Live, Die, Upgrade, Repeat: The Modern Roguelike

By Jon Bailes and Luke Shaw on at

When most games today focus on streamlined experiences that promote steady progress, there’s something almost anachronistic about the roguelike, with its constant restarts and uneven level design. And yet it’s in this environment that the formula has blossomed and spread into countless different genres, precisely because of those factors. With this summer’s Dead Cells the latest of its kind to find success, it’s clear that there are still countless ways to tweak that formula into subtly different, compelling results.

So what is it that’s made some of these games especially memorable in the last decade? What makes a roguelike (or 'roguelite') work, and how have the core features evolved with their expanding appeal?

As we discovered in this discussion of our own experiences, it’s a design approach that’s used to hook us in numerous ways. What was clear, not least with Dead Cells, is that we had often taken different things from the roguelikes we’d played. Perhaps, as much as anything, their enduring attraction is in meeting such varied expectations.

Jon Bailes: The base reference for me with roguelikes is Mystery Dungeon: Shiren the Wanderer. It was ten years ago now, and before I'd even played it I'd heard about this turn-based dungeon crawler with randomly generated levels and permanent death. At the time, those concepts didn’t sound all that enticing, but were obviously resulting in some fascinating experiences. It was apparent that Shiren created stories; not in terms of in-game narrative, but in the way players would share tales of their adventures, usually based around terminal misfortune.

Shiren the Wanderer

I soon developed my own love-hate relationship with Shiren, enjoying the sense of mystery and discovery, while cursing its near-bottomless capacity for sadism. It is often a mean game, allowing you to build in confidence as you obtain strong weapons, a sturdy travelling companion and a healthy supply of consumables, only to smash it down again by suddenly plunging you into darkness, ruining your gear with a rust trap, teleporting you into a monster house, or turning your rice balls rotten.

But as much as Shiren seems to revel in your failure, it also provides the tools to cope, and your ability to manage gear, develop contingency plans, assess risk and improvise becomes crucial. The wild unpredictability is what grants you space to be resourceful and imaginative, leading to tense, exciting journeys with emergent scenarios that are often worth recounting.

So this experience is what I’ve taken forward into games that have repurposed the core roguelike elements – randomness and permadeath – in various genres since. There are so many different versions of this design approach now, but I think the measurement of their success or failure remains constant: whether or not we end up telling stories.

Another obvious leader in this regard for me is Spelunky HD, which basically embodies the same philosophy as Shiren. If anything, it’s even more punishing. There are more instant death scenarios and, without the crutch of turn-based play, it rarely affords the player time to stop and consider their options. A slight platforming malfunction or moment of indecision can wipe a promising run in a second.

Yet again, the flipside of that is the scope for improvisation, and how each run is markedly different and demands your full attention. Sometimes there’s a sudden curveball, like a darkness or beehive level, that instantly complicates your progress. But on a more moment-to-moment level there’s the constant temptation of risk-reward opportunities. Often the reward is much smaller than the risk (a gold bar or ‘damsel’ on one hand, instant death on the other), but still hard to resist. It creates an adventurer’s mentality, which makes each game play out differently, and explains why so many stories end in the player cursing their own stupidity.

Spelunky HD

Luke Shaw: roguelikes have always intimidated me. They are precise, exacting, and punish lax play with a hard restart. When I first tried Spelunky HD, I had some idea of what I was getting into. I'd heard of Shiren, and I knew the general elements that define a roguelike. Sadly, I found it very frustrating and bounced off . Subsequently I grew to admire how it’s constructed by watching it being played by others, and reading about the game's secrets. That was the real value of Spelunky for me, because it set off an interested in exploring other roguelikes as I realised the depths that this genre could create, how their discrete parts needed to be constantly assessed and reacted to,and the way in which players gradually came to understand requiring each game’s systems.

As esoteric as many of them are, the action layer of some modern roguelikes provides a simple entry point. The first I really invested time into was Crypt of the Necrodancer due to its punchy immediacy. The fundamental alchemy behind most roguelikes is in the slow absorption of knowledge that empowers the player, but to getting to that point is helped enormously if the game has a clever and immediate hook.

