What 15,649 TFD Comments Say About Why Players Leave, Return, and Leave Again
I analyzed a corpus of 15,649 YouTube comments and replies from 13 videos discussing the state of The First Descendant, why players stopped or reduced their playtime, and how opinions changed over the game's lifecycle.
The dataset contains 9,341 top-level comments from 8,184 unique authors and spans discussions from TFD's launch period through 2026. The analysis was compiled on August 17, 2026.
TL;DR
- Repetition is the strongest recurring signal, appearing across all 13 videos — but the deeper complaint is not simply “too much grind.” It is repetition without enough experiential variation.
- Acquiring a build and having something meaningful to do with it are separate problems. Many comments move from “How do I get this?” to “What do I use this for?”
- More content does not automatically mean more gameplay variety. Different rewards, regions or modifiers can still ask players to solve essentially the same gameplay problem.
- A meaningful portion of disengagement looks like dormancy rather than rejection. Many former players still like the combat, characters or foundation and remain willing to return.
- My overall interpretation: TFD has many answers — Descendants, weapons and builds — but not enough different questions for those answers to stay meaningful.
## 1. Repetition is the strongest signal, but “too much grind” is an incomplete diagnosis
The broadest recurring complaint in the corpus is repetition. Conservative heuristic coding tagged 1,539 of 9,341 top-level comments, or about 16.5%, for repetition/grind. More importantly, the signal appears strongly enough in all 13 of 13 videos.
That sounds like a simple result: players dislike grinding.
But the comments are more complicated than that.
A number of people explicitly compare TFD with Warframe, Destiny, Diablo, Vermintide, Darktide, and other games built around repeating activities for long periods of time. Some of them have hundreds or thousands of hours in those games. Their argument is not necessarily that repetition itself is unacceptable.
The distinction I kept seeing was closer to:
Repetition is not automatically the problem. Repetition without enough experiential variation is.
A grind can remain enjoyable when repetition creates mastery, different situations, social interaction, meaningful intermediate goals, or enough moment-to-moment variation that individual runs do not collapse into autopilot.
A bad grind has the opposite effect: the player's main objective becomes reducing the number of times they have to play the activity.
This also helps explain why complaints about “grind” can refer to several different things at once:
- The time cost is too high.
- RNG makes the time cost feel effectively uncapped.
- The repeated activity itself is not enjoyable enough.
- After the grind is finished, the reward does not unlock a sufficiently different reason to play.
Those are different problems even though players may use the same word for all four.
That is why simply improving drop rates can make sentiment noticeably better without necessarily solving the deeper issue. A higher drop rate reduces the amount of repetition required. It does not automatically make the underlying activity something a player wants to return to after the reward has been obtained.
The distinction matters because a tolerable grind and a good grind are not the same thing.
A tolerable grind minimizes how much repetition the player must endure.
A good grind can make the repeated experience enjoyable enough that minimizing it is no longer the player's only goal.
## 2. Acquisition and utilization are two different problems
One of the clearest patterns across the videos is that burnout can happen at two separate stages of progression.
Before completing a build, the player asks questions such as:
- How do I obtain this Descendant or weapon?
- Where do I farm the materials?
- How many Catalysts, rerolls, duplicates or other resources do I need?
- How long will the process take?
- How much RNG stands between me and the thing I want?
Those are acquisition questions.
But after the build is complete, another group of questions starts appearing:
- What do I actually use this for?
- Where does the completed build matter?
- What different experience does this character enable?
- Why should I build another weapon or Descendant if most activities are solved in roughly the same way?
- I finished the farm. What is the next reason to play?
Those are utilization questions.
That distinction is important because a game can make major improvements to acquisition while leaving utilization relatively unchanged.
Recent comments in the corpus do acknowledge that some farming friction, pity systems, quality-of-life features and new-player progression have improved over time. That matters. The research does not support saying that TFD simply “never fixed anything.”
But friction fixes and motivation fixes are different kinds of work.
A friction fix asks:
“How do we make this less frustrating to do?”
A motivation fix asks:
“Why does the player want to do this in the first place?”
Making an item easier to obtain can improve the first question while having little effect on the second.
This is also why I think “TFD needs more endgame” is too vague by itself.
An endgame activity can exist on the menu and still fail to solve the utilization problem if it is primarily a harder version of an existing loop, or if its main purpose is to unlock yet another progression layer.
