Why Dota Sucks — 5. Auto-Attacking

Synopsis: Defense of the Ancients and most of the games in the dota genre use the auto-attack mechanisms common to real-time strategy games, and the concept has wide crossover appeal with the fans of many computer game genres. To understand its role in the genre, you must understand that most dota games line the battlefield with powerful defensive towers, and the goal of the early-game is to control the pace of action in the “farming” phase. Quite simply, the purpose of auto-attacking is to provide an element of skill in this phase. However, much like the free-roaming camera, the auto-attack mechanisms were intended for an RTS genre where you are playing a commander who has indirect control over the action. The end result, once again, is that the player is disconnected from the action because the dota genre should be providing direct control of your character. It is a system where the character initiates attacks, rather than the player. The body of game history and even certain dota games will demonstrate that, within the context of a character-action game, auto-attacking is the inferior opinion.


Note (12/14/14) – Edits have been made addressing the argument that dota games feature a toggle for auto-attacking.

With discussion of the core visual choices out of the way, we can now commit our time and energy to the game mechanics. In beginning this discussion, we should once again point out that most of the design decisions in the dota genre boil down to the following: “Warcraft III and Defense of the Ancients did it, so we’re doing it.” However, the matter of auto-attacking—and the entire circus of design choices surrounding it—goes the furthest in demonstrating how ubiquitous this really is.

Because custom maps played on Battle.net were inextricably bound to network and input latency, Warcraft III offered its mapmakers minimal support for more conventional controls and combat schemes. Most games in the dota genre now follow that precedent and use the auto-attack mechanisms found in real-time strategy games. Instead of using a series of inputs to initiate and perform your range of basic attacks, dota games allow the character to automatically attack a target so long as the target is visible and it is in their attack range. You can interrupt and stall the automation by using other commands: Movement, casting spells, and even repeatedly issuing the “Stop” command. But generally speaking, if your character can attack something, they will attack it. To govern how fast a character or unit can attack, the process is tied to their “attack speed”, which can be increased by gaining levels, buying items, or being the target of enchantments. While some dota games feature a toggle for auto-attacking, the toggle was conceived well after the genre’s conception, and auto-attacking is intended to be a central concept.1

Auto-attacking provides two core appeals to fans of the dota genre. The first is that the system poses as a marginal or superficial improvement on the combat systems in a number of different games.2 The fans of dota games are very often coming from other computer games—StarCraft II, Warcraft III, Diablo II, World of Warcraft—which lack support for the complex input strings found in more conventional action games, whether the combos in fighting games or the multiple-button inputs in brawlers.3 For fans of RTS games, auto-attacking will feel comfortable and natural. For fans of Diablo games for the personal computer, auto-attacking “streamlines” much of the clicking that defines those dungeon crawlers. And for fans of the MMORPG, dota can present similar concepts in a fast-paced action game that is not bound to the significant server latency that comes with those MMORPGs.4 Even if the combat in dota is limited, it’s appealing to audiences who don’t think it is.

But in addition, auto-attacking presents an interesting psychology of design. Very commonly, you will hear that fighting games are “button-mashers” and real-time strategy games are about “who clicks faster”. Basically, the argument is that good players are only beating bad players because they can perform more commands, instead of performing those commands efficiently. Auto-attacking ensures weaker players won’t be “out-clicked” or “button-mashed” by those better players. And in order to increase your damage—to put yourself in a position to “out-click” those players—you must outplay opponents when everybody starts the match on a level playing field. Proper management of your limited attack power will create separation between stronger and weaker players, and will do it in a very overt and obvious way.

But if you’re wondering how auto-attacking creates depth, we need to understand that it is not the only piece of the puzzle. And try to bear with me on this one, because this is a bizarre circle of interconnected design decisions. Traditionally, versus multiplayer games use some sort of mechanic or concept which forces players to engage their opponents aggressively or cede all initiative and get slaughtered.5 Your game has to do this. If it doesn’t, the game will become slow and defensive and it will compare unfavorably to games that are fast and aggressive. The simple rule in these games is that if you’re not attacking, you’re probably losing, and it works.

Everything in the dota genre suggests this “initiative” should come from the weave of “towers” or “turrets” that litter the battlefield. Anyone who has played Warhammer 40,000: Dawn of War, Company of Heroes, Team Fortress 2, the Call of Duty series, or even Unreal Tournament 2004‘s Onslaught mode should be familiar with the concept of node control, where players contest control points and try to capture them. However, both StarCraft and Warcraft III offered limited support for the concept or idea of capture points, because these games were using their resource nodes (gold, minerals, gas) as the control points.6 Aeon of Strife and Defense of the Ancients are simply treating the control points like all of the other structures in a conventional RTS: They cannot be captured and they can only be destroyed.

