Open-World Blues: On Mechanical Dynamism and Difficulty Curves

Paper written for the 2023 American Society for Aesthetics Rocky Mountain Division Meeting.

Following C. Thi Nguyen, games can be understood as a form of artistic expression in which aesthetic effects are produced through player agency.¹ Game designers constrain player agency by imposing structured mechanical rulesets and parameters that enable and reward certain kinds of action while discouraging and disallowing others. In modern action-adventure video games, two major approaches to overall game design have emerged: linear and open-world. In what follows I explore some problems that such open-world design presents for achieving a traditional feature of action-adventure games, namely a consistent difficulty curve. Through an analysis of several recent major video game releases, I argue that much modern open-world design tends to undermine the possibility of satisfying play mechanic progression over the course of the game, leading either to flat difficulty curves or to erratic difficulty spikes. Moreover, I suggest that the open-ended exploratory mode valorized in such design presupposes a merely negative concept of agential autonomy.

Open-world design in action-adventure video games has become increasingly popular and critically acclaimed in recent years. Generally speaking, open-world design can be understood in contrast to more classical ‘linear’ design, in which progression through levels and areas in the game is tightly controlled by designers. One can think for instance of the classic Legend of Zelda titles A Link to the Past (1993) and Ocarina of Time (1998), in which, with only a few very minor exceptions, players will work through the 10 or so major dungeons in a predetermined order.² Such games can include designed possibilities for players to sequence-break, as in the hidden Star Road warps in Super Mario World (1990) or more infamously and extensively in the Metroid series, without ceasing to be linear in overall design. Linear games like this are thought to better allow designers to craft player experiences with tight narrative arcs and dynamic mechanics, but at the expense of more limited player freedom.

By contrast, open-world games feature areas and/or levels that players are free to explore in any order they choose, even if there is a supposedly ‘best’ nonmandatory order. While such games have existed for a long time—some histories point all the way back to Elite (1984), a space-trading game for the BBC Micro home computer system designed by David Braben and Ian Bell³—their modern-day iteration was arguably born with Rockstar Games’ immensely popular Grand Theft Auto III (2001), which allowed players to freely roam about its 3D world from a third-person perspective. In retrospect, perhaps even more influential still was Ubisoft’s Assassin’s Creed (2007), which introduced many of the core features found in popular and successful open-world games to this day, including the Far Cry series (2008-present), the Batman Arkham games (2009-present), the Red Dead Redemption games (2010 and 2018), Horizon: Zero Dawn (2017) and Zero West (2022), and Ghost of Tsushima (2020). Just as linear games can have sequence-breaking and still be linear, open-world games do not necessarily need to give players unconditionally free rein to explore in order to be considered open-world. Many of these games divide their worlds into a small number of self-contained areas which unlock as the player progresses through key points in the game’s narrative; to give you a sense for just how standardized this practice is, the first and last games just mentioned—Grand Theft Auto III and Ghost of Tsushima—both break their world maps into three areas that unlock sequentially. Some games are more absolutist in this regard: many of the games in the Elder Scrolls (1994-present) and Fallout (1997) series, as well as the most recent Zelda title Breath of the Wild (2017), truly do let players go almost anywhere from just about the first moments of the game. But in all cases, open-world games are valued for their comparatively high degree of player agency, and freedom of exploration is generally considered to be their most highly prized virtue and most attractive feature.

Let me give you two examples to back up my claim that games with open-world design have become increasingly popular and critically acclaimed. These two games will be my focus in much of the following analysis, so this context is important. 2017’s Breath of the Wild was the best-selling launch title in Nintendo’s history,⁴ and sold some 30 million copies by 2022;⁵ it was named Game of the Year by numerous industry publications, and to this day has the highest number of perfect review scores of any game in Metacritic’s database.⁶ Five years later, FromSoftware’s Elden Ring (2022) similarly received universal acclaim from critics,⁷ racked up a staggering number of Game of the Year awards, and sold over 20 million copies within its first year of release.⁸ Both games were praised especially for their exploration mechanics and their expansive, largely unrestricted open worlds. I think it is plausible to claim that at this point open-world games are at least as, if not more, popular and critically acclaimed than games with more traditional linear design, and Breath of the Wild and Elden Ring are the crowning achievements of this approach both commercially and critically.⁹

