I first encountered Darkest Dungeon in June, 2016. Everything about it looked enticing: tactical turn-based combat, dark comic art style, Lovecraftian setting & plot, high quality narration. In particular, I liked the stress mechanic: the idea that psychological trauma was as dangerous as physical damage was both cool and thematically appropriate. I was recovering from a sports injury at the time, so I had plenty of idle hours. I bought it and spent a month playing for a few hours after work each day. I completed almost all of the content, playing on "Darkest" difficulty with all of the intended settings enabled. My B-team wiped during my first attempt at the first of the four titular darkest dungeons (the end game content), and I didn't have the motivation to train their replacements. By that time, I had completed the rest of the game and my injury had healed, so I dropped the game without finishing. Overall, I considered it a positive experience.

Almost a year later, I discovered Joseph Anderson's critique of Darkest Dungeon on YouTube. Unlike me, he finished the game in its entirety, and had a profoundly negative experience doing so. In short, his criticism is that the game masquerades as being difficult while actually being tedious. The "rules" for each zone can be mastered within a few runs, after which each successive run consists of implementing those rules over and over again through effectively identical dungeons. Every dungeon (with two exceptions) plays exactly the same. Every boss needs to be killed 3 times, with zero changes to their mechanics. The only way the game can punish players is by wasting their time, as opposed to imposing interesting trade-offs or dilemmas. Losing heroes simply prolongs the campaign by forcing the player to spend hours training up replacements. This is made extra tedious by dungeon level caps which make grinding dangerous because high-level heroes cannot venture into low-level dungeons to shelter new recruits. This means playing conservatively, which makes grinding take even longer. The dread of Darkest Dungeon isn't found in the fates that might befall the heroes, it's in the time wasted to recover progress after heroes die.

I remember watching the review for the first time, comparing his critique with the experience of my play-through as I listened. I couldn't help but admit that his criticism was correct on several counts. All of the dungeons do play identically. There is a lot of grinding. The RNG is unforgiving, especially when one hero is chosen to be the butt-monkey for the entire run. The factual nature of his arguments nagged at me, yet somehow I still felt that he was wrong because I remembered having a good time with the game. Well, in June 2026 I tried playing Darkest Dungeon again - a full decade after doing so for the first time - and it was a completely different experience this time around. It was like revisiting a county faire I hadn't been to since my childhood, and realizing that it's actually a poorly-oiled tourist trap. The games aren't tough, they're rigged. The decor isn't rustic charm , it's neglect. Overall, the place only manages to maintain operations because it lacks competition and the customers don't value their own time. Joseph Anderson was right, and I was wrong.

Without getting caught up in a discussion about systems design, it suffices to say that a good game challenges players to master a system of interconnected mechanics. It presents a series of interesting decisions that force context-dependent tradeoffs, where deeper understanding allows for better decision making and increased risk tolerance. Feedback is legible, meaning results are quantifiable in kind & magnitude (signal clarity) and directly attributable to a particular decision (credit assignment).

Darkest Dungeon isn't like this at all. The marketing of deep party synergy is all veneer. In reality, it's a single rigid, unforgiving knowledge-check hiding behind layers of obfuscation and RNG. The challenge of Darkest Dungeon isn't in mastering systems for tactical turn-by-turn improvisation, it's about discovering a few specific meta compositions that solve the suite of unrelated combat mechanics. Every decision with which it presents the player is either irrelevant or trivialized by applying one of two dominant strategies. The only qualities that matter for a Darkest Dungeon player are patience and luck: can he endure repeatedly implementing the same strategy in a hyper-conservative way to survive any bullshit the RNG throws at him, and how much bullshit does the RNG throw at him when he cuts corners.

Dominant Strategies

Let's get this out of the way: Darkest Dungeon is a solved game. "Dominant strategy" is a game theory term for a strategy that is mathematically or practically superior, regardless of context. Darkest Dungeon has "two" such strategies. "Two" is in quotes because both strategies rely on the same underlying broken mechanics. If you understand which mechanics are broken, you can get away with all kinds of meme builds (though generally all of them are worse versions of one of the two dominant strategies).

