Launch Notes: Ranked Duels and Real Practice

AlgoArena launched around one core belief, that programming practice should feel closer to real problem solving than memorizing answers in a static editor.

What shipped

The first public version centered on input-based algorithm problems, ranked 1v1 battles, ELO, match history, replayable rounds, public leaderboards, and AI-assisted review. A learner could enter a duel, write code against real inputs, submit, and see how the round unfolded.

That starting point mattered, because the goal was pressure, feedback, and repeatable practice in one focused place instead of a prettier problem list.

Ranked play used ELO, the chess rating system in which beating a stronger opponent earns more points than beating a weaker one. The launch model also weighed each problem's difficulty. A round on a hard problem moved rating more than a round on an easy one, and a win and a loss on the same problem were not worth the same number of points. Matchmaking paired players by rating alone, dropping the tag and category preferences that earlier builds had matched on. An ELO history chart showed each player's rating over time.

A player who left a match lost it. The opponent got the win right away, and the player who left saw a defeat screen. Matchmaking was hardened against queues that never resolved and against phantom matches, where a player was paired with someone who had already left. Practice bots could take the other seat, so a battle was always ready to start. Players could report a broken problem or an opponent. After a round, a player could read the opponent's solution. Battles also recorded each player's keystrokes, which is the data that match replay runs on.

Between battles, the practice page drew from a live problem bank, loaded a page at a time, with filters for tag, difficulty, and rating. Error messages in the practice IDE were reported test by test. A daily challenge came with a streak that counted the days in a row a player came back to solve it.

Players could add friends, invite them to a battle, get notifications, and open each other's public profiles. Profiles recorded achievements for practice and competitive milestones. A global ranking ordered players by rating, and a promotion screen marked each step up to a new rank.

Why it matters

Input-based practice changes the feel of a problem. You parse the input, handle the cases, and produce output the way real contest and interview systems expect, with no multiple choice to fall back on. Ranked duels add another layer of pressure by putting your work up against the clock and an opponent's pace.

That pressure is useful only if the review loop is strong. Match history and AI-assisted analysis gave learners a way to revisit the round instead of treating a loss as a dead end.

Where it points

The launch version became the foundation for Compete and still shapes the rest of AlgoArena. Classroom borrows the live surface. Assessments borrow the evidence mindset. Vibecoding borrows the idea that process should be visible.

For more on the practice philosophy, see Ranked Duels and Deliberate Practice and Why Input-Based Problems Matter.