Designing Roguelikes for the Browser: Run Length, Death, and Meta-Progression
· ArcadeDeck Team · #design #roguelike
Roguelikes are the trickiest genre we ship at ArcadeDeck. The pitch sounds simple — procedural runs, permadeath, every attempt is fresh — but the genre's history is littered with games that nailed the first run and collapsed by the tenth. The difference between a roguelike that holds players for a hundred runs and one they bounce off after three is rarely the art or the variety; it is three tuning decisions made before any content gets built. This post walks through those three decisions using our own catalogue: Dicefall Chronicles, 40th Century: Void Survivor, and Kingdom's Last Stand.
Decision one: how long is a run, really?
The most-cited number for browser roguelikes is "twenty minutes per run." It is repeated everywhere because it sounds right, and because the genre's most-played PC titles converged on similar lengths. The number is misleading. The right metric is not the length of a winning run; it is the length of an average run, including the early failures. If a player's first ten attempts each end inside three minutes because they have not learned the systems yet, the game effectively trains the player to expect short, cheap sessions. When they finally earn a 25-minute run, the late game lands as a pleasant surprise rather than a slog.
Dicefall Chronicles enforces this curve deliberately. Early-floor failures resolve in 90 seconds; mid-floor runs settle around 8-12 minutes; only deep runs into the lower dungeon stretch past 20. The progression of run lengths matches the progression of skill, which means the time investment grows with the player's confidence rather than dropping it on them in run one.
40th Century: Void Survivor takes the opposite approach: every run is roughly the same length because the wave timer is fixed. The variance is not in duration but in how deep into the artifact pool you reach before the ceiling catches you. Both shapes work, but they imply different player behaviours. Dicefall rewards patience; Void Survivor rewards intensity. The wrong call would have been to copy one shape into the other game; the right call is to pick a shape that matches the run's emotional arc.
Decision two: what does death actually cost?
Pure permadeath — the original Rogue model — costs everything. Modern roguelikes almost universally soften this, but the softening is where most designs go wrong. Two failure modes show up consistently:
- Too soft. Death costs nothing, every run carries forward all upgrades, and the genre dissolves into a slow grind. Players stop reading the procedural content because the unlock economy is the real game.
- Too hard. Death costs everything, including knowledge — players who lose a 40-minute run learn nothing they can apply differently next time, because the next run starts at the same baseline with different RNG. Frustration compounds.
The sweet spot is what we call a knowledge tax: death costs the run's progress but not the lessons learned. In Kingdom's Last Stand this shape is explicit. You lose your army on a wipe; you keep the elemental matchup notes and the formation patterns you discovered. The next run starts at zero gold but with a smarter player at the keyboard. The death economy works when the second run is meaningfully better than the first, even with no carried items.
The shape of a healthy run distribution: many short failures near the start, fewer long survivals, and the median run lengthening as the player learns.Decision three: meta-progression — how much, and what kind?
Meta-progression is the most-debated lever in modern roguelike design. Two extremes:
No meta. Every run starts identical. Pure skill expression. Beloved by genre purists and a small minority of players. The retention numbers, in our experience, are unforgiving — most browser players bounce after their first three losses if there is no off-ramp.
Heavy meta. Every run permanently unlocks something — characters, items, starting bonuses. Retention soars in the first week; in the second week, players realise the unlocks are the actual progression and the procedural runs are a delivery mechanism for them. The genre's appeal — that every run is a fresh strategic puzzle — quietly evaporates.
Our internal rule is roughly 10-20% of total run impact comes from meta-progression, never more. Dicefall Chronicles unlocks new starter dice between runs, but the dice rolls themselves and the order they appear remain the heart of the game. Void Survivor's artifact pool is permanent but the pool is large enough that even after 50 runs you are still seeing combinations you have not built before. Meta-progression should make the player feel like they are getting better at the game, not getting more powerful in the game.
What this looks like in practice
Dicefall Chronicles ran a balance pass in March that cut starting-deck size by one die. The intent was to make early runs feel less generous and force riskier dice purchases earlier. The result was that median run length dropped from 11 minutes to 8 minutes, but new-player retention into a third session jumped roughly 18%. Shorter early runs, faster failure cycles, more attempts per session — exactly the loop the genre needs to keep working.
40th Century: Void Survivor is in the opposite tuning state. Top players have started clearing the standard mode reliably, which means the run length has compressed in a different way: the late game is no longer the question, the artifact mix is. The next pass, scheduled for May, will widen the artifact pool rather than buff enemies — adding decision space rather than damage. The change matches the genre rule: when the existing decisions stop being interesting, add more decisions, not more difficulty.
Kingdom's Last Stand is the youngest of the three and is still finding its balance. Early data suggests the elemental-matchup system is doing the right work for new players (the second run is meaningfully better than the first), but the late game converges on a small number of dominant formations. We are watching specifically for whether top-ten leaderboard entries are using the same loadout. If they are, we have a meta problem and will rebalance the relevant elements before adding new content.
Meta-progression should make the player feel like they are getting better at the game, not getting more powerful in the game.
The browser constraint, explicit
Everything above is true for any roguelike. The browser context tightens the screws further. Our players show up with seven minutes of attention; some have a tab full of work waiting. A roguelike that demands a 40-minute commitment for any meaningful progress has already lost the median session. We tune for the player who has time for one or two runs in a sitting and budgets accordingly. If a build cannot pay off inside two attempts, it does not ship.
Try the games
If you want to feel these decisions firsthand, three runs each will tell you more than this post does. Start at Dicefall Chronicles for the dice-deck shape, 40th Century: Void Survivor for the fixed-length intensity, and Kingdom's Last Stand for the formation-driven version. Each has a global leaderboard; tuning notes for any of the three land in this devlog when they ship.
If you have a roguelike-specific design question we have not covered, write in to help@arcadedeck.net. The most useful prompts are concrete ("why does Dicefall feel harsher than Void Survivor on a first run?") rather than general — those are the ones we can answer with the actual numbers.
More from the ArcadeDeck blog · Browse all games