However, I will argue in what follows that these two games illustrate a problem inherent to open-world design, namely its incompatibility with (a) dynamic mechanics and (b) a satisfying difficulty curve. My claim is that open-world games will tend either to have static mechanics and a flat difficulty curve, or to have dynamic mechanics but will be unable to avoid problems in difficulty dynamism. To make this case, I will first define mechanical dynamism, and explore how it relates to mechanical difficulty and difficulty curves. Then I will argue that Breath of the Wild embodies the first horn of the above dilemma while Elden Ring embodies the second. In a brief concluding section, I will argue that the kind of exploratory freedom for which these games are so venerated implies a generally negative conception of agency, and make the case that there are good reasons to want to prefer a more positive conception of agency in gaming experiences.

Mechanical Dynamism and Difficulty Curves

A game’s mechanics are the different actions that a player can perform. I define mechanical dynamism as the degree to which a game’s mechanics change over the course of the game. More mechanically dynamic games generally introduce new mechanics, while less mechanically dynamic ones give the player fewer new options for interaction as the game progresses. 

Mechanical dynamism is not the same as mechanical complexity. Super Mario Odyssey (2017) is quite mechanically complex, with around a dozen different moves and an enormous set of combinations in which they can be performed; but the game is mechanically static, not dynamic, because these moves are all available from the beginning of the game and basically no additional mechanics are introduced. Mechanically static games feel more or less the same to play throughout their runtime. Mechanically controlling Mario in the game’s first level, Cap Kingdom, is experientially identical to controlling him in its final secret level, Darker Side. 

At the other end of the spectrum, Super Metroid (1994) is highly mechanically dynamic: one begins the game with only a basic suit and beam weapon, but in time will discover a large number of items that change the player experience considerably: missiles and super missiles, bombs, energy tanks, five different beam modifiers that can be selectively combined, two different suit upgrades, and numerous movement modifiers such as the morph ball, grappling beam, space and high jump, speed booster, and more. The range of mechanical choices open to the player controlling Samus is much larger by the end of the game than it was at the beginning. Most Zelda games are also very mechanically dynamic, as each dungeon provides the player with a new item involving a new gameplay mechanic. The experience of mechanical dynamism is generally one of increasing positive agency. 

How do these mechanical considerations relate to the question of game difficulty? A difficulty curve maps the degree of a game’s mechanical difficulty over the course of its runtime. Let’s consider a game that is neither mechanically complex nor mechanically dynamic. In Assassin’s Creed II (2009) pressing a single button performs an assassination, holding one button down while pushing the control stick forward performs all climbing and parkour actions automatically, and the few features that are added to the simple combat system over the course of the game do not significantly alter the playing experience. Controlling Ezio feels virtually identical at the start of the game as at the end, with a few minor changes, but the range of mechanical choices open to the player is always small. There is also little enemy variety: only four different kinds of enemies and no mechanically unique boss fights. As a result, the game’s difficulty is also basically static from start to finish. There are just not many ways that the designers can make things more difficult for the player, given that the game is both mechanically simple and mechanically static. Assassin’s Creed II has a very flat difficulty curve. 

Many games strive to implement more dynamic difficulty curves, where the degree of difficulty changes over the course of the game. This can take different forms. A linear upward curve represents gradually increasing difficulty; a parabolic upward curve represents increasingly increasing difficulty, a game that is ‘easy to learn, hard to master’. Some games, like FromSoftware’s Bloodborne (2015), arguably have downward curves, where high degrees of early game difficulty progressively give way to a mechanically easier experience; Bloodborne’s first level, Central Yharnam, will likely be the hardest part of the game for a new player.¹⁰ None of these different forms of difficulty dynamism is necessarily better or worse than the others; all are valid design decisions, and different kinds of players will prefer different difficulty dynamisms, represented by differently shaped difficulty curves. 