The important question is not only whether endgame content exists.
It is whether finishing a build changes the kinds of experiences available to the player.
## 3. “Nothing to do” usually does not mean the menu is empty
One of the easiest phrases to misread in the corpus is “there is nothing to do.”
Taken literally, it makes little sense. TFD contains missions, bosses, farming routes, progression systems, weapons, Descendants, seasonal objectives and various things that can still be completed.
But when those comments are read in context, “nothing to do” often means something closer to:
“There is nothing I actively want to do once I no longer need the reward.”
That is a difference between content availability and content desirability.
A player can still have:
- weapons that are not fully upgraded
- Descendants they have not farmed
- medals or titles they have not earned
- records they have not completed
- resources they could continue accumulating
- progression systems that are not technically finished
and still feel that they have “nothing to do” if completing those remaining goals requires repeating activities they no longer enjoy.
Completion percentage is therefore not a complete measure of perceived content depth.
A game can have a very large checklist while still feeling narrow if most of that checklist is completed through the same gameplay language.
This is also how apparently contradictory comments such as “the combat is fun” and “the game is boring” can both be true.
A player can enjoy shooting, movement, grappling, skill effects, animations and character fantasy at the scale of seconds while becoming tired of the same routes, spawns, objectives, optimization logic and reward structure at the scale of hours or months.
Moment-to-moment gameplay and session-to-session structure are different layers.
That distinction becomes especially important in a live-service game because it helps explain why people can leave while still saying that the foundation itself is good.
## 4. More content is not necessarily more kinds of gameplay
Another recurring disagreement is whether TFD actually lacks content.
On one side, the game clearly has more missions, bosses, systems, regions, weapons and Descendants than it did at launch.
On the other side, people continue describing the experience as repetitive.
Those statements do not necessarily contradict each other.
The more useful question may be:
How many genuinely distinct activity identities does TFD have?
Imagine an update that adds a new region, several weapons, another progression system, one boss and a new Descendant.
The quantity of content has objectively increased.
But if all of those additions ultimately feed the same loop:
farm faster
→ build stronger
→ clear faster
→ farm the next thing
then experiential breadth may increase much less than content quantity.
Several comments describe activities with different names, modifiers or rewards as still feeling like variations of familiar patterns: linear speed-running, wave clearing, AoE spam, boss nuking, repeating a room or route, or “old content with something added.”
This is not an argument that those additions have zero value.
It is an argument that content cadence and experience cadence are different.
Content cadence asks how often the game adds new assets, regions, items, characters and systems.
Experience cadence asks how often the player receives a genuinely different kind of activity.
A live-service game can have a high content cadence while still having a relatively low experience cadence.
This also changes how “more content” should be evaluated. Instead of counting only how many things were added, other useful questions are:
- How many new kinds of decisions were introduced?
- How many new forms of cooperation became relevant?
- How many builds gained a new reason to matter?
- How many new failure states or pressures appeared?
- How many reasons to play exist independently of acquiring the new reward?
That is the level where the recurring variety complaints start making more sense.
## 5. TFD may have a collection-expression problem
TFD has a major strength: a broad collection of Descendants, weapons, skills and build fantasies.
But collection creates more long-term value when owning different tools allows the account to solve different kinds of situations.
If most activities primarily reward some combination of:
- movement speed
- AoE
- burst
- boss DPS
then characters designed around:
- healing
- defense
- taunting
- zone control
- stealth
- sustain
- resource support
- team utility
have fewer contexts in which their differences can become important.
This can create a balance problem that is partly external to the character.
On the surface, the community may say a particular Descendant is weak and needs a buff.
Sometimes that may simply be true.
But another possibility is that the game rarely asks the type of question that character is built to answer.
If the solution to every underused character is more damage so that everyone approaches the same clear-speed benchmark, the roster can become numerically stronger while its identities become less distinct.
A different diagnosis is:
Not every weak character needs to become a better answer to the same question. Sometimes the game needs to ask different questions.
This is what I mean by Collection Without Expression.
There is also a useful distinction between two kinds of collection:
Collection as progression: owning another Descendant or weapon means the account contains more things.
Collection as capability: owning another Descendant or weapon means the account can solve more kinds of situation.
If collection mostly expands the first, a large roster can still converge on a small number of meta choices in most activities.
If collection expands the second, an older character or weapon can become relevant again because a new context values what it already does, without requiring a direct numerical buff.