Necrodancer is immediately playable and fun, with little knowledge required. The focus is squarely on the memorisation of easily identifiable enemy patterns and acting to the beat of the soundtrack, which in itself is a smart twist on the classic early roguelike motif of actions happening when the player moves. Instead of allowing the player to consider the consequence of every action, Necrodancer encourages momentum and reactive play – asking the player to move with the flow. Dexterity (initially) trumps knowledge, making it easy to pick up.

There is still depth, but the process of learning the constructed language of play is less pronounced. A roguelike with a similar immediacy but significantly more depth, and what I moved onto next, is The Binding of Isaac. Including its two expansions, Rebirth and Afterbirth, I’ve played around 150 hours, which is only a tenth of what some of my friends have. Like Spelunky it takes the form of a familiar game: a dungeon crawler in the 2D Zelda mould, but with the added bonus of near-bottomless depth.

The Binding of Isaac

Unlike Spelunky and Necrodancer, which have eminently knowable lists of enemies and items, Isaac is about fecundity. A gross hoard of items, modifiers, room configurations, enemies, and secrets all combine together to create a huge, overflowing sandbox of possibilities. The key to its longevity is in how it couples the quick gratification of a 30-minute completion time with the continual drip of new items and mysteries. The sheer wealth of variation in items, and how they interact with each other in absurd but nearly always logical ways, means that there’s something to look forward to in each run.

Isaac lets players experience something that most games simply don’t have the budget or scope for - a near impossible amount of replayability, and a comically huge roster of pickups. Its appeal for me lies in the fact I will never conquer it, and that each run is largely consequence free. There is little negative about playing Isaac, it both enthrals completely and acts as a perfect podcast or second screen game. Playing in its possibility space is rarely anything other than a positive experience, one where you can either embrace the randomness of a session, or exert your agency as a player to push the systems in your favour.

JonIsaac is very good at what it does, and I think also exemplifies how roguelikes started to move away slightly from the strict masochistic pleasures of Shiren or Spelunky. Presumably the aim with many modern roguelikes is to reduce the sense of unfairness that often goes with the territory. With permadeath in particular there’s always potential for demoralisation, as a game repeatedly boots you back to the start. In Isaac, the brevity of each run clearly helps, and that sense of constant progress that comes from unlocking new gear, and that’s really become a fixture of modern roguelike design. There’s some accumulative continuity even in Shiren (store houses) and Spelunky (shortcuts), but only as an option that integrates into the risk-reward mechanism. A steady drip of new powers is different, and in some cases I think it reduces the stakes too much.

A case in point is Rogue Legacy, where an ‘inheritance’ system between characters allows you to spend gold collected on your previous run on permanent stat increases and class upgrades. Combined with level design that packs screens with HP-sapping enemies and traps, it’s clear that early-game characters aren’t meant to survive long, and the goal is to merely grab enough gold before their demise to ensure the next character is slightly more resilient. The result is gradual but consistent progress, but that makes it less a learning process than a grind.

Rogue Legacy

In the best roguelikes, once you’re familiar with the pieces every new start is exciting because it could be ‘the one’: that attempt where fortune, skill and judgement finally align and lead to victory. In Rogue Legacy, each time you set out you have a sense for far you’ll get. You know when you’re ready to beat an area boss, after which you'll enter a new area where you struggle for a while, until you power up further. If you keep playing, you’ll win, because you’ll become strong enough to beat the toughest enemies. Until then, you won’t. This is why I’ve never really read a Rogue Legacy story, and nor do I have any to tell. I simply kept playing until it was done.

In general, I’d prefer to see more creative approaches to permadeath, rather than an over-reliance on consolation prizes. One good example in this respect is Invisible Inc., which almost shouldn’t work as a roguelike because of its hours-long run time, but in fact uses its identity as a turn-based squad-tactics game to its advantage.

Invisible, Inc.

Crucially, it’s not a game that wants you to fail repeatedly. Instead, it wants you to see the end and then refine your strategies with new characters at higher difficulty levels. It’s therefore structured so that losing a character or failing an objective is a setback rather than game over. Missions also go bad over multiple turns, rather than with one false move, and there are often opportunities to save situations even as the guards close in (including the last resort ‘rewind’ option). The onus is on how you prepare for the finale and get yourself out of jams along the way, regularly resulting in audacious last-minute escapes, and unlikely tales of success.