It is generally agreed, however, that one wants to avoid difficulty spikes, where the game suddenly and unexpectedly becomes much more difficult than it had previously been. Perhaps the most infamous example of a difficulty spike is Ocarina of Time’s Water Temple, which is considered not only significantly harder than what came before it, but harder even than anything else in the game after it. Difficulty spikes can be very frustrating experiences, and often serve to turn players away. More rarely, sometimes the opposite problem arises, where a game suddenly becomes much easier. We can call these difficulty drops, and while they are not frustrating in the same way as difficulty spikes, they too tend to be unsatisfying experiences. Thus a general rule for difficulty curves is that, whatever its intended shape, it should not involve too many spikes or drops; it should be relatively smooth.

Static Mechanics and Difficulty in Breath of the Wild 

How does what we’ve established thus far interact with the principles of open-world design embodied in Breath of the Wild and Elden Ring? Let’s start with Breath of the Wild: how mechanically dynamic is it? It has an impressive degree of mechanical complexity. Its combat system is relatively simple, with only three different kinds of melee weapons, a few different special arrows with elemental characteristics, and low enemy variety. But outside of combat there’s a lot going on. Link can famously climb nearly everything in the world, and he possesses a paraglider. There are four different magical abilities, which allow Link to freeze water, stop objects in time, manipulate metal objects, and produce bombs. A complicated food system allows link to cook many different dishes using combinations of raw ingredients, eating which can restore his health or provide buffs. Moreover, the game has robust physics and even a ‘chemistry’ engine, amounting to a very large set of physical rules and enabling many of the aforementioned systems to interact with one another.¹¹ For instance, grassy fields can be lit on fire with fire arrows; the fire will spread thanks to the wind, it will cook any raw food items it comes into contact with, and it will generate an updraft that Link can use to propel himself into the sky with his paraglider. 

All of this adds up to a very mechanically complex game: there are a huge number of action options available to the player at any time. But it is also very mechanically static. All the above systems are active from the very beginning of the game; Link leaves the initial tutorial area, the Great Plateau, in possession of all four magical abilities and key items. Few new mechanics are ever introduced. Four characters give Link special abilities for completing their optional quests. One generates an updraft for his paraglider, which as we’ve noted can be accomplished in other ways. One will revive Link if he receives enough damage to kill him, while another produces a shield that protects him from damage. And the last attacks nearby enemies with lightning. All of these have long cooldowns, some nearly 20 minutes long. Besides this there is a special armor that enables Link to swim up waterfalls. None of these significantly affect gameplay; some are usable only in highly specific conditions; others are mechanically redundant; some are passive effects that leave gameplay unaffected entirely. I therefore argue that none of them change the fact that the game is essentially mechanically static.

Importantly, this lack of mechanical dynamism is necessitated by the game’s open-world design aspirations. Once you’ve left the Great Plateau, the intention is that you are able to go anywhere in the massive open world. Previous Zelda games had more linear world design that was typically enforced precisely through the introduction of new mechanics; for instance, in Ocarina, one can access the Gerudo Valley only after acquiring the Longshot from the Water Temple, or else by acquiring Epona, Link’s horse, which requires completing a complicated sidequest, both of which constitute significant new mechanics. But using mechanics to gatekeep access to world areas runs counter to the open-world design intentions of BotW. 

The game’s mechanical stasis interacts with the question of difficulty in three different ways, given the game’s three primary forms of mechanical challenge. First there is combat. The combat system itself is very simple, with just three melee movesets; the only difference between a tree branch wielded as a sword in the early game and a powerful late-game sword is numerical damage output. Similarly, late-game enemies are distinguished from early-game ones only in their numerical damage output and larger health pools. Each of Link’s kills adds points to a hidden counter; when various point totals are reached, enemies in the world level up: red enemies become blue, then black, then silver.¹² This means that as Link’s HP and damage output rises during the course of the game, enemy stats scale along with him. But this yields a basically flat difficulty curve in combat: most enemies are never particularly easy or difficult or easy, but have stats that are evenly matched to Link’s. On the other hand, the exceptions to this general rule, such as the Lynels that can be found in the world, often represent intense difficulty spikes. Indeed due to the game’s damage calculation and weapon degradation systems, encountering a Lynel early in the game can amount to an enemy that is not just difficult but literally mechanically impossible.