The quantitative layer gives this idea some additional support.
Gameplay-variety complaints and collection-expression complaints co-occurred in 51 top-level comments, about 11.3 times the rate expected if the two tags were fully independent.
Post-build-purpose and collection-expression complaints co-occurred in 32 comments, about 8 times the expected independent rate.
That does not prove causality.
But it is consistent with the hypothesis that narrow activity architecture and narrow roster expression are related.
## 6. Difficulty looks more like an audience-segmentation problem than a single consensus
The corpus does not support a simple conclusion that TFD is either “too easy” or “too hard.”
One group wants deeper mechanics, encounters that survive long enough for mechanics to matter, meaningful team synergy and challenges for optimized builds. Another prefers power fantasy, fast clears, lower mechanical burden and important rewards that are not locked behind highly demanding content.
There is also overlap: many players on both sides dislike “difficulty” created mainly through huge HP pools, one-shots, extreme resistances or mechanics that simply invalidate characters.
So “I want harder content” does not necessarily mean “give enemies more HP,” and “do not nerf my character” does not necessarily mean “I want no challenge.”
Challenge can instead come from positioning, objectives, formations, reinforcement sources, timing decisions, weak points, protection, escort, attrition, battlefield state or multi-stage encounter systems.
The broader implication is that TFD may not need one globally correct difficulty. It may need different activity identities for different motivations.
One activity can serve power fantasy. Another can serve mastery. Another can emphasize adaptation, social play or experimentation.
Not every player needs to enjoy every mode.
Trying to make one universal activity simultaneously fast, difficult, relaxing, rewarding, social, soloable, deep and simple is likely to create permanent balance conflict.
This also complicates the “hardcore vs casual” framing. A player with a few hours per week can return for years, while a very high-hour player can consume an update in days. Hours per week and long-term loyalty are different variables.
## 7. Co-op has at least two separate structural problems
Co-op is not one of the highest-volume themes in the corpus, but the comments that discuss it are unusually specific.
The first issue is co-op progression alignment.
Personal RNG can cause friends who begin farming together to drift into different objectives.
One player gets the desired drop.
Another still needs the Amorphous Material.
A third has to return to an earlier material farm.
They may have similar account strength and still no longer share the same reason to run the same content.
That creates a useful distinction:
Co-op compatibility is not the same as co-op progression alignment.
A game may technically allow four people to play together while its progression systems repeatedly pull those four people toward different tasks.
The second issue is co-op agency.
When one optimized Descendant deletes the room or boss almost immediately, the other three players can technically be present without meaningfully participating.
They run behind, watch, collect the result or become functionally passive.
That creates another distinction:
Matchmaking is not the same as meaningful co-op.
Simply adding a six-player activity, raid, or larger party does not automatically solve either problem.
The group still needs aligned reasons to be there, and individual players still need meaningful ways to participate.
The corpus also contains a smaller but interesting social signal: some veterans describe helping newer players, encountering random players or playing with friends as independent reasons to log in.
That matters because social motivation does not have to be the most efficient farm.
A veteran may need nothing from an old activity and still join because they can help somebody, save a failing run, test a build with other people or simply spend time with friends.
This is one reason some older or less efficient systems can still be remembered positively: they sometimes created more natural social texture.
Efficiency and social health do not always point in the same direction.
## 8. Progression bloat becomes an investment-trust and returnability problem
Progression bloat is relatively weak in the launch-era data and becomes more visible later, which is logical: launch players cannot complain about years of accumulated systems that do not exist yet.
Later comments increasingly mention secondary progression layers, heavy build management, balance changes, power creep, completed investments becoming outdated, returning players not knowing what to prioritize, and new friends struggling to catch up.
One veteran describes maintaining builds after repeated meta shifts as something like “spreadsheet repair.”
This is not only a complexity complaint. It is also an investment-trust complaint.
Players invest farming time, materials, Catalysts and rolls, but also rotations, muscle memory, playstyles, familiarity and attachment to a character.
If that value repeatedly shifts because of balance churn or power creep, future investment starts to feel less durable. Even a buff or rework can damage trust if it removes the behavior a player originally enjoyed.
The issue is not that nothing should ever change. It is whether earned value and established playstyles remain recognizable often enough that building an account feels trustworthy.
This connects to Returnability Debt.