Luke: Invisible Inc. is definitely one of the greater modern roguelikes out there. That has a lot to do with how perfectly pitched its alarm system is, how well it presents information to the player, and how the gradual descent of chaos is preferable to the sudden deaths of Spelunky and downward spirals of Isaac. I also agree Rogue Legacy's slow incremental unlocks don't really work; Isaac is full of unlocks too, but they broaden the experience, rather than 'levelling up' the player in a linear way.

Counter to the aforementioned, Darkest Dungeon employs a lot of smart design decisions to make its blend of knife-edge drama and success by attrition work. Firstly, its randomness and its apparent vindictiveness are made clear from the offset. Characters die fast, they gain negative traits at an intimidating pace, and victory often seems out of grasp. Unlike Invisible, Inc. things don’t go wrong by degrees in Darkest Dungeon – plans fall apart apart before your eyes. Enemies get critical hits, traps spring because you weren't paying attention, investigating a curiosity ends up turning sane men insane. It’s a house of cards that’s constantly falling apart as you race to patch it back together.

Darkest Dungeon

Like Rogue Legacy, it also requires you to grind for resources to improve your estate, allowing your roster of characters to upgrade and survive longer. Unlike Rogue Legacy, it’s utterly stuffed with stories. Every player will experience a mission where the odds swing out of favour, where an innocuous fight will spiral into disaster as your party loses their sanity, becomes unresponsive, and self-flagellate as the darkness closes in. But then there are those glory moments, when a fiercely pitched battle seems over and your entire crew's at death's door but, improbably, they pass their instant death tests, and manage to slay the monstrosity in one final heave.

The constant progression allowed by upgrading the estate and a stream of fresh recruits blunts the harsh outcomes of these encounters, but the rhythm is never as dull as Rogue Legacy. Both rely on a similar progression system to different ends. Legacy is a slow grind, and almost inevitable victory with hiccups along the way, whereas Darkest Dungeon is the reverse – moments of glory that shine amidst a quagmire of a failure.

Playing smart and learning the game is always more worthwhile than pursuing a grind. Randomness, or at least some semblance of unpredictability is often a roguelike’s greatest asset, but it needs to be tempered with the ability to learn it so that the player isn’t solely forced to put away hours to guarantee success. Darkest Dungeon never stops being gruelling, but it always feels like learning is rewarded. You’ll never fully temper the harsh roll of the dice, but at least it’s a battle between your brain and the systems at work rather than a slow plod towards a certain win state.

Jon: Randomness is of course the other important factor in roguelikes, and for me those cruel twists of fate are vital to the experience. But something like Darkest Dungeon almost seems anomalous now, as many games narrow the scope of procedural generation. They may be more accessible for it, as they reduce the chances of unfair convergences of events, yet can then struggle to provide the necessary variety.

I’m particularly thinking of Enter the Gungeon here. For a start, it’s rarely important how the self-contained rooms in each level connect together. For the most part, you enter a room, shoot everything in it, move to another and repeat. How you enter, from which direction, or the order you clear the rooms in aren’t major strategic choices.

Enter the Gungeon

In contrast, the way each level works as a single live space in Shiren or Spelunky means slight layout variations have significant effects, even before we consider the interactive components. And Invisible Inc. shows how this factor can produce great variety and a smoother challenge. Something as small as finding the level’s exit near your starting point, as opposed to at the end of the facility, or a central hub where numerous guard patrols intersect, affects your strategy and route plan. When you juxtapose these possibilities with the gameplay variables – enemy types, optional goals, character skills, the rising security level – no two missions are the same, yet most end up well-balanced.

With Gungeon, it all comes down to the content of the rooms, which are too safe. There are differently shaped and sized spaces, and some environmental features, but it rarely matters. You either take cover or move depending on the type and proximity of the enemy. It’s the same template as Isaac, of course, but again the briskness and variety of a run makes the difference: the relative compactness of Isaac's levels, the speed of play, and the greater variety of  scenery arrangements add up to something much more significant.