Second there are the game’s take on dungeons, which are the four Divine Beasts. Because they can all be accessed from the start of the game, and in any order, there can be no significant difference in difficulty between them. Each is basically a small self-contained puzzle box, and although some of these puzzles are arguably a bit more difficult than others, none are very challenging. If (say) Vah Naboris was much more difficult than Vah Ruta, this would constitute a practical barrier to the aim of allowing players to freely explore the world in any order. Thus, once again, the difficulty curve here is very flat.

Finally, there are the 120 shrines scattered throughout the world. These too can be accessed in any order. 29 shrines are empty inside, featuring only a treasure box.¹³ 25 of these require that some quest or puzzle in the overworld be solved before the shrine can be accessed, and these are generally quite simple. 70 shrines feature a puzzle of some kind. Many of these do not iterate on or explore their core puzzle mechanic; indeed, when puzzles require the use of a magical ability, the solution is often functionally identical to the tutorial that introduced players to the mechanic in the first place. For these 99 shrines, the difficulty curve is basically flat. 

The remaining 21 shrines are combat challenges, each of which contains a variant on a single enemy type. These are broken into three tiers: minor, medium, and major tests of strength. Now if the player accesses the minor tests in the early game, medium tests partway through, and major tests toward the endgame, the difficulty curve will be flat just like overworld combat. But players in the early game can stumble onto medium and major tests, which repeat the Lynel problem: these spikes can range from extremely difficult to actually impossible. And late-game players can come across minor tests, which amount to significant difficulty drops, where the enemy poses no risk whatsoever.

Notably, none of this content is mandatory: the player does not need to do any of the Divine Beasts or even any of the shrines other than the tutorial ones in the Great Plateau. It is possible to leave the tutorial area and go directly to face the game’s final boss, Calamity Ganon. In order for this to be possible, it is necessary that he must be defeasible even for a character with weak, early-game gear and no health or stamina upgrades. Indeed doing this is arguably the only way for Ganon to present any real challenge. Completing the Divine Beasts progressively simplifies this fight mechanically, and with late-game gear and upgraded player stats Ganon will be extremely easy.

The general pattern that emerges here is that the static mechanics and open-world design aspirations make it impossible to establish a satisfying difficulty dynamism. The difficulty must generally be kept relatively low so as not to discourage player freedom, but the resulting flat difficulty curve can be punctuated by sharp difficulty spikes and drops depending precisely on how the player takes advantage of this exploratory freedom.

Mechanical Dynamism and Difficulty in Elden Ring

Elden Ring’s mechanics are almost entirely combat-based, and they are complex as well as dynamic. To give you a sense for its extraordinary mechanical complexity: the game’s leveling system allows players to customize their characters by investing points into one of eight different attributes. There are 308 different weapons in the game with a large variety of movesets. There are also special abilities: 70 sorceries, 101 incantations, and 91 weapon arts. While there is some redundancy here, many of these are mechanically unique and most have distinct statistics and attribute scaling. Because these weapons and abilities are scattered throughout the world and all have specific attribute level requirements for usage, the range of action options continually expands throughout the course of the game. I have left out the game’s hundreds of consumable items, armor sets, summoning abilities, and the various systems governing special statuses such as poison, frostbite, and bleed. The result is such a high degree of mechanical dynamism that players have near-total positive freedom in build customization.