Many players appear to cycle through:
update → return → play → complete goals → leave → wait
That can be normal for a live-service game.
The problem is when every absence becomes more expensive because returning means learning new systems, repairing builds, understanding balance changes and figuring out what is still relevant.
A player can still want to return while the cost of returning keeps increasing.
A healthy live-service relationship does not necessarily require permanent daily retention. It may be equally important that someone who leaves for several months can come back, understand their account quickly, trust that old investments still matter and begin meaningful play without turning the return into a restoration project.
## 9. A lot of disengagement looks like dormancy rather than rejection
This was one of the more surprising qualitative patterns.
Negative comments often still contain praise for:
- combat
- gunplay
- movement and grappling
- character designs
- visual quality
- buildcraft
- the underlying foundation
That means “I stopped playing” and “I dislike the game” are not equivalent statements.
In the quantitative layer, 282 top-level comments were coded as clearly positive about the core game — combat, gameplay, characters or the foundation.
Among those comments, 45.7% were also tagged for repetition.
That is one of the strongest results in the entire analysis because it changes the shape of the criticism.
It is less like:
“I hate the gameplay.”
and more like:
“I like the gameplay, but I am being asked to use it inside too few kinds of experience.”
A broader “dormant affection” group combined positive-core comments with comments showing attachment, hope or intention to return.
That produced a base of 581 comments.
Within that group:
- 40.4% also mention repetition
- 11.9% mention lack of variety
- 8.8% mention post-build purpose
- 8.4% mention quit/return behavior
Again, these are not population estimates.
But within this corpus they support the idea that TFD's attraction remains alive among a meaningful number of people who have reduced or stopped playing.
This is why I describe the game as being stronger at creating attraction than converting that attraction into durable habit.
An update can therefore produce a large return spike without necessarily proving that retention has been repaired.
A possible cycle is:
new update
→ curiosity
→ player returns
→ farms the new Descendant or weapon
→ tries the new activity
→ acquisition ends
→ reason to keep playing falls
→ player becomes dormant again
Peak concurrent users after an update can show renewed interest.
They cannot by themselves tell us whether the new activity created a lasting independent reason to play after acquisition ended.
## 10. Some visible complaints are real, but they are not equally strong churn signals
Monetization is prominent at launch: prices, bundles, paint, crafting timers, pay-to-skip systems, RNG, shop value and comparisons with Warframe.
Later discussion shifts somewhat toward a different criticism: why does cosmetic or character production sometimes feel more reliable than the arrival of new ways to meaningfully use those characters?
At the same time, many players genuinely like TFD's attractive or sexualized character designs and consider them part of the game's appeal.
So the corpus does not support “sexy skins are the main reason players left.” A more defensible interpretation is that character appeal can be a strong acquisition and monetization pillar, but it cannot substitute for retention gameplay.
Story concerns are also real — the main narrative is often described as generic or weakly presented — but they do not appear stronger than repetition, variety, grind or post-build purpose as churn signals. Descendant-specific stories are sometimes viewed more positively, suggesting character attachment may be a more natural narrative strength.
“Dead game” and “dying” language is even noisier. People use it for Steam population, clickbait, personal disengagement or general frustration, while others reject the label or continue playing.
For that reason, I do not treat “dead game” as a cause. It is a judgment or outcome label, not an explanation of the mechanism behind disengagement.
## 11. The timeline shows the problem changing as the player relationship matures
The timeline is important because launch-era discussion and later discussion are not asking the same questions.
The broad cohorts in the dataset are:
Launch / early (~2 years), n = 4,306 top-level comments
- Repetition: 9.8%
- Variety: 1.2%
- Post-build purpose: 1.1%
- Investment trust: 0.6%
- Progression bloat: 0.3%
- Quit/return: 0.7%
- Onboarding: 1.2%
~1 year, n = 2,438
- Repetition: 26.6%
- Variety: 5.5%
- Post-build purpose: 5.1%
- Investment trust: 7.0%
- Progression bloat: 1.3%
- Quit/return: 5.8%
- Onboarding: 2.1%
8–11 months, n = 1,042
- Repetition: 16.5%
- Variety: 4.0%
- Post-build purpose: 3.1%
- Investment trust: 2.6%
- Progression bloat: 1.3%
- Quit/return: 3.1%
- Onboarding: 3.5%
4–7 months, n = 1,151
- Repetition: 19.3%
- Variety: 4.3%
- Post-build purpose: 3.6%
- Investment trust: 3.4%
- Progression bloat: 2.1%
- Quit/return: 4.9%
- Onboarding: 2.9%
<4 months, n = 404
- Repetition: 18.6%
- Variety: 3.5%
- Post-build purpose: 3.2%
- Investment trust: 4.7%
- Progression bloat: 2.0%
- Quit/return: 5.9%
- Onboarding: 4.0%
These should be treated as broad phases, not precise month-by-month telemetry. The downloader preserved relative comment-age strings rather than exact publication timestamps for every comment.