The first dominant strategy is stun-lock. It's the safer strategy, and solves (almost) every encounter. The strategy involves picking a combination of heroes who are able to stun multiple ranks at once with a single move. By stacking speed (to go first) and accuracy (to guarantee a hit), the party can keep all of the enemies stunned while methodically picking them off and healing through any stress / health damage. The second dominant strategy is burst damage. It's slightly more risky, but works better against bosses and mini-bosses (the best way to kill a boss is to ignore whatever mechanic it uses and just burst it down). The strategy involves picking off dangerous enemies before they act, particularly focussing on the back ranks. It involves using the "Mark" debuff (which increases damage taken) and heroes that deal bonus damage to marked enemies. This strategy also involves stacking speed and accuracy, as the goal is to kill enemies before they can act.

It's noteworthy that rather than reacting to shifting combat conditions through calculated improvisation, both strategies ignore whatever skills differentiate one set of enemies from another and instead focus on denying them the ability to act at all. The community refers to this concept as the "turn economy": any turn an enemy gets is an opportunity for the RNG to screw you over, so the fewer turns the enemy gets, the better it is for your party. Stunned enemies skip their turn, dead enemies have no turn. It's also noteworthy that both strategies rely on maxxing out speed and accuracy stats. This is a natural corollary of the previous point: a miss is a wasted turn, and going first means denying multiple turns to your enemies. These are the aforementioned broken mechanics: once a player learns about the +8 SPD and 115 ACC breakpoints, he's effectively read the Dungeonomicon and is capable of building all kinds of meme builds. For example, 3 Antiquarians + 1 Leper: use Invigorating Vapours with each Antiquarian to grant +60 DODGE to the entire party, then bonk the enemies with the Leper. While off-meta, this strategy still abides by the turn economy: a dodged attack is a wasted turn for the enemy.

Just because a game has a dominant strategy, though, it doesn't necessarily mean that it's broken or un-fun. Many games have exploits, but are nonetheless enjoyable when engaged with on their own terms. Darkest Dungeon isn't one of them. The reason for this is that the game mechanics don't synergize, but instead exist in parallel. Light, food, health, stress, stun, bleed, blight, and debuffs are all just things to be managed to various degrees in isolation. Not even combat priority syngergizes with any of these mechanics: going first is just always the best strategy. Since there are almost no trade-offs between the mechanics, constraints on the solution space are generally 'AND' instead of 'OR', and the solution space rapidly collapses. The game unintentionally pushes players towards what might be considered exploitative strategies because they're the natural conclusion of the lessons it teaches players about what is effective. Whether on purpose or by accident, whether implemented well or poorly, every effective strategy in Darkest Dungeon is either a turn economy strategy, or a burst strategy. Thinking back to my first playthrough in 2016, my A-team used a burst strategy and my B-team used a stun-lock strategy ineffectively.

Obfuscating Failure

The viable solution space for the game is narrow, but the total solution space is very large: 17 heroes, each with 7 skills, able to stand in 4 positions and to move between them. If exploring the solution space is fun, then the game can still be enjoyable as a "journey, not destination" type of experience. This is not the case. Darkest Dungeon despises creativity & exploration, and punishes players for engaging in them at every opportunity.

Poor party composition is the number one thing that will doom a run. However, this is obfuscated by a lot of factors: skills, quirks, trinkets, and positioning. The game is also highly RNG dependent (when enemies get turns, that is). Practically, this means new players often experience doubt about whether or not issues during a run should be attributed to party composition: maybe the skills and positioning need tuning; maybe the hero trinket choices were suboptimal; maybe the run was a statistical outlier. In addition, attribution has a temporal component, too. 90% of the time, there's nothing actionable a player can do while being punished. Depending on the situation, the decision that's being punished could have been made as much as 30 minutes ago. In game designer lingo, this is known as "bad credit assignment".