How does this interact with the game’s open-world aspirations and the difficulty question? FromSoftware’s games are notoriously difficult, and Elden Ring is no exception. But where previous games in their catalogue had broadly linear structures, enabling the designers to craft consistent and satisfying difficulty curves,¹⁴ Elden Ring’s gigantic open world, which has only a few gates, makes this completely impossible. Let’s briefly consider its structure.¹⁵

As with BotW, most of the game’s content is nonmandatory, though it has more imposed structure. A wide-open early- and middle-game gives way to a relatively linear endgame. Players will need to defeat any two of four major bosses found in the early areas of the world map in order to enter the city of Leyndell. (Godrick, Rennala, Radahn, and Rykard can all be defeated in any order, though this is the most obvious and arguably the easiest order in which to face them.) At that point, the game’s possibility space narrows considerably:¹⁶ players will need to beat this area’s two bosses, then progress to the Mountaintops of the Giants and defeat the Fire Giant; from there they are teleported to the dungeon Faram Azula, where they will need to defeat the Godskin Duo and Maliketh the Black Blade. At that point they must return to Leyndell to defeat a final series of bosses: Sir Gideon Ofnir, Godfrey/Hoarah Loux, Radagon, and finally the Elden Beast. All of this must be done in that precise order.

The worldmap is split into eight areas, five of which can be accessed from the start of the game. Players begin in Limgrave, and will need to pass through Liurnia and the Altus Plateau on the way to Leyndell. In the simplest case, Caelid, the Weeping Peninsula, Mt. Gelmir, and the Consecrated Snowfield can all be skipped entirely.¹⁷ There is also a whole series of fully optional underground areas: Ainsel River, Siofra River, and Deeproot Depths. In addition to ten or so ‘legacy dungeons’—that is, large, intricately-designed dungeons with multiple bosses, resembling the tightly designed areas of FromSoftware’s earlier games—the world map is peppered with minor dungeons: 17 caves, 15 catacombs, 9 tunnels, and 3 hero’s graves. These optional areas all feature enemies and terminate with a boss fight, the reward for beating which will be a unique item, such as a weapon or spell.

This enormous open world creates two kinds of major problems concerning difficulty. First of all, unlike BotW, enemies in Elden Ring do not level up to scale with player progression. So as with the combat shrines in BotW, designers have no way to ensure that players encounter any of the nonmandatory content in any particular order. Thus, the same difficulty spike and drop problems arise even in the overworld: a low-level character may discover a catacomb in Mt. Gelmir that is far beyond their capabilities, just as a high-level character may explore the Weeping Peninsula only to find that its main dungeon Castle Morne and its final boss present no mechanical challenge at all. While BotW’s enemy scaling entailed a mostly flat difficulty curve, Elden Ring’s static enemy placement combined with its open world and high mechanical dynamism opens the door for extremely uneven, even seemingly random difficulty spikes and drops. Designers can only guess at what level players will be when they encounter any particular piece of content, and it is very easy for players to break the game’s mechanical difficulty entirely by accident by being over-leveled or under-leveled relative to where they explore.

Second, while it is possible for a player to do only the mandatory content in progressing through the game, this is highly unlikely on a first playthrough, and it seems clear that the designers intended for players to complete at least some nonmandatory content while exploring the open world, especially in the early and middle game. As they do so, they will unlock new mechanics and level up their characters, increasing their damage output and their health pools. The problem that arises here is that a character who has completed much of the nonmandatory content—including seven major world map areas, numerous legacy dungeons, and 51 minor dungeons—will be extremely high-level and be highly mechanically capable by the time they reach the lategame. Where BotW had a very easy final boss, the designers of Elden Ring wanted to ensure that their mandatory final bosses presented a challenge even for a character who has thoroughly explored the massive open world. They accomplished this by giving those bosses extremely large health pools and enormous damage outputs. 

The result is that many of the lategame bosses represent intense difficulty spikes even for very high-level players, and will be almost impossible for anyone who has stayed mostly or only on the mandatory path. This is difficult to quantify and is in some ways subjective, but perhaps the following examples can illustrate my claim. One of the four semi-mandatory bosses that must be defeated before accessing Leyndell is Rennala; in her active phase, she possesses 4100 HP. Players dealing 200 damage per hit, not unreasonable for that stage of the game, will need to hit her around 20 times. By the very lategame, players with maxed-out strength statistics and equipped with fully-upgraded greatswords will deal somewhere around 600 damage per hit. The mandatory Fire Giant has 42000 HP, and so will need to be hit around 70 times. He also deals enough damage to kill players even with very high health pools in 2 or 3 hits. The fight is thus a long, drawn-out affair requiring players to successfully avoid taking almost any hits while landing dozens of blows of their own. The Fire Giant is not unique among lategame bosses for being so unforgiving thanks to its numbers. The Godskin Duo has 26000 HP, so maxed-out players will need to land 45 hits with the added difficulty that the player is fighting two enemies at once. Some lategame bosses, like Maliketh the Black Blade, the Elden Beast, and Malenia, Blade of Miquella have tens of thousands of hit points and abilities that will kill even high-level players in a single hit. Because the difficulty here is increased simply by inflating damage and health pool statistics, it is not mechanically interesting and is more often frustrating than satisfying.