The ~1-year cohort also contains several videos explicitly framed around quitting or players leaving, so its high values should not be interpreted as proof that the game was objectively “worst” at exactly that moment.
The directional pattern is more useful than treating the table as population telemetry.
Early players mainly evaluate the immediate value proposition:
- Is the farm worth it?
- Is monetization reasonable?
- Is RNG too punishing?
- Is the combat good?
- Is the game polished enough?
- How does this compare with Warframe?
Players with more account maturity can evaluate a different relationship:
- Is my investment durable?
- Does this completed build have a purpose?
- Is the new activity genuinely different?
- Are old characters still useful?
- Will another progression layer make my account harder to maintain?
- If I take a break, how much work will it take to return?
A simple version of that lifecycle is:
Stage 1: “Is this farm worth it?”
Stage 2: “I finished the farm. Now what?”
Stage 3: “I took a break. Why do I need to relearn or repair so much when I return?”
That is why I do not think the story is simply “TFD failed to fix its problems.”
Some acquisition friction has genuinely improved.
The problem appears to have partially shifted from entry friction toward structural motivation, investment trust and returnability.
## 12. Could this just be one creator, repeat commenters, or launch-era bias?
This was important to test because the dataset is not a representative survey.
First, 93.3% of unique top-level authors appear in only one video. Only about 6.7% appear in more than one. Cross-video recurrence is therefore mostly coming from different commenter pools rather than a small group posting everywhere.
Second, the two largest launch-era videos account for approximately 47.1% of all top-level comments. I removed both and recalculated several structural themes.
They became more prominent:
- Repetition: 16.5% → 22.2%
- Gameplay variety: 3.1% → 4.9%
- Post-build purpose: 2.8% → 4.3%
- Investment trust: 3.0% → 5.2%
- Quit/return: 3.1% → 5.0%
- Collection expression: 1.5% → 2.4%
- Progression bloat: 1.0% → 1.6%
So the broad launch audience does not appear to be creating the structural diagnosis. If anything, it dilutes it with pricing, monetization, first impressions and acquisition concerns.
Third, I removed each video one at a time. The repetition share remained between approximately 13.7% and 19.3%, so no single video drives the main result.
Using a minimum of at least three top-level comments in a video before counting that video as containing a signal, repetition reaches 13/13 videos. Gameplay variety, quit/return, investment trust, post-build purpose, onboarding and collection expression each reach 12/13.
The Korean-language comments also contain the repetition/grind signal. That is not enough for a serious cross-cultural comparison, but it is a useful sanity check that the dominant complaint is not entirely dependent on one English-speaking creator ecosystem.
Likes also need caution. Only 302 of 9,341 top-level comments received at least 20 likes — about 3.2% — yet they account for roughly 88.2% of all likes.
I therefore treat likes as resonance, not prevalence. A highly liked comment may strongly connect with readers, but its like count is not the number of independent people who would spontaneously name the same issue.
## 13. Why some small percentages can still matter
Gameplay variety is 3.1%, quit/return 3.1%, investment trust 3.0%, post-build purpose 2.8% and collection expression 1.5%.
Those can look small if read like fixed-choice survey results, but this is an open-ended corpus. People are not answering the same question from the same list of options: some write essays, some only say “I quit,” some joke, argue with the creator, or discuss unrelated issues.
That is why I put more weight on the combination of frequency + cross-video recurrence + different author pools + timeline consistency + robustness tests + co-occurrence, rather than treating one pooled percentage as a population estimate.
The study is much better at answering “Which patterns keep reappearing in these discussions?” than “Exactly what percentage of all TFD players believes this?”
## 14. A deeper structural model: rewards may be carrying too much of the engagement burden
After connecting the major themes, one broader hypothesis kept appearing.