In addition to credit assignment, game designers typically also consider "signal clarity": how clearly feedback is identifiable as good or bad. Darkest Dungeon struggles with this, too. Apprentice dungeons are tuned to have high tolerances, but those tolerances get tighter for higher level dungeons. In apprentice dungeons, badly put-together parties can scrape by ~50% of the time. At veteran level, this goes down to ~20% of the time. For champion dungeons, you can forget about scraping by: these dungeons cannot be completed without good party comp. The issue should be readily apparent: parties which seemed viable at lower levels, into which a great deal of resources have been invested, will suddenly become useless once they reach veteran level (level 3). In order to be useful again, an entirely new party of heroes need to be levelled up to the same level, and the party composition shuffled until it's viable for veteran dungeons. This means the player is forced to restart party composition experimentation with higher-value assets in an environment that's less forgiving.

Failure Cascades

The whole game, at both the combat and the logistical levels, operates on a system of failure cascade, where each mistake and misfortune makes the next one more likely. Critical hits from enemies inflict both stress and health damage. Healing wastes turns that could otherwise have been spent on combat (turn economy). High stress precipitates affliction: a state in which the hero is partially removed from player control and increases stress for the other party members. This of course makes additional heroes becoming afflicted more likely. At the precise moment when player agency is absolutely crucial to stabilize a run, the game partially relegates the player to the role of spectator. Rationally, it would seem that reaching this point would make retreat look like the wiser option. However, once a player has started a dungeon, he's under immense pressure to complete it.

Leaving a dungeon without completing the mission means not only is the dungeon reward forfeited, but the heroes who were sent in also take stress damage, contract diseases, and develop negative quirks. All of these things require spending money to fix. Meaning, retreating from a dungeon risks coming back with fewer resources than were spent to enter the dungeon. Meaning, players who enter a dungeon and discover that the party composition is bad are under pressure to push on anyways. But as mentioned earlier, dungeon runs suffer from a devastating snowball effect: every point of stress and health damage makes additional damage more likely. Once a player is in a tough situation, every encounter becomes a game of chicken: financial factors pressure him to keep going to at least break even, but doing so risks inflating the break-even target even higher. There's also the risk of passing the tipping point from which losing one or more heroes becomes unavoidable.

Roguelike Elements

The terms "roguelike" and "roguelite" have become overloaded. Purists would describe a "roguelike" as being a game of short, high-variance runs. The game must be completed in a single attempt, and the challenge is based around mastery over the mechanics. Deaths are frequent, absolute, instructive, and result in a complete reset of game state. Mastery is built through trial and error during gameplay. A "roguelite" is a roguelike with meta-progression that lowers the skill barrier over time in order to soften the blow and allow players to progress further into the game through not only mastery but time investment. All of the most popular roguelikes of the past decade are actually roguelites: Binding of Isaac, Nuclear Throne, Enter the Gungeon, Slay the Spire, Hades, Balatro. "Roguelike elements" is a meaningless term that can refer to anything: randomized loot, procedural generation, or limited permadeath.

Despite its rougelike marketing, Darkest Dungeon barely resembles one. While a single dungeon run can be compared to a roguelike run, in terms of duration and minor variations achieved through procedural generation, that's where the similarities end. While hero deaths may be swift, and the run is sometimes instructive, the key difference is that the deaths are not absolute: while the heroes are dead, the campaign goes on. It doesn't really resemble a roguelite, either, where each run is still completely independent and incremental progress is permanent. Darkest Dungeon is unique in that runs are not independent and the player can lose progress. The closest thing to which I can compare it is a Fire Emblem challenge run: "roguelike elements" injected into a 40-hour RPG campaign. It ends up being an astonishingly bad combo, because major mistakes provide neither interesting narrative divergence, nor a blank slate. Instead, they impose cleanup labour necessary to salavage the existing campaign, something players only tolerate because a few hours of busywork is still less than starting over from the beginning.