The second problem, the lategame difficulty spike, could have been avoided if the designers had not so drastically inflating boss statistics, but then of course high-level characters would be punished for having explored the vast open world—and exploration is the selling point of such world design—as the lategame bosses would then amount to a series of underwhelming difficulty drops. The first problem, of uneven difficulty spikes and drops owing to high mechanical dynamism and open-world design, cannot have been avoided at all, unless the designers opted for an enemy scaling system such as that of BotW, but as we saw, that would yield a flat difficulty curve. 

Conclusion

I have argued, using Breath of the Wild and Elden Ring as case studies, that open-world design undermines the very possibility of consistent and satisfying difficulty dynamism. In the first case, the game’s low degree of mechanical dynamism entails that the game’s difficulty curve is basically flat. In the second, the game’s high degree of mechanical dynamism makes it impossible to prevent the difficulty from being uneven and random. Additionally, where BotW’s commitment to freedom meant that it needed a very easy final boss, Elden Ring’s designers sought to prevent this by artificially inflating the difficulty of its final bosses beyond what many players will find enjoyable or even feasible.

These games are highly critically acclaimed precisely for their open-world design, and are lauded for the exploratory freedom granted to players. And it is true that both of these games allow players a high degree of freedom to explore their vast worlds, and anyone who has played either will attest that it can feel great to do this kind of exploration. In closing I want to make two critical remarks about this, and suggest that we have good reasons to prefer the virtues of more linear world design.

First, the thrill of exploration has relatively low replay value. Players will no doubt be very excited the first time they discover the Master Sword in the Great Hyrule Forest, or when they realize that there is an entire series of underground areas in the Lands Between. But that momentary excitement does not survive on subsequent playthroughs. Players come back to games because they are mechanically satisfying, and I would suggest that it is not worth it to sacrifice satisfying, dynamic mechanics for the sake of making such moments of open-world exploration possible. Second, the conception of player agency behind this valorization of exploratory freedom is merely negative: it involves an absence of obstacles; go where you like, nobody is stopping you. But I think positive player agency is much more interesting. This involves giving players a wide variety of action options, the ability to create unique outcomes and determine the shape of their experiences based on their active choices. Both of these games have high degrees of positive player freedom by virtue of their many interacting and highly complex systems, as I noted. But there is no reason why this kind of positive freedom needs to be placed in an open-world sandbox. Ironically enough, I think, abandoning the control over intentionally crafted experiences, made possible by more linear design, does not give these positive-agency enabling systems enough room to really breathe.