The basic mechanics of Enter the Gungeon are solid, and the recent Gungeons and Dragons expansion helps a little, but the sense of repetition still persists. Restarts are especially problematic, as you shoot the same low-level enemies in the same rooms with the same basic pistol all over again. In the end, too much rests on the randomness of weapon and item drops but, despite this, they lack real tactical variety, and the item pool is so diluted you may unlock interesting gear then hardly ever see it again.

Gungeon doesn't link rooms in interesting ways, or have enough risk-reward mechanics; say, rooms with time targets, destructible items in the line of fire, or enemies that must be shot in a specific order. Without that kind of variety, it's no surprise that Gungeon is a commercially successful roguelike but, beyond a few boss battle heroics (the game’s recognition of no-hit victories is valuable here), there just aren't any great stories about it. In most cases, my attempts merely petered out, as I made a few too many careless mistakes, or, more depressingly, just didn’t have the firepower to tackle the later stages.

Luke: I was really disappointed to find that Gungeon didn’t have the depth past its initial spark. As you’ve pointed out, the basic premise and mechanics are solid, almost masterful, but there’s no real sense of learning beyond attack patterns. One of the best experiences in a roguelike is your knowledge tipping you off to the possibility of an unexpected combo and Gungeon doesn’t really offer many chances.

In Isaac combos are everywhere, from the homing tears synergising with lasers and brimstone beams, to mass minion spawning, infinite bombs that don’t hurt you, and many of a number game breaking builds. Gungeon’s immediate peers – the crunchy Nuclear Throne and sci-fi Galak-Z – fizz with the energy of mutations or weapon randomisations that all slot together in interesting ways and make Gungeon feel slight in the process.

A recent and excellent example of how knowledge can allow you to build these combos is Slay the Spirean exceptionally moreish card battle based roguelike. Like the best in the contemporary genre, it is incredibly immediate, but staggeringly deep. Combat seems simple enough – you have energy you can spend to play cards that either deal damage, or stack defence on your character. The numbers are small, usually in the single digits, modifiers are also small, but tend to scale up quite quickly.

Slay the Spire

The twist comes from each character having a library of cards that work around specific keywords and concepts either universal to the game – such as drawing, or exhausting – or unique to them, such as the Ironclad’s Strength which increases damage, or the Defect’s Orbs that can attack or block at the end of the player’s turn, amongst other things.

A typical session involves progressing through a randomised map of rooms that contain various combat and non-combat encounters, before tackling a boss and ascending to the next floor. After each fight, you’re rewarded with the ability to draft one of three cards to your deck. It’s initially simple, but the depth is in learning what action is optimal, or most enjoyable to take in a given instance. With few hard limits on abilities activating, many cards can react off themselves and other relics exponentially, which leads to powerful and potentially infinite combos.

Slay the Spire initially dependent on playing enough to unlock the relics and cards that lubricate the smarter synergies, but the real power is how your knowledge of the game lets you parse the choices of each run. Weighing up items and cards against each other to make the best decision, and learning when not to take cards and items when your deck has achieved the sort of elegance and pacing it needs is incredibly important. It’s possible to beat the game at the highest difficulty time and time again, proving that the slow accretion of knowledge, and knowing how to use that knowledge to bend the statistics and randomness to your advantage, is the key to success.

Jon: I can’t comment on Slay the Spire (those PC gamers with their early access), but it’s certainly true that one of the great things in roguelikes is the way stacked systems and random elements encourage creative solutions, and that openness means there’s always a possibility you can upset the balance in your favour – abuse of cloaking rigs in Invisible Inc., for example. That said, I’m usually playing at a lower level, where the fun comes from having to improvise each run, rather than bending the game to my will.

This is a good point to bring in Dead Cells. In many ways it represents a refinement of the trends we’ve been discussing, in that it combines low-level randomisation with a focus on complex weapon synergies and incremental progress through permanent upgrades. But it’s also relevant in regards to this last point about players controlling the game’s systems, which here I think actually works against the classic roguelike elements.

Dead Cells

There’s a huge amount to like in Dead Cells, of course. The speed and responsiveness of the main character produces an exhilarating flow and precision in combat. The range of gear offers plenty of tactical scope, equipment drops are frequent enough to encourage experimentation, and the alternative routes and personalities of each level add variety. So early on, there are definitely stories to tell, especially due to the powerful combined effects or amusing lop-sidedness of particular weapon selections. Indeed, the centrepiece of any good Dead Cells story is a screenshot of the player’s set up: the picture that paints a thousand violent words.