At this point, it should be clear why the stun-lock and burst strategies are dominant. When every turn the enemies take leaves a party exponentially worse off, and failing a dungeon might impose hours of busywork, the best strategy for a risk-averse player is a hyper-conservative one that doesn't allow the enemies to make any moves. This is really the heart of the problem: playing the game in interesting or thematic ways is dangerous, whereas playing the game optimally is slow and repetitive. This is how you end up with a meta where players abuse stun-lock to whittle away at enemies and drag out fights to infinitely cycle healing and stress-relief skills. Even if these issues were resolved, several other issues mentioned previously would still contribute to boring / repetitive gameplay: irrelevant dungeon goals, poor dungeon layouts, recycled bosses. Ultimately, the worst sin the game commits is being disrespectful of the player's time.

Getting Good

The issues mentioned so far are real, but don't on their own differentiate Darkest Dungeon from any other poorly balanced RPG on Steam. What does differentiate Darkest Dungeon is its overwhelmingly positive reviews, its categorization as a "tough but fair" game along the likes of Dark Souls, and its collection of superfans who defend it as an extension of their own ego.

One of the under-discussed negative consequences of Dark Souls is its role in spawning "git gud" culture. I think the origins of "git gud" culture are rooted in genuine constructive criticism and a competency counter-culture that emerged during the early 2010s. At the time, AAA gaming was pushing hard to broaden the market appeal of games by making them accessible to anyone, regardless of skill or commitment. It was the beginning of the modern era of gaming, with studios beginning to introduce many of the most reviled mechanics we know today: waypoints, yellow paint, narrator babysitters. Dark Souls directly rejected this trend, setting hard minimums on pattern recognition, reaction speed, and deduction. Much of its reputation for difficulty isn't based on actual difficulty, but the perception of difficulty when compared to other contemporary major releases by swathes of people who had only started playing games recently. It was in this context that "git gud" was born. Though sometimes used in this shorthand as a pejorative, often the sentiment underlying the statement would be elaborated in full: "There are a lot of play styles and a lot of bosses. Some bosses are VERY difficult with some play styles. To beat these bosses, you're going to need to really master both the play style and the boss mechanics. If you're unwilling to put in the time to do that, then use one of the many items the game includes to lower the difficulty, or change your build / play style. If you don't want to do that, or you're not having fun, that's okay. But don't blame the game. The community has already proven that any boss can be reliably beaten with any weapon, at any level, without taking any damage."

Surprisingly quickly, what started out as an honest defence of games with consistent rules that respected player agency and the assertion that the value of an achievement is proportional to the effort invested to earn it was perverted into a cargo cult for masochists. The directional relationship between exclusivity and mastery became confused, and exclusivity became an end in itself. Completing "difficult" challenges rooted in mastering broken mechanics, enduring boredom, and flipping coins against the computer until you won was reframed as a marker of merit because doing so was exclusive (forgetting that true mastery is elegant, reproducible, and appears effortless to onlookers). Criticism of a game's balance or design directly threatened the merits of the achievement for anyone who "mastered" those mechanics, particularly those who incorporated the achievement into their self-mythology of being "good" at games. "git gud" was repurposed as a defense mechanism for both bad game design and their own egos: "I suffered through it and you should, too. If you can't or won't it means you're just bad." Modern "git gud" culture is rooted in survivor bias and the Sunk Cost Fallacy. It's the guy with the "ate the most shit" award mocking you for refusing to eat any shit at all.