TFD is a looter-shooter, so rewards and progression obviously need to matter.
The issue is not that the game should somehow stop having loot or grind.
The question is how much motivational work the reward is doing compared with the activity itself.
In a healthier reward relationship:
reward attracts the player to the activity
→ gameplay gives the player a reason to keep enjoying the activity
→ the activity can retain some value even after the reward is no longer optimal
In a highly reward-dependent relationship:
reward disappears
→ much of the reason to play the activity disappears with it
Many veteran comments describe some version of the second pattern:
- the weapon is already obtained
- the character is already maxed
- the build is complete
- the records are done
- there is no reason to return to that map or boss
That makes every activity fragile.
Developers must continuously create “the next thing to obtain” because progression is carrying not only reward, but also content, motivation, novelty and endgame purpose.
There is another possible reinforcing loop here.
If vertical progression makes players increasingly powerful, normal enemies can become less capable of engaging them.
If gameplay resistance decreases, progression and loot have to carry more of the engagement burden.
More progression is then added to provide new goals.
Players become even stronger.
The underlying combat becomes even easier.
Another progression layer is needed.
That does not mean vertical progression is bad. It is fundamental to the genre.
The risk appears when progression is expected to function as reward + content + motivation + endgame + novelty all at once.
This is why some comments asking for “for-fun content” are interesting.
They are not asking for a better reward.
They are asking for an activity they would still want to play even when it is not the most efficient farm.
That is a different kind of retention value.
## 15. Local fixes can make sense while still producing a global problem
Community feedback is very useful for locating pain points, but pain points and solutions are not the same thing.
Individually reasonable responses might look like:
grind hurts → increase drop rates
bosses fail too often → reduce difficulty
a character is underused → buff it
players dislike nerfs → buff alternatives
progression feels slow → increase rewards
mechanics slow farming → simplify them
Each can improve an immediate problem.
But accumulated over time, local optimizations can potentially produce stronger players, less meaningful enemies, faster content consumption, more progression systems, shifting builds, less distinct roles and harder catch-up.
This is also why apparent contradictions should not automatically be dismissed as “the community doesn't know what it wants.”
“The game is too easy” + “do not nerf my character” can both be true if the desired challenge is able to absorb earned power rather than remove it.
“The grind is too long” + “the game runs out of content too quickly” can both be true if acquisition is stretched while experiential breadth remains narrow.
“Attractive skins are why I play” + “the game focuses too much on skins” can both be true if acquisition appeal and retention are different concerns.
A useful framework is:
Symptom — “boring,” “nothing to do,” “too easy,” “too hard”
Immediate cause — same mission repeatedly, one-shots, huge HP, obsolete build
Structural cause — narrow activity variety, reward dependency, progression bloat, low role expression, returnability debt
Proposed solution — raid, PvP, buff, nerf, higher drop rates, guilds
The mistake is jumping directly from symptom to proposed solution. This research tries to stay longer in the middle: identify the immediate and structural mechanisms before deciding what feature should be built.
## 16. What the data supports strongly — and what it does not
After the qualitative and quantitative layers, I would separate the conclusions by confidence rather than pretending every idea has equal support.
Very high confidence within this corpus:
- Repetition/grind lacks sufficient experiential variation.
- The repetition complaint recurs across many different videos and audience pools.
- Likes should not be treated as prevalence.
- Launch-era audiences are not the sole source of the structural complaints.
High confidence:
- Lack of post-build purpose is a recurring issue.
- Lack of gameplay variety is a recurring issue.
- Investment trust and power creep become more visible lifecycle concerns.
- Dormant affection and quit/return behavior are real recurring patterns.
Medium-high confidence as broader diagnoses:
- Returnability Debt.
- Collection Without Expression.
- Progression Substituting for Gameplay.
- TFD lacking enough horizontal reasons to play.
Lower-confidence hypothesis:
- Old-world depopulation/content ecology being a major retention driver.
There are comments suggesting that concentrating rewards in new activities moves veterans away from older areas, which can make those spaces quieter for newcomers and weaken social continuity.
It is plausible.
The recurrence is not strong enough for me to present it as a major established pattern.
And several claims are not supported strongly enough by this corpus:
- “The community wants PvP.”
- “Raids are the solution.”
- “Guilds are the solution.”
- “Sexualized skins are the main reason players leave.”
- “One specific balance change caused the population decline.”
- “One creator represents the community.”
