Sun et al.: Why Do Players Lose Themselves in Punishingly Hard Games? — Fukai Reads

TL;DR

This paper takes on a single question: why do people voluntarily pay to immerse themselves in games engineered to defeat them? Its subject is the Soulslike (the action-RPG subgenre typified by FromSoftware's Dark Souls, defined by punishing difficulty and frequent death). The authors push back on the conventional wisdom that friction-free comfort makes a good game, and try to explain how struggle itself becomes the raw material of enjoyment.

Their method is a qualitative analysis of 600 user reviews of Elden Ring, Sekiro: Shadows Die Twice, and Dark Souls III posted on Steam (the large PC-game storefront and community platform). From these they propose the concept of "resilient flow": a cognitive state in which absorption is maintained not by avoiding frustration but by meaningfully framing it. What is new here is reading players' psychology from the words they wrote themselves, rather than from a laboratory. Note that this is an arXiv preprint (submitted March 2026, not yet peer-reviewed).

Introduction — who, where, and why I chose it today

The paper is led by Zhehao Sun with eight authors in total, based at the Faculty of Creative and Critical Studies and the Department of Computer Science at the University of British Columbia (Okanagan), Canada. The work was funded by Canada's NSERC and SSHRC. It appears as an arXiv preprint (arXiv:2604.15318, in the cs.HC / Human-Computer Interaction category, submitted 2 March 2026), with no stated peer-reviewed venue. So the accurate reading, for now, is that this is a freshly posted paper not yet widely discussed.

I chose it today because it treats difficulty design — a daily headache for anyone making puzzles or games — carefully, in the vocabulary of psychology. Is difficulty a friction to be kindly removed, or a wellspring of enjoyment? I was drawn to the move of approaching that question through the voices of real players of commercially hugely successful titles. I re-brewed a strong hot drip coffee this morning and read it marking up a printed PDF with colored pens.

Background — the 'be kind' orthodoxy and the paradox of failure

Conventional game-design wisdom holds that players play to relax and relieve stress. Designers should therefore be "benevolent guides" who smooth away friction, and too much difficulty invites anxiety and disengagement — that is the standard reading of classic flow theory (the psychologist Csikszentmihalyi's account of an optimal state of absorption that arrives when skill and challenge are balanced). Flow is often drawn as a narrow channel between boredom and anxiety.

Soulslike games deliberately defy this benevolent model. With high difficulty they push players toward the anxiety boundary and force a state of hyper-arousal (heightened tension that sharpens attention). Yet players do not leave; they become devoted. Here lies the "paradox of failure" (the game scholar Juul's term for the paradox that people voluntarily seek out games that make them unhappy in the short term, for long-term gratification). What was not yet known was the concrete picture: how this paradox actually surfaces in players' own words.

Why does this matter? Difficulty is a variable nearly every game and puzzle designer touches every day. Lower it and the doorway widens but the bite thins; raise it and the bite grows but people leave. If there is a third road on this old tug-of-war — where the right way of raising difficulty increases both absorption and satisfaction — design guidance changes. The market data the paper cites projects the Soulslike market at roughly USD 1.5 billion in 2024, rising to about USD 3.2 billion by 2033. The market, too, suggests difficulty can be a driver rather than a barrier — that is the authors' starting point.

Approach — reading 600 reviews by hand

The authors' method is qualitative thematic analysis (reading large amounts of text and extracting and classifying recurring themes by hand). Rather than equations or survey tallies, the raw material is the words players wrote naturally. Bringing people into a lab risks distorting behavior through the awareness of being watched; the idea here is to avoid that and gather "naturally occurring testimony" from the Steam community.

There are three phases. First, building the corpus. From roughly 2,000 reviews they applied a strict filter: more than 50 hours of playtime (to capture skilled veterans rather than beginners' momentary irritation), reviews the community tagged "Most Helpful," and manual removal of low-effort or joke posts. This narrowed it to 600 reviews across three titles: Sekiro (rhythmic), Dark Souls III (reactive), and Elden Ring (open and exploratory) — representing three distinct variations of the combat loop.

Next, coding (labeling passages to classify them). First they pick up high-frequency words players use themselves — "dance," "rhythm," "greed," "zen" (open coding). Then they group these into broader meanings such as "internal locus of control" (the sense of attributing failure to oneself) and "cognitive occlusion" (explained below) (axial coding). Finally they map these onto flow theory to construct the three dimensions of resilient flow (rhythmic mastery, fairness of rules, mindful stress). It is a careful hand that builds theory by stacking up words rather than filling in unread details. No equations appear.

Findings — from violence to dance, from punishment to pedagogy

The findings are organized into three dimensions. The first is the "metaphorical shift." Early on, players describe combat with the vocabulary of violence ("beating," "killing," "surviving"). But upon mastery (50+ hours) the lexicon moves toward art and music. In Sekiro it becomes "rhythm game," "a violent waltz," "the music of swords"; in Elden Ring, "choreography," "a beautiful dance." One review (126 hours played) writes: "once you master the parry, you don't fight enemies anymore... you dance with them. The combat becomes a violent waltz." The perception of the enemy shifts from obstacle to dance partner.

The second is the "taxonomy of failure." The authors classified over 200 descriptions of "death" and found a strong tendency for players to blame not the system but their own psychological failings (an internal locus of control). Prominent in Dark Souls III is "greed" — dying because you got greedy for one more hit, a self-attribution. A representative voice: "I got too greedy trying to get another swing in and it cost me... The difficulty is sometimes unfair but 99% of my deaths are my fault." Precisely because failure is consistent and predictable, the authors read, the game functions as a "fair pedagogue."