1. C. Thi Nguyen, Games: Agency as Art (New York: Oxford, 2020). ↩︎
2. These are quite rare exceptions: in Ocarina, for instance, leaving aside how the game’s (notoriously numerous) glitches allow for illegitimate sequence breaking, it is possible to beat the Fire Temple (5^th) prior to the Forest Temple (4^th), but the first 3 must all be done in order; the Forest Temple must be beaten before it is possible to proceed to the Ice Cavern (6^th), which must be beaten in order to enter Water Temple (7^th), Ganon’s Castle (10^th) can only be done last, etc. ↩︎
3. “The complete history of open-world games (part 1)”, Computer and Video Games, 24 May 2008 (http://www.computerandvideogames.com/article.php?id=189591), archived at the Wayback Machine. ↩︎
4. “What Does the Nintendo Switch, Breath of the Wild Launch Success Actually Mean?”, IGN, 10 March/27 April 2017 (https://www.ign.com/articles/2017/03/10/what-does-the-nintendo-switch-breath-of-the-wild-launch-success-actually-mean). ↩︎
5. Sal Romano, “Switch worldwide sales top 122.55 million, Pokemon Scarlet and Violet top 20.61 million”, Gematsu, 7 February 2023 (https://www.gematsu.com/2023/02/switch-worldwide-sales-top-122-55-million-pokemon-scarlet-and-violet-top-20-61-million). ↩︎
6. Brenna Hiller, “Zelda: Breath of the Wild has the most perfect review scores in Metacritic’s history”, VG247 (https://www.vg247.com/zelda-breath-of-the-wild-has-the-most-perfect-review-scores-in-metacritics-history). ↩︎
7. “Elden Ring for PlayStation 5 Reviews”, Metacritic, retrieved 20 April 2023 (https://www.metacritic.com/game/playstation-5/elden-ring). ↩︎
8. Anthony Wood, “Elden Ring Sells 20 Million Units”, IGN (https://www.ign.com/articles/elden-ring-sells-20-million-units). ↩︎
9. Two of the other highest-rated games in the years between these, according to Metacritic, were also open-world: the already-mentioned Red Dead Redemption 2 (2018); and Divinity: Original Sin 2 (2019), an RPG that has been described as “[leaving you] wide open when it comes to determining a course of action, with very few moments forcing you down a particular path” (Brett Todd, “Divinity: Original Sin 2 Review”, Gamespot, [https://www.gamespot.com/reviews/divinity-original-sin-2-review/1900-6416769/]). For 2020 and 2021 the highest-rated games were also RPGs (Persona 5 and Disco Elysium), but these were not open-world. Thus since 2017, 4 out of the 6 highest-rated games were open-world. ↩︎
10.  According to the PlayStation Network’s Trophies system, only 64% of players defeat Father Gascoigne, who serves as the boss of the first level, while 43% defeat Mergo’s Wet Nurse, the final mandatory boss. In other words, more than a third of players never beat the first level, but two-thirds of those who do go on to beat the whole game.
    ↩︎
11. Kate Gray, “Is The Legend of Zelda: Breath of the Wild the best-designed game ever?” The Guardian, 30 May 2017 (https://www.theguardian.com/technology/2017/may/30/the-legend-of-zelda-breath-of-the-wild-nintendo-game-design-open-world-player-explore). ↩︎
12. “World Level”, Zelda Dungeon (https://www.zeldadungeon.net/wiki/World_Level). ↩︎
13. On the topic of shrines I rely on Joseph Anderson’s analysis in “Breath of the Wild – Not Enough Zelda” (https://www.youtube.com/watch?v=T15-xfUr8z4). ↩︎
14. These broadly linear worlds still left players to make some decisions about progression; oftentimes linearly-organized blocs of areas can be done in any order. For instance, in Dark Souls (2011), players (1) must ring the two Bells of Awakening in any order, then (2) progress through Sen’s Fortress and Anor Londo, then (3) must seek out and fight the four Lords in any order, and finally (4) proceed to fight the final boss. In this way they are perhaps closer to what God of War (2018)’s lead designer Cory Barlog has described as ‘wide-linear’. See Dean Takahashi, “God of War full interview – The definitive story behind the crafting of God of War”, VentureBeat (https://venturebeat.com/2019/01/06/god-of-war-full-interview-the-definitive-story-behind-the-crafting-of-god-of-war/view-all/). ↩︎
15. Details about the world and enemy statistics for Elden Ring are taken from the game’s fan wiki (https://eldenring.wiki.fextralife.com/). ↩︎
16. See Mark Brown’s excellent “The World Design of Elden Ring” (https://www.youtube.com/watch?v=LvnlvB9n6ic). ↩︎
17. The distinctions between ‘worldmap areas’ and ‘legacy dungeons’ is somewhat leaky; I go back and forth as to whether Miquella’s Haligtree should be considered one or the other. I’m classifying it as a legacy dungeon here. ↩︎