There is a genre of football game that has nothing to do with playing football. You do not pass, shoot, or defend. You draft. You pull players out of a hat, slot them into a formation, and then watch a simulated result tell you whether your hunch was any good. A drafter that went viral last year ran on exactly this loop, and I could not stop thinking about it. So I built my own version in Godot, World Cup themed, and made it a roguelike. It is called World 11, and it is free to play in your browser. This is the devlog. What it is, how the pieces fit, and the handful of problems that were more interesting than I expected.

The Idea, A Draft, Not A Match

The pitch is simple enough to explain in one breath. You pick a formation. Then, for each of your eleven slots, you spin two reels: a nation and a year. They land on a real World Cup squad, say West Germany 1974 or Brazil 1970, and you draft one player from that squad into an open role. Spin again, draft again, until your eleven is full. Then you run the gauntlet and find out how far your patchwork team gets. The appeal is that your team is part dream team, part accident. You wanted a striker and the reel gave you Saudi Arabia 1994, so now you are deciding whether their best forward is better than leaving the slot for a later spin you might not like either. Every run is a string of small bets. That tension, draft now or gamble on the next spin, is the whole game. The football is just the dressing. Because you are drafting and not playing, the genre lives or dies on two things: the quality of the choices, and whether the simulated result feels fair. Most of the work went into those two.

One Spin At A Time

The draft screen is where players spend most of their time, so it had to feel good to touch. Each player is a card. Picking someone out of their natural position costs a rating penalty, so a brilliant winger shoved in at centre back is worth less than his number suggests. And there is a chemistry system: line up players from the same nation or the same decade and they connect, which nudges your team strength up. That last bit is what turns a list of names into a puzzle. Do you take the 88-rated player who shares nothing with your squad, or the 84 who is from the same country as three others you already have? The chemistry lines literally draw themselves between connected players on the pitch, so you can see your team becoming more than the sum of its ratings, or failing to. You also get three rerolls per run, which exist purely so that one catastrophic spin does not end your fun before it starts. Rerolls are a pressure valve. Use them too early and you have nothing left when the reel hands you a squad with no one who fits.

Real Players, Real Faces

A drafter is only fun if you recognise the names. So the game ships with 58 real historical World Cup squads, roughly 1,234 unique players, spanning 1966 to 2022 and every confederation. The data is just names, positions, years, and a calibrated rating. No invented players. The faces were the rabbit hole. I wanted the cards to show the actual footballer, but I was not going to ship copyrighted press photos. So I wrote a small tool that pulls only free-licensed images from Wikipedia and Wikimedia Commons, the ones explicitly released under Creative Commons or in the public domain, and skips everything else. It writes an attribution file crediting each photographer, because the licenses require it. Players with no free image fall back to a clean initials badge instead. The trap with that approach is disambiguation. Search Wikipedia for a one-word name and you do not always get the footballer. My first pass cheerfully gave me the Greek philosopher Socrates instead of the Brazilian midfielder, the comic-book Hulk instead of the Brazil forward, and an American businessman for a backup keeper named Luis. The fix was a verification pass that checks each page is actually described as a footballer, re-fetches the wrong ones through a football-specific search, and, where the name is too ambiguous to be sure, reverts to initials rather than risk showing the wrong human. A wrong famous face is worse than no face. Coverage landed around 84 percent, which is plenty to make the cards feel alive.

A Match Engine You Can Trust

You do not play the matches, so the simulation has to earn your trust. If it feels random, the drafting feels pointless. Under the hood it is a possession and event model. Midfield strength sets possession share. Possession and the gap between your attack and the opponent's defence set how many shots each side takes. Every shot carries an expected-goals chance to become a goal, weighted by the shooter's finishing, and goals get a real scorer and a minute. Knockout draws go to extra time and then penalties, weighted by the better team but never a coin flip. The important property is that the whole thing is deterministic from a single seed. Same seed, same draft pool, same fourteen results, every time. That sounds like an implementation detail, but it is what makes two features possible for free: a daily challenge where everyone in the world faces the identical bracket, and shareable challenge links where a friend gets your exact draft and tries to beat your score. The simulation being pure and seeded is the foundation both of those sit on.

Making It Hard Enough To Matter

Here is the problem I did not see coming. The first version was too easy. A decent team won the cup almost every time, which means winning meant nothing. I did not want to guess at the fix, so I wrote a headless harness that drafts hundreds of optimal teams and runs them all the way through the gauntlet, then reports how often they win and where the losers fall out. It told me the truth immediately: a strong, captained, attacking team was winning about 65 percent of the time. That is not a tournament, that is a coronation. The fix was two data-driven levers, no code rewrite. First, a difficulty ramp that makes each knockout round tougher than the last. Second, a match-day variance swing, a little randomness on the opponent's strength each game, so the favourite is never a sure thing. I tuned both against the harness until a strong team wins roughly a third of the time, with a healthy spread of heartbreak losses in the final. Now going unbeaten is the dream, not the default, and the harness is a permanent dial I can turn whenever the balance drifts. The lesson I keep relearning: when balance matters, measure it. A few hundred simulated runs tell you more in thirty seconds than an hour of playing on a hunch.

Watch It Your Way

The last piece was presentation. The original gauntlet just blitzed all fourteen results at you in a few seconds. Efficient, but it threw away the drama. So there are now two modes. Fast Sim is the old behaviour for when you just want the verdict. Match by Match steps through one game at a time with a live minute-by-minute ticker: a clock ticking up, goals popping in at their real minutes, your eleven on the pitch, and a bracket on the side showing how far you have come. Same simulated result, very different feel. One is a spreadsheet, the other is a broadcast.

What Is Next

World 11 is a complete, playable game today, but it is the foundation for a deeper one. The seeded engine is begging for a daily-challenge leaderboard. The collection that already tracks every legend you have ever drafted wants to grow into a proper codex. And the roguelike half is barely tapped, between-round boosts and modifiers are the obvious next layer. For now, the core loop is done and tuned, and I would rather ship it and watch people play than polish in the dark.

Play It

World 11 runs free in your browser, no download, no signup. Spin a nation, draft your eleven, and see if your frankenteam can go all the way. Play it at games.kevingabeci.com, alongside the other small games I am building one nights-and-weekends project at a time.