The Last of Us Tops the 100 Greatest Games List: The Real Dispute Is the Missing Rubric

Twelve hours after Kinda Funny Games published its list of the 100 greatest games of all time, the comment section split into three groups. The first defended The Last of Us at number one. The second argued that The Last of Us Part II deserved the slot, because its combat systems and structural ambition run deeper than the 2026 original. The third dragged Sekiro into the argument, and that group was the loudest.

What is worth recording: all three groups agree the game is good. Nobody disputes quality. What they dispute is a document that does not exist — the rubric that placed one hundred titles in that particular order.

I read the list three times. The first pass, looking for data. The second, looking for methodology. The third, looking for anyone at the outlet explaining why the other ninety-nine titles sit where they sit. All three passes came back empty.

The issue is not the number one slot. The issue is the ruler.

Context: a ranking built like a match

Kinda Funny Games is a creator-led games outlet operating inside Western commentary culture. It published a hundred-title ranking with The Last of Us — developed by Naughty Dog, released in 2026 — at the top.

The construction of the piece matters more than its content. The sequence runs: open with controversy, then reveal the ranking, then quote both camps, then close by relativising the whole exercise with the observation that choosing the greatest game of all time has never been easy. Inverting cause and effect generates reading tension, and reading tension is the product being sold.

Naughty Dog appears as developer, behind the Uncharted line and then The Last of Us. That is a canon built by a single studio, and the lineage is itself a statement about criteria: when one studio defines the quality standard of a decade, a list built by critics tends to reproduce that standard.

The piece carries no sales figures, no completion rates, no player counts. The only quantitative element in the entire text is a rank position: one out of one hundred.

It also records, verbatim, the question the community itself asked — does this list evaluate by gameplay quality, by popularity, or by historical influence — and then leaves it hanging. That is the most important detail in the article.

And it builds a rhetorical ladder: from simple products such as Pong to modern games such as Clair Obscur: Expedition 33. A ladder running from 2026 to 2026, with the newest release placed at the upper end.

I cover esports for the Korean market from Seoul. In this job, arguments end at a scoreboard, and a scoreboard does not negotiate. But even there I have seen enough rankings assembled by feel to know one thing: when the ruler is not printed, the rank becomes an assertion with a number attached.

A ranking is a match with no referee

In the LCK, the standings can be recomputed by hand. Match wins, game differential, head-to-head, then tiebreakers. Anyone with a spreadsheet can reproduce the order. In the VCT, circuit points work on the same principle. The value of a sports table is that it does not need to be believed — it needs to be recomputable.

A cultural ranking has no scoreboard. The rubric therefore has to carry the load the scoreboard carries in competition. When the rubric is absent, the ranking loses the right to be recomputed and retains only the right to be trusted or rejected.

The smallest detail on the field usually says the largest thing. Here, the smallest detail is a missing methodology line.

This is not a rare technical defect. It is the genre default. Rankings are the cheapest format to produce and the most expensive in engagement, and their value sits in the comment section rather than in the ordering. A published rubric would cut dispute volume to near zero. The ambiguity is functional, not accidental.

Four axes with no common unit

The piece lists the criteria readers are actually applying: gameplay quality, popularity, historical influence. It also quotes one side arguing that emotional storytelling, character building and unforgettable moments are the real measure.

Together that is four axes in four incommensurable units. Gameplay is measured by systems depth. Story by emotional residue. Popularity by headcount. Influence by historical lag. Adding them is adding metres to kilograms.

In statistics, composite indices are routine — but they always arrive with two mandatory components: declared weights and a described normalisation. Without those, a composite index is just a preference ordering dressed as a measurement.

A ranking that does not publish its weights is a personal preference presented as a measurement.

And the deeper problem: when weights are withheld, the burden of proof shifts entirely onto the reader. Defenders must guess which criterion they are defending. Critics must guess which criterion has been violated. Two camps argue inside two different frameworks and never touch. That is why this dispute cannot end, and why it lasts exactly as long as the article's traffic cycle.

Canon lag and the endpoint trap

The ladder from Pong to Clair Obscur: Expedition 33 is where I stopped longest.

It compresses more than fifty years of industry history onto one axis, then places a title released in the same period as the ranking at the modern end of that axis. Methodologically, this is an endpoint error: works that have survived decades of scrutiny are being compared against a work that has had no time to accumulate consensus.

Reception variance for a new title peaks at launch. It decays as discussion settles and only the surviving assessments remain. Canon needs lag — usually a decade — before a work can sit on the same scale as stabilised works.

The gap in reception variance between the two ends of the time axis is too wide to be placed on a single scale.

That is why I do not use lists like this as reference data. Not because they are wrong. Because they have not had time to be either.

Based on my experience tracking matches during my years in data analysis in Seoul, I learned a fairly harsh rule: before collecting anything, fix the observation window. In 2026, when the Korean league returned behind closed doors, I collected the first forty-two matches and found the home win rate had fallen to roughly twenty-five percent, against the pre-pandemic level. But to say that sentence I first had to define what counts as a match, what counts as home, and which window counts. Without those three definitions, any rate can be manufactured to say anything.

A hundred-title ranking has no observation window, no variable definitions, and no weights. It has a rank.

Safe versus ambitious: the only frame that outlives the argument

The one thing in the piece with a longer shelf life than the ranking is the pairing of 2026 against 2026.

The first The Last of Us was a refinement: it took the Uncharted template, tightened the pacing, cut the spectacle, and turned restraint into style. The Last of Us Part II was a gamble: non-linear structure, inverted character roles, expanded combat systems, and an acceptance that part of the audience would react violently.