But the further I progressed, the more this started to fall away. For one, there’s little incentive to take alternative routes after you’ve seen them once, and many levels are surprisingly rigid in structure. Particularly strange, however, is the way some permanent unlocks actively undermine the random factor. Once you can reforge buffs on your weapons, choose the weapon type each shop sells, and then pay to switch its stock, you’re pretty much creating an optimum build each time, rather than constructing a strategy from a configuration of options unique to that run.

By the time I was actually trying to finish the game, I repeatedly chose a preferred strategy – following the default path and aiming towards specific weapons (often involving an ice bow/blast and sinew slicers). And with the levels themselves lacking potential to surprise, this meant I was effectively playing the same game every time. Worse still, I could nearly always breeze through to the final boss, only to get killed and sent back to the start before learning its patterns. After several near-identical failures, and some attempts with alternative builds that were less effective, it seemed ridiculous to keep repeating that 40-odd minute slog, regardless of any new upgrades. I just wanted to finish the game without playing through it all again. The stories had dried up.

Luke: As you might expect, my experiences with Dead Cells were slightly different. I wonder if this is due in part to the way that I played it – routinely putting around 10 to 20 hours into various patch builds, appreciating how they chopped and changed various elements whilst retaining the core structure. That's obviously not the same perspective as a player with the 'finished' game.

I think Slay the Spire achieves the dense and intricate ruleset that only the likes of Spelunky and Isaac have managed before it, Dead Cells has been a revelation of sorts amongst many of my friends and peers, who maybe aren't quite as interested in the roguelike history that fed into it. The game is instantly captivating with its well-oiled combat mechanics and impressive presentation, but under the surface the same roguelike heart is ticking away.

Dead Cells

Dead Cells' weapon list is varied and situational, from brutish broadswords and cruel whips to magical orbs of death and elemental powers. In the early game it’s easy enough to fumble through with a haphazard selection of weapons, unlocking any blueprint that takes your fancy and picking mutations based on whether they feel right at the time. As you progress it becomes clearer that the game’s roguelike heritage allows for the same build-up of knowledge that the best in its class invites. Weapon synergies become obvious and as intoxicating as anything in Isaac – simple combos like an ice bow and a crossbow with a bonus to frozen targets allows you to delete enemies with the satisfactory and bombastic crunch of the DOOM shotgun.

Further progression past the first boss introduces the legendary forge, and the game’s main source of progression and investment. After dumping hundreds of cells into the Forge the rate of higher rarity weapon drops is increased, which means more effects, and more chance for tight synergy. Throw in the ability to reforge weapons, mutators that play off specific stats and damage types, and the combinations start to grow exponentially.

It’s true that Dead Cells feels more limited in scope than some of the greats, and the chance for interesting stories is narrower, but as you climb through the difficulty levels, knowing when to stick or fold with certain items is incredibly important. Coupled with the danger of risky modifiers the ‘one more go’ factor is high. It keeps its randomisation slight but meaningful, and you’re rarely left feeling like you can’t nudge a bad hand in a slightly better direction. Restarts are swift, and death is even swifter, and the pool of items and modifiers is judged such that it always feels like, even after hitting a triple-digit playtime, there's more to see.

Jon: I certainly appreciate the some of the design subtleties in Dead Cells, especially around the flexibility of the weapon system, but it overly dilutes the experience that first attracted me to roguelikes. Guess I'm more ‘purist’ about the implementations of roguelike elements I enjoy, despite the frustrations they bring, or leftfield approaches like Invisible Inc. which still rely on core mechanics for replayability. I’m certainly looking forward to Spelunky 2. I just hope they haven’t made it any harder.

Luke: Despite my initial claims that the exacting nature or roguelikes was what had pushed me away from them, it turns out that it was that very nature that turned them into my genre of choice. Whilst Dead Cells was the initial impetus for us to write this article, and I agree it presents a great synthesis of elements, it isn’t the be all and end all. I love that it has found purchase with a new audience that, at least in my limited experience, were warded off the esoteric nature of many roguelikes, and I am excited to see what they make of the meticulous Slay the Spire when it lands on switch in the near future.