- “Horizontal modes would automatically solve retention.”
- “A highly liked proposed feature should therefore be built.”
The comments contain proposals for PvP, PvEvP, raids, guilds, vehicles, roguelike systems, social hubs and other ideas.
Those are useful solution proposals.
They are not evidence of consensus on the treatment.
Consensus on a problem does not imply consensus on a solution.
## 17. What this research cannot establish
This is still a self-selected YouTube comment corpus. Several videos are explicitly about quitting, the state of the game or dissatisfaction, which naturally attracts a different audience from a build guide or positive showcase.
Active Reddit/Discord polls have the opposite problem: survivorship bias. People still active are more likely to have accepted or enjoyed the current loop.
Neither source is “the truth.” They answer different questions.
Active-community data is useful for asking:
“What do the people still here want the game to preserve?”
Lapsed-player data is useful for asking:
“What was missing for people who left or reduced their engagement?”
The quantitative tagging here is heuristic rather than a professionally validated supervised classification model. Comments can belong to multiple themes, and the percentages should be read as indicators within this corpus rather than population estimates.
Association is also not causality. For example, the 11.3x co-occurrence between gameplay variety and collection expression supports a hypothesis, but cannot prove that one causes the other.
Questions about actual churn causality require data the community does not have: retention cohorts, session length, activity completion, return rates, progression state before churn, friend/party behavior and player-level longitudinal telemetry.
So the safest rule is:
Treat strong findings as patterns. Treat causal chains as hypotheses. Treat solutions as proposals that still need testing.
## Overall conclusion
After all three analytical layers, I do not think the most accurate description is:
“TFD failed because of grind.”
“TFD lacks content.”
“TFD is too easy.”
“TFD is too hard.”
“TFD only sells skins.”
A more complete interpretation is that TFD has a combat-and-character platform attractive enough that many players still want to like the game, but its activity and progression ecosystem has not consistently converted that attraction into a stable long-term relationship across enough player types.
The game is very good at asking:
“What item am I still missing?”
“Which Descendant do I not own yet?”
“Which system have I not maxed?”
It is less consistent at asking:
“What kind of experience do I want to play tonight?”
“What different problem does this build let me solve?”
“Can I play with a friend for fun without being on the exact same farming step?”
“If I already own everything I care about, what still makes me want to log in?”
“If I leave for six months, can I come back without doing homework?”
That is why I think “more reasons to play” is a stronger diagnosis than simply “more content.”
A wider ecosystem could allow different motivations to exist independently:
want an item
→ farm
want to test a build
→ challenge
want co-op
→ team activity
want power fantasy
→ large-scale combat
want adaptation
→ variable activity
want to relax
→ social/casual play
want mastery
→ endurance or high-difficulty play
want collection expression
→ situations where different kits have meaningfully different value
Rewards do not need to disappear.
Grind does not need to disappear.
Buildcraft does not need to disappear.
The point is to place them inside a wider ecosystem where progression serves gameplay instead of having to substitute for gameplay.
So the shortest version of my conclusion remains:
TFD does not appear to lack “answers.”
It appears to lack enough different “questions” for those answers to continue meaning different things.
## Methodology and raw data
The corpus contains 15,649 comments/replies, 9,341 top-level comments, 8,184 unique top-level authors and 13 YouTube videos/Shorts spanning discussion from launch through 2026.
Top-level comments were the main quantitative unit because reply chains can inflate one original viewpoint. Replies still mattered qualitatively for nuance, disagreement and counterarguments.
The analysis used three broad layers:
- Identify recurring complaints, defenses, contradictions and viewpoints across videos.
- Connect surface themes into possible structural mechanisms.
- Quantitatively test them through theme coding, author checks, cross-video recurrence, broad time cohorts, robustness tests and selected co-occurrence.
The videos were selected because they focus on TFD's state, why people left or reduced playtime, changes over its lifecycle, and debates about grind, content, balance, direction and long-term potential. The videos themselves are not treated as direct representations of “community opinion”; their comment sections are the corpus being analyzed.
Raw datasets and source-video information:
https://drive.google.com/drive/folders/15xH7823PSA4ZBrj0SBO7kpu_l4Rlocts?usp=sharing
If you read all the way through, I am especially interested in where your experience doesn't match these patterns.
If you stopped playing and later returned, what actually brought you back — and what made you stay or leave again?