The third is the "therapeutic paradox." The most surprising finding is how many players describe these stressful games as "relaxing" or "a release." The authors explain this through "cognitive occlusion" (because surviving demands 100% of the brain's processing capacity, external anxieties and intrusive thoughts are crowded out). One testimony: "I was in a very dark place in my life, full of depression... this game taught me patience and perseverance. It put me in a state where the only thing that mattered was the present moment." The three titles are also contrasted: Sekiro as speed-and-precision "Mushin" (no-mind), Dark Souls III as rhythmic endurance, and Elden Ring as the "wonder" born of scale and the unknown.

Use cases — applications for puzzle and game makers

From here I want to think about applications for those of us who make puzzles and games. First, stop thinking of difficulty as a single dial to turn down or up. What this paper suggests is that the key to absorption may lie less in difficulty itself than in fairness. If you are building something Sokoban-like (a box-pushing logic puzzle), reducing unfair dead-ends (unwinnable states recoverable only by resetting) and making boards where the cause of failure is visibly the player's own misreading leads to a design where players can feel "99% of it is my fault."

Second, treat failure cost and retry speed as explicit design variables. The loop the paper describes is "death → acquiring knowledge → retry." If you mass-produce levels via PCG (Procedural Content Generation, automatically generating levels and the like) for a hyper-casual game (a quick, casual mobile game), make what's lost per failure small but not zero, and keep the wait before retry short. The paper's point that completely painless failure lets attention drift connects directly to the feel of retry design.

Third, observe the "metaphorical shift" as an experiential metric. In the paper, the move in players' vocabulary from "violence" to "dance/rhythm" was the proof of mastery. If you gather reviews or playtest comments for your own puzzle game, track the change in the words players use. If the lexicon shifts from "frustrating" to "satisfying" or "I can get into the rhythm," you can read it as a sign that difficulty is converting into bite. It is a cheap, easy-to-run qualitative health check that needs no special equipment.

Fourth (as an adjacent application), reframe difficulty as a "device for focus." The cognitive-occlusion discussion offers hints for serious games (games with non-entertainment goals such as education or healthcare). If a moderate, focus-demanding difficulty temporarily crowds out real-world anxiety, that itself can be experiential value. But this is double-edged, and difficulty does not work for everyone (see the limitations section).

Limitations — what the authors admit, and what I noticed

Let us look at the limits. First, the weaknesses the authors themselves acknowledge. One is survivorship bias. The analysis is limited to players motivated enough to write reviews, and never sees the psychology of those who quit in frustration and never returned. So the conclusion that "difficulty produces absorption" is skewed toward the voices of those who stayed. Two, text analysis cannot measure physiological markers of flow such as heart rate. The authors list correlating with biometric data — heart-rate variability, skin conductance — as future work.

Now what I noticed reading it. What I, Fukai, point out here is, first, the danger that "fairness" is defined by working backward from the outcome. If a work that players call "my own fault" is deemed fair, that edges close to the tautology "people who got absorbed feel it was fair." Concrete design principles such as telegraphed attacks and consistent hitboxes are named, but the nature of text analysis means there is no ablation-style test (removing design elements one at a time to verify their effect) of how much each actually contributes.

One more point. This is an arXiv preprint, freshly posted, so citations are still scarce and it is not yet a widely validated piece. The sample is also limited to three FromSoftware titles, and whether it generalizes to other high-difficulty genres such as roguelikes or competitive shooters is unconfirmed (the authors themselves admit this as future work). "Resilient flow" is an attractive coinage, but for now I think it is healthy to receive it as a hypothesis and test it in small ways on your own project.

Fukai's Reading

From here is my own reading, Fukai's. I want to situate this study as one attempt to invert, in the vocabulary of psychology, a premise game design has long carried: that difficulty is a friction to be kindly reduced. In the language of design criticism, this is close to the work of redefining difficulty not as "punishment" but as "pedagogy." The claim that "fairness" holds the dividing line — whether failure is felt as punishment or as information — resonates well with my own instincts as a puzzle maker. Logic with no unwinnable states, telegraphed danger, consistent rules: in the end these are, it reads to me, another name for respect toward the player, something this paper teaches in a roundabout way. That said, I should add that this is my interpretation, and the authors do not assert it quite so flatly.

Closing — a map to related work

Finally, a map to related work. Those who want to go deeper into the psychology of difficulty and failure can pair this with Juul's "paradox of failure" (Juul, 2010), which it builds on, and with Hefkaluk et al. (2024), who studied how players recover from failure in the hard game Celeste — together they give a layout of the debate. For the origin of flow theory there is Csikszentmihalyi's Flow (1990), and for dynamic difficulty adjustment (automatically changing difficulty to match a player's skill) Fisher and Kulshreshth (2024) is a good entry point. How to design difficulty is a never-ending homework common to both puzzles and games.

References

Papers and related resources referenced in this article:

・Struggle as Flow: Challenge, Design, and Experience in Soulslike Games (Sun, Xu, Zhen, Lin, Thorogood, Lasserre, Dulic, Smith, 2026, arXiv preprint, cs.HC)

・DOI: 10.48550/arXiv.2604.15318

・Related work: Juul, J. (2010) "In search of lost time: on game goals and failure costs" (Foundations of Digital Games)

・Related work: Hefkaluk, N., Linehan, C., Trace, A. (2024) "Fail, fail again, fail better: How players who enjoy challenging games persist after failure in Celeste" (International Journal of Human–Computer Studies, 183, 103199)

・Related work: Csikszentmihalyi, M. (1990) "Flow: The Psychology of Optimal Experience" (Harper & Row)