That pairing transfers directly into esports without a single edit.

A team playing a proven system carries low variance and a known ceiling. A team playing experimental carries a higher ceiling and higher variance. Across a long series, variance does not add — it multiplies by game count. That is why champions are rarely the teams with the boldest ideas, but the teams that can repeat their bold ideas three times in one week.

Faker is the canonical case. Five world titles across 2026, 2026, 2026, 2026 and 2026 were not built by wholesale reinvention each season. They were built on a stable base with reinvention at the edges. When the base broke, the team collapsed. When the edges went stale, the team got read. Both failures happened, and both took multiple seasons to repair.

Safety does not oppose ambition; it is the condition that makes ambition repeatable.

The Part II camp is right about art and wrong about structure. They say the 2026 game is more ambitious, and it is. But an ambitious work does not automatically beat a restrained one, because the two are measured on different scales — and the ranking never says which scale it is using.

The consensus machine and the quantifier shift

One line in the piece deserves to be read on its own: even those who object concede that The Last of Us is one of the most outstanding games in history.

Read the quantifier. One of. That is a claim about membership in a set. It says: this work belongs among the best. But the ranking asserts something else entirely: this is the maximal element of the whole set.

Conceding the weaker claim to manufacture the feeling of consensus around the stronger one is a measurable rhetorical move, and it works because it lowers the evidentiary threshold to exactly the level of the available data.

This is a quantifier shift: from set membership to maximum rank, executed without a single additional piece of evidence.

In sport, competition rules block that shift. A top-four team is not the champion. Here, nothing blocks it, so it passes unopposed.

Evidence with no sample

Every description of community reaction in the piece is built from uncountable phrases: one gamer said, some players argue, many people feel.

No counts. No platform. No timeframe. No sampling method.

Based on my experience tracking matches and argument spikes, a voice only becomes a sample when you know how many voices there are and where they were counted. Before that, it is a quotation, and quotations have illustrative value, not statistical value.

This matters because it changes how the whole piece should be read. If the controversy was not measured, it was not observed — it was produced. And a produced controversy cannot be used as a sentiment indicator, nor as input to any model.

Mislabel and real damage

The piece was filed under esports.

Across its main body there is no team, no player, no patch, no tournament, no organisation. Only the appended headline block touches esports at all.

To a statistician, a mislabel is not a small error. A mislabelled row does not merely contribute zero — it biases the aggregate. Entity counts skew, topic distributions pull, sentiment baselines blur. If an esports monitoring system ingests this under an esports label, it records a coverage signal that does not exist.

The fix is simple: a pre-filter requiring at least one esports entity as subject matter — a competitive title, a team, a player, or an organiser. This piece fails that filter.

Appendix: a lead list, not evidence

The related headlines at the end carry a few notable fragments. One references LCK viewership doubling after six years of franchising. Another names GAM Esports as a VCT Pacific 2027 partner. There is VCT Champions 2026. There is MLBB Vietnam and its potential for global competitiveness. There is K/DA and a controversy over a new appearance. There is a sponsor-named VISA event.

None carries figures, measurement windows, or sourcing. They are leads for verification, not evidence.

The structure of that appendix is worth reading, though. A general games outlet maintaining a standing esports vertical across multiple titles means esports has become a stable editorial category rather than a seasonal topic. That is the only genuine transmission signal in the whole piece, and it sits at the edge.

The real value of the appendix is structural, not editorial: a general games outlet keeping a standing esports vertical means esports has become a stable editorial category rather than a seasonal topic.

I do not listen to the crowd; I read the rubric. In this case, the rubric was never printed.

Where I could be wrong

One possibility I have to put on the table myself: the absence of a rubric may be the correct choice, not a defect.

Canon is not a measurement. It is a social process. A published weighting framework would freeze that process, turning a live argument into a dead spreadsheet. Lists like this exist to keep the conversation going, and ambiguity is the fuel. If I force it into a composite index, I may kill the exact thing that makes it useful.

Second possibility: I am applying competition logic to a product with no win condition. The Last of Us has no scoreline, no overtime, no season table. Demanding a rubric for a narrative work is demanding a referee for a concert. My question may be the wrong question, not their answer the wrong answer.

Third possibility, and the one I like least: the number one slot may be correct, and my entire objection may be procedural pedantry. If so, I am using the language of method to disguise an aesthetic disagreement, and that is an occupational failure.

My condition for being fully wrong: Kinda Funny Games publishes a reproducible weighting framework, applies it across all one hundred titles, and the order barely moves. At that point the ranking shifts from assertion to instrument, and every procedural objection I have made loses its value.

People say I object to get attention; I simply look one step ahead. The step is this: when the argument fades, what remains is not the number one slot. What remains is an unanswered question about which scale to use.

What to watch next

Over the next twelve months, if the outlet publishes a weighting framework, dispute volume around the top slot will drop sharply and the next list will outlive a single news cycle. If it does not, the next list will generate equivalent dispute with a shorter lifespan — because audiences are learning to doubt the ruler before learning to read it.

On the esports side, the testable criterion is whether the 2026 season produces independent LCK viewership figures published with a stated measurement window and with platform migration excluded. If so, the franchising argument gains ground. If not, it remains a headline.

If you are right before the moment, you are called a lunatic. If you are right after, you are a genius. This ranking is not on either side yet — it is standing in the safest place available, which is the place where the ruler never has to be published.