...but I wouldn't blame anyone whose only exposure to his works is Metal Gear Solid 3 for thinking otherwise.
(Before you ask, this article is written about Metal Gear Solid 3, and not Metal Gear Solid Delta. If you think a remake can take the place of an original, or even eclipse it with the endless march of technological progress, please do not participate in any conversations involving culture, ever. Also, if you're playing any of the re-releases of the game, be sure to set the camera to the original Snake Eater style overhead camera and not the third-person camera, which makes an already flawed game even worse.)
By now, the story around Metal Gear Solid 3 is well known -- Kojima doesn't want to make more Metal Gear Solid sequels, having attempted to kill the series with Metal Gear Solid 2, but his overlords at Konami force him to. He trudges onwards, and despite himself, creates one of the best games of a console generation. Like most stories, there is a kernel of truth here. It should be (this sentence originally read "is", but the intelligence of the average gamer forced revisions) very obvious to anyone that plays Metal Gear Solid 3 that Kojima did not want to make it. The rest of this story is a lot more dubious in the face of the evidence.
Metal Gear Solid 3 is ultimately an ugly patchwork of unrelated scenes and mechanics, with deep flaws remaining unfixed but simply band-aided over because the director didn't even want to make the game he was forced into directing, and certainly wasn't going to go to the effort to rethink basic assumptions of content he resented from the start. The most obvious of these is with its approach to resources and "survival". Becoming hungry and running out of food is a condition that creates a feedback loop -- because your character is less capable when hungry, this also makes you less able to get more food, which makes the hunger situation worse. Heavy injuries require limited resources to heal, and if you can't heal them, your life will drain out until you die, so if you run out of these resources, you're just screwed. Unlike the survival horror games I've been writing so much about recently, Metal Gear Solid 3 came out in that unfortunate era when developers were seemingly banned from considering their meta-mecahnics, when "save anytime" and checkpoints on every screen became pre-requisite for the commercial success of any title before Demon's Souls reminded everyone that developers, in fact, are allowed to think of the implications of save/reload/checkpoint/game over, and as such, there simply isn't any room to allow for "dead end" gamestates or encourage strategic use of saves. Unlike Metal Gear Solid 1, there is no retreating to previous areas for more resources; items missed are missed permanently. As such, softlocks were clearly a major concern for Konami -- they knew the game's design inherently had them, had feedback loops that made them worse, and that its design fundamentally could not work with them, and so instead of fixing the core problems, they deployed two different band-aids.
The first of these is at least clever in a classically Kojima-esque sense. If you quit the game for a while, when you come back to it, your character will have taken a nap and have his energy/hunger bar full, his life full, and all of his injuries healed. While it's unclear exactly how Snake takes a nap in the middle of a pistol duel with Ocelot, leading to all sorts of silly ludonarrative dissonance in a game that generally takes its ludonarrative aspects extremely seriously, it's at least a nice idea to get a player to "come back fresh". It doesn't change the fact that this is an ugly kludge to fix fundamental issues with the game, it doesn't come anywhere close to actually fixing the game's horribly broken core, but it's at least clever.
The second of these is appalling. If you die on any given screen enough times, the game just fills up your health and energy and makes you completely invincible for the duration of that screen. This is the ultimate admission that this is a videogame that has nothing to teach the player about how to play it in most of its scenes; that skipping a scene now won't leave you underprepared for something later. It is, to quote a great article by Danbo, palliative care for a terminally bad video game. What Kojima is saying about his game by including this is that there's nothing of value for us to take from one scene to a later one, no interesting set of questions we can ask the player and then build on with increasingly complex level design, and not even any pretension towards creating such. This is a series of scenes to be seen one-at-a-time as "content", that build on each other only in the most superficial way of story cutscenes that are barely related to the game itself, philosophically indistinguishable from Dragon's Lair.
Kojima might be the first director ever to include his own one-star review of a game with the game itself. Metal Gear Solid 3 is an openly waved white flag, a complete capitulation to everything Kojima railed against with Metal Gear Solid 2 1. While he previously mocked the idea of single-screen challenges with no relation to each other being any kind of "Substance", here he begrudgingly accepts it and creates it for a playerbase ready to lap it up. If this was meant as satire, or as a cutting critique of the state of the medium, it would still be bad as a videogame, but it would at least be laudable as criticism. Unfortunately, other evidence indicates that this isn't the same kind of act of rebellion that Metal Gear Solid 2: Substance was, but merely an act of resignation, a justifiably bitter director saying "fine, here you go, have what you want, I don't care anymore," and half-assing his way through a title destined to sell millions.