This is extremely relevant to Darkest Dungeon, noticeable if you spend any time engaging with discussions about it online. As an example, take this thread from the Steam forums. It's truly archetypical of Darkest Dungeon discussion board interactions:

• In 2015, a guy makes a post saying that he has trouble with slimes in The Weald. Not that they're particularly difficult or dangerous, just that fights with them drag on.
• Even in 2015, while the game is still in Early Access, already players are unknowingly grasping onto the skeleton key of Darkest Dungeon strategy: speed and stun-lock. Even the people offering competing advice like "AoE them all at once" and "focus one and DoT the rest" qualify it with admissions that his heroes will probably need more speed to implement it.
• In 2021, the thread gets necro'ed by someone sharing that he's had dungeon runs ended by slimes before
• By then, the stun-lock meta is already at least two years old, so we get our first "git gud" comment
• Another user joins in as devil's advocate. He posits that if a party is composed of low-speed heroes, is in a bad situation from an earlier fight, or is particularly unlucky, then a group of slimes could be a tough - even fatal - fight. He provides a link to a video of a Twitch streamer for whom all 3 of these things were true, and who experienced a party wipe as a result.
• User 'No One' arrives to set the record straight: that particular streamer is playing the game wrong. The reason he's playing the game wrong is because playing correctly (i.e. optimally) is boring (he literally says this).
• Another apologist chimes in to claim that slimes are a particularly high-variance enemy encounter, so it's possible to have an unusually tough fight against them even with good party composition
• User 'No One' sets the record straight again: if you ever have trouble with slimes under any circumstance in Current Year, you should git gud or quit. Back in 2015, it was okay to be ignorant, but not anymore.

In other threads, you can find superfans (including 'No One') acknowledging that turn economy strategies are exploitative. That's not to say that they admit any of the flaws listed above, though. Instead, they frame turn economy strategies as low-hanging fruit - a way of beating the game for those that sort of understand the mechanics, but not really. They assert that a player who has acquired the requisite knowledge the game attempts to impart can play the game "with basically any party comp". Some even claim that the game can be "easily" completed on Stygian difficulty with meme parties like Quad Antiquarian. Digging into these claims reveals the supremely degenerate mathematical exploits mentioned previously: maxxing out SPD, ACC, and DODGE ratings. Every build relies on a single formula, bolstered by specific quirks and trinkets to address class-specific shortcomings. What "You can win with any comp if you understand the game," actually means is, "If you exploit these 3 hidden stat caps, then it doesn't matter which heroes you field." Even so, many of these meme builds are unplayable outside of a single zone, excruciatingly slow, susceptible to getting stomped by RNG, or all of the above. Somehow, this is meant to refute that the game is mechanically flawed and repetitive: that a player profficient at exploiting flaws playing patiently can get by with suboptimal strategies for short periods of time, though this is both slower and riskier than using the optimal strategy.

It's worth dicussing who user 'No One' is. He's the author of one of the most highly-rated guides for Darkest Dungeon on Steam (literally titled "Git Gud"). In this guide, he teaches people how to play the most conservative variants of the stun-lock meta build. He has a whole section on stalling fights. In this guide, 'No One' offers insight like:

• Only 20 party combinations are playable, and only 15 of them are "good"
• Every party must have at least one hero dedicated to nothing but stuns, though 2 would be better and 3 doesn't hurt
• Stun-bot heroes should be equipped with 2 +STUN trinkets. If you don't have any +STUN trinkets then use +ACC until you find some.
• +SPD quirks are overpowered, so lock those in on all of your heroes
• Don't bother bringing any supply items into dungeons, other than food, torches, and keys

In a comment on his own guide, he declares it as proof that the game is rather easy, contradicting his own public statements in order to glaze himself. Factually, he is correct: the game can be beaten by using a simple strategy for the entire campaign that takes only 5 minutes to explain. The game even nudges players towards this strategy through harsh punishments, as they learn to ignore 99% of the 83,000 possible party combinations and nearly 300 unique trinkets. There's something else 'No One' is correct about: Darkest Dungeon isn't for everyone. However, I would adjust that statement slightly to make it more accurate: Darkest Dungeon isn't for anyone.