Human Fall Flat

First Impressions

Start with the numbers. The Steam store page reads Overwhelmingly Positive: 95% of 30,811 reviews. The last thirty days alone brought 951 more, 93% of them positive. A 2016 game is still collecting reviews at that rate a decade later.

Then you open the ten most helpful positive reviews and stop short. Seven of the ten say almost the same thing: give us a level editor, add Steam Workshop support. They were posted between October 2017 and September 2018.

The ten most helpful negative reviews are just as uniform. Six of them are about Linux support being dropped. Not about the puzzles. Not about the game at all.

So across those twenty reviews, barely a sentence scores the puzzle design. What, then, is the 95% a verdict on? That is the question I want to follow through this pool.

Human Fall Flat — Steam store screenshot #1

Putting the Mechanics into Words

There is very little you can do. Walk, jump, and reach out with each arm separately to grab. The arms follow where you look, so turning the camera is itself a way of applying force. One top Japanese review calls it a strange machine where the camera generates the power.

The positive side reaches for hilarious, funny, physics. The negative side reaches for imprecise, floppy, and for the idea that the physics stops being an ally and becomes an opponent. One negative review from November 2017, with 128 helpful votes, describes fun curdling into irritation.

What interests me is that both are pointing at the same design. One verb — grab — covers pushing, pulling, climbing, throwing and dropping. The verb count has been cut to the bone, exactly the subtraction I wrote about in the case for fewer verbs.

But what was cut is the number of verbs, not the precision of the input. In Sokoban a tile is a tile. Here the same grab lands slightly differently every time. The difficulty of this game sits in the body, not on the board.

Human Fall Flat — Steam store screenshot #3

What Was Not Subtracted: the Width of the Solution

The store description states plainly that its puzzles have more than one solution, and the review pool bears that out. Positive reviewers get past the same obstacle in visibly different ways.

This, I think, is why the reviews do not talk about puzzles. When a board has one intended answer, you can say "here is how I solved it." When the answer is wide, you cannot — you cannot even tell yourself whether your route was the solution or an accident.

So what gets told is not a solution but an incident. One top Japanese review jokes about workplace-accident footage. A review section full of pratfalls is not laziness on the writers' part. It is what the design produces.

I read something similar in the Poly Bridge pool. Where physics rules, players struggle to own their answers. It also rhymes with the structure I described in the grammar of solving together.

Human Fall Flat — Steam store screenshot #5

A Request Frozen at the Top

The Workshop support those top positive reviews kept asking for did arrive. It was reported in April 2019, with the developer publishing two example levels to get creators started. The store page now advertises more than five thousand community-made levels.

Eight-player online co-op was not there at launch either; it was added in October 2017. The game became a different pastime partway through its life. And yet the top of the review section is still held by posts from 2017 and 2018, because helpful votes accumulate on old reviews.

There is one aggregate box. The votes stacked in it were cast on at least three different games: the 2016 solo game, the eight-player game from late 2017, and the Workshop-equipped game from 2019 onward.

The negative side has the same shape. The end of Linux support was confirmed by the developer on the store forum on 29 October 2018, and a matter that is not a design judgement has held the top slots for nearly eight years. A review section is not a mirror of a game's present. It is sediment. In Mini Metro I noted that the aggregate has no box for "mixed"; what shows here is the same gap along the time axis.

Human Fall Flat — Steam store screenshot #9

Sources

This piece was written from the Steam store page and the Steam Community review pool as they stood on 27 August 2026.

・Steam: Human Fall Flat (Overwhelmingly Positive; 95% of 30,811 reviews. Last thirty days: 93% of 951.)

・SteamDB (roughly 226,000 reviews across all languages, 93.5% positive) and Metacritic (PC: 70 from 22 critics; user score 7.2 from 359 ratings).

・Read in full: the ten most helpful positive reviews, the ten most helpful negative ones, the eight most recent, and the ten top-rated Japanese-language reviews, plus a store forum thread about Linux support.

・Background facts — studio, post-launch additions, sales, sequel announcement — were checked against the Wikipedia article and the April 2019 Destructoid report on Workshop support.

・The images are screenshots published on the Steam store page. This session could not reach the image CDN directly, so their contents could not be inspected; the captions therefore give only the store ordering rather than describing what is shown.

Closing

Metacritic's PC page reads 70 from 22 critics, with a user score of 7.2 from 359 ratings. That is a long way from Steam's 95%. The critics scored a puzzle game; the Steam voters scored an evening with friends. Read that way, the gap is not a contradiction.

From a design standpoint I give it 7.0. Cutting down to a single verb was the right call. But by leaving the solution wide without tightening the input, the game gives up the feeling of having solved anything. That is a trade, not a fault: it hands back ownership and receives shareable accidents instead.

It suits people who want to laugh at failure on a shared screen. It does not suit people buying hours alone with a board. Hours shown on the top helpful reviews range from 7 to 51, but the complaints about solo play cluster on two words: monotonous and short.

What stayed with me after reading was not a verdict on the game but a doubt about aggregates. A game can keep moving for ten years; each buyer still votes only once. The argument in when the control scheme decides the difficulty folds back here onto the time axis.

Human Fall Flat — Steam store screenshot #14