Let me be clear here, Metal Gear Solid 3 is not a work of subversion, it is no "External Gazer". As a stealth game, it is a clear response to the Thief series, Splinter Cell, and Far Cry (the original, not the sequels). The basic stealth mechanics, where the player crawls across a surface that provides a good background for their chosen camoflage, is clearly intended to maximize player satisfaction. The tension provided as a player acts almost as an observer, hoping that their camoflage is good enough that they won't be spotted as they await their moment to act (in contrast to the first two Metal Gear Solid games where stealth consists mostly of not being in the line of sight of enemies, and as such is based more around movement) is clearly an answer to western stealth games on the PC and XBox, and their usage of shadow and foilage, and it is safe to say that it is not played for any irony, given how intrinsically satisfying these moments are when they work well (which, in turn, robs the narrative of any guile as characters extol the exhilaration of battle). This is a game that desperately wants to be a videogame, not a referendum on the medium itself. Kojima has always had a difficult relationship with boss fights, being a director who makes videogames that are usually not focused on direct combat but still feeling a need to respect traditional videogame structures, but the ones here are about as traditionally, mechanically, good as they could be, given how poorly stealth games tend to handle the idea of a forced, direct confrontation (which is to say "they're mostly bad, but not in an ostentatious way like Metal Gear Solid 2 where they are clearly designed to be unsatisfying to drive a narrative point"). The survival mechanics detailed above that laid bare the contempt the creators had for their creation weren't added as a way to make sure that even the dumbest player saw the mechanical ironies and was ready to raise pitchforks and torches against the idea of one-screen bits of content as laid bare; it was the opposite, a desperate way to try to tie together a game of moving from one screen to the next in unrelated setpieces, with the Metroidvania2 structures of the first two games having been consigned to the flames of "progress" (remember, this game was released at the end of 2004, when backtracking was most definitely out unless you were a sidescrolling handheld game), because we lived in a post-Deus Ex world (lol) and "realism" (lol) was the future (lol).
But, it's actually even worse than the above makes it sounds, because as much as it attempts to ape the superior western-developed stealth games of the late 1990s and early 2000s, it fails to learn their most important lessons. "Chokepoints with guards" abound with few to no options for navigational circumvention (the original Metal Gear Solid was actually far better in this regard!) And, once the alarm is tripped, you might as well drop the controller to get to the game over screen as quickly as you can if you're not next to a screen transition that turns off alarms, because fighting isn't an option because of infinitely respawning enemies and there's zero hope of re-establishing stealth, completely ignoring the lessons in elasticity that should have been learned from the Thief and Splinter Cell games it tries so hard to ape. I have written in the past that Metal Gear Solid V must have been designed by a team who were smoking crack, given that they thought that regenerating health in a stealth game was a good idea, but condemnation where it is due, Metal Gear Solid 3 beat it to the punch on this particular mystifyingly bad decision by over a decade. It's just absolutely dire all around, taking the mechanics that didn't work in the past while removing the ones that did, and marrying them to the mechanics of the game's present but without understanding any of the "why"s behind those mechanics.
So, why is this obviously bad game heralded as a classic?
Simple -- it came out in 2004, when aesthetic separation was at its peak in videogame criticism, and when the affirmative answer to "are games art?"3 consisted solely of "yes, because they contain bad movies within them." If your review has a separate section labelled "gameplay", and your review score is the average of that, "graphics", "music", "sound", and "story", as was the norm in that cursed era, then Metal Gear Solid 3 will score about as highly as any videogame conceivable, with its production values being what they are. Kojima's sense of style is proven to come to him effortlessly naturally in this game, because a product that was so obviously the result of band-aids and a "eh, good enough" mentality is dripping with style everywhere. That seems to be where the commonly held idea that "Kojima wants to (or should) make movies!" comes from (I certainly never heard anyone saying that about Metal Gear Solid 1). However, that seems to be a misreading of the actual evidence; attempts at mechanical expression are constant, just not very well done. This is probably most obvious in the boss fights; most representing an emotion that Snake needs to overcome, but also being designed so that the player themselves needs to overcome it (i.e., the player needs to overcome their fear of The Fear's arrows and just stand in first person and blast him; he needs to overcome impatience and romanticism against The End and just knuckle down for an hour-long fight; he needs to overcome a need to push the action against The Fury and mostly let the boss come to him; etc), with touches like the "What a Thrill!" song coming in against The Boss as the timer ticks down obviously marrying a mechanical conceit (hunting down a camouflaged enemy as a timer ticks down) with an emotion (the adrenaline of battle, knowing that you're operating on the clock to get this job done). However, it's consistent in the stealth mechanics as well, with the exhortations to "become one with nature" on this mission being complimented by the stillness of using camo as opposed to the motion-based "keep a wall in between us" stealth of earlier Metal Gear games. No, Kojima wasn't trying to make a movie here instead of a game; he just made such a flawed game that many people didn't even notice that the baseline efforts of mechanical expression had been made. The End may be a test to see if the player has actually "become a soldier" and will knuckle down and do the job, but it's also about as simple as a boss fight can be -- take your time, don't do anything stupid, and you'll never lose this fight. In failing like this, it also becomes a pretty damning indictment of the whole camouflage system. The Boss may be intended to be a climax visually and mechanically, with the white-on-white visuals fueling the battle, but it fails at all of this because the player will invariably realize that the Thermal Goggles are the easiest way to approach the fight, thus completely ruining the aesthetic effect. Almost every screen of the game has some similar flaw, and because almost every screen of the game introduces some sort of "unique" wrinkle that's discarded on the very next, the player never enters a flow where flaws can at least be understood and accepted; instead, it becomes a milder case of the same syndrome that makes The Legend of Zelda: Twilight Princess the worst videogame ever made, where the game becomes more of a sequence of minigames, each failing to get its point across and therefore preventing mechanical immersion to the point that most players don't even notice that it's failing4.
Kojima doesn't want to make movies, but quality comparisons to the brilliant Death Stranding also make it clear that he sure didn't want to be making Metal Gear games, either.