressive passes, defensive duels, regains, pressing, carries and take-ons, finishing, box entries...) it measures the change in detections per minute between before the break and the first 5 minutes after. With the full tournament, the answer is that yes, something does change: progressive passes go up 13%, defensive duels and regains up 10%, pressing and carrying up 9%. Only sideways passes go down.
One concrete example illustrates it well. In the Norway Vs England quarter final, England controlled the pitch before the break (up to 74% pitch control in the stretch right before it), but Norway clawed back control as soon as play resumed, edging toward a 50-50 split, until an English goal in the 46th minute tilted the balance back their way. The break woke Norway up. The goal woke England up again.
4. Player Lens: turning "who stands out" into a tool - Pawel Basiak, Senior Frontend Developer
Instead of answering one specific question about the World Cup, why not build the tool that lets anyone ask that question about any player?
The result is Player Lens, a prototype that compares a player's performance in one competition, the World Cup say, against his usual level in another, within his position group. The comparison runs on a "lens": a weighted combination of metrics the user picks and weighs themselves, from off-ball movement to goals or shots, which can be created, saved, and iterated on live. Once a lens is set, the tool classifies each player as a breakout performer, hidden contributor, or consistent performer, and generates a natural-language explanation of why he stands out, all on top of a player map and a rankings table. It's essentially the tool version of one of the ideas the hackathon itself suggested: new ways of grouping players by filtering on metrics, except here the axis is defined by the user, not the analyst.
5. Putting data in context: a chat that understands what you're looking at - Marlon Becker, Senior Frontend Developer
This project doesn't start from a specific football question either, but a product one: when someone is looking at a team's or player's page, how do we give them more context without making them go dig for it? The exploration combines several pieces connected to Kognia's MCP and other internal sources, more as a demonstration of what could be built than a single finished feature.
The first piece is a match momentum chart with range selection. You can drag across the match timeline to zoom into a stretch, minute 15 to 30 say, and instantly see how both teams compared just within that window.
The second is an "explain this view" button on every metric row, which generates, on the spot, a natural-language read of that number against the rest of the competition. For example, France's 6.53 shots on target per 90 nearly double the tournament average, putting them at the 78th percentile.
The third is a chat connected to the MCP and to Kognia's metric documentation, which answers using as context the player, team, or season currently on screen, with no ambiguity, since that context travels as an internal identifier. Asked about Mbappé's goals, it doesn't just list each one's minute. It points out that 6 of his 8 goals came in the second half, and suggests the natural next question. The same chat also lets you look up, without leaving the screen, the full definition of any metric.
Comments