While there are ways to quickly wipe out the towers—commonly known as “push” strategies—they are foremost an anti-rushing mechanic, a way to give weaker players breathing room as they need it.7 Players usually need to gain several levels in order to sustain any assault on a tower, and as a result, the most common early-game goal is to manage the pace and flow of combat. This circus is known as the “farming” phase. As the computer-controlled soldiers head down their lanes and into battle, players want to strike as many “last-hits”8 as possible, to control the gold and experience that comes with making the fatal blow.9 As characters grow strong enough, the spoils of last-hitting become the backbone of the mid- and late-game situations where players are more likely to engage in large team battles and coordinate attacks on enemy structures.10

This is where auto-attacking comes in. The dota genre has had many issues providing players with the fast-paced input that defines many point-and-click action games, because in using the RTS genre its base, it is reducing the scale of RTS down to a small handful of units. In order to maintain skillful player input, the auto-attack function exists to perpetuate an element of skill in last-hitting and the farming phase. Early on, timing your automated attacks in order to score the killing blow may seem difficult enough. But as the level of play improves, you will then have to measure all the other variables in play—where enemies are located, what your teammates are doing, the health totals of soldiers—against the last-hitting process.11 The process of competing for last-hits will eventually become its own game and skillset.12 Thus, the simple reasoning is that when you combine auto-attacking with last-hitting, it makes for a deeper game with a more defined separation of skill between weaker players and stronger players. If you can’t last-hit, then you aren’t going to beat the people who can.

Auto-attacking exists to validate a circle of ridiculous design choices, and it’s all for nothing. One fundamentally misunderstands the purpose of auto-attacking when you transplant it from an RTS game into a character-action game. Because again, in RTS games, you are playing the role of a commander. And even if you are playing as a participant on the field, you are controlling most of the soldiers on the field through indirect means. You are not that soldier. You have no direct connection to the soldier on the screen. You’re superior to that soldier, and he is to follow your orders as you issue them. But at the same time, it wouldn’t make sense for your soldiers to stare down the enemy and fail to take initiative. Unless you’ve told the soldier to stand their ground, they don’t need orders to fire on the target that’s moving towards them.13

As the large collection of individuals in your command apply their marginal intelligence and capabilities to the situation, this helps to make the game world feel like a real place. That was precisely the appeal of Populous and SimCity from the moment those games went to market at the end of the eighties. Auto-attacking makes sense in a strategy game where the player is providing indirect control for most actions: Telling others where to go, what to kill, what tactics to use, and then allowing those participants to apply their own intelligence to those goals.14 But you know what happens when you apply it to action games which rely on immediate, forceful, and direct control of one participant on the screen? You have compromised the repetitive elements in your game.

This is a failure that runs two layers deep, a pair of mechanisms that must be understood in tandem. That first is that the dota genre is co-opting that indirect control from RTS games, whereas character-action games demand the most direct and complete form of control possible. One day, we’ll probably be using neural or full-body controls to fulfill that direct control. But right now, since we’re still using mechanical input devices (controllers, keyboards) to interact with game worlds, we have to improvise. That means when a character swings a sword, it is most commonly mapped out to one button press. One button press, one swing.15 That’s how you create the illusion that you are swinging the sword. But in dota, you are simply automating the attack process and continuing the frenzy until the target is dead, you are dead, or you acquire a different target.

One can merely look at the similar games which also use a high-fantasy motif— the Fable series, Dragon’s Dogma, Kingdoms of Amalur: Reckoning, the Souls series—and you will find a list of games which never considered the concept of an auto-attack feature and would feel mushy and inconsequential with one. But if none of those games are convincing, then look no further than 2011’s The Last Story. The role-playing brawler allows the player to make their own decision on the matter of auto-attacking, providing those players with a toggle for the “Attack Type”, and the Automatic Attack Type allows the character to initiate their melee attacks by simply running into the target. What was likely a concession to the fans of Hironobu Sakaguchi’s previous games, a “Who’s Who?” of Japanese turn-based role-playing games, will compromise the experience if you do not realize the option for a Manual Attack Type is burrowed in the options menu.

“But what about games like World of Warcraft and Diablo? Both of those games feature elements of automation, and people seem to like those.” And with it, we get our second layer, the one where dota becomes significantly and demonstrably worse than those games. As substandard as the approach to combat in MMORPGs and isometric brawlers may be, those games still require players to initiate combat. They require you to press the button on your mouse, keyboard, or controller, and it effectively forces you to acknowledge the combat state. But in co-opting the RTS mechanisms, the characters in Defense of the Ancients, League of Legends, Dota 2, Heroes of Newerth, and Dawngate initiate combat simply if they are within range of a target, no matter how much attention that the player is otherwise giving to their character. The character is initiating the combat, rather than the player. And even if you disable auto-attacking in Dota 2 or League of Legends, concepts such as “attack-move”16 still exist and still use the automation.

Quite remarkably, the inferiority of the auto-attack concepts can even be demonstrated within the dota genre! In playing Awesomenauts and Dead Island: Epidemic—both games which use more familiar approaches to character-action combat—you will find that the combat is superior for the simple fact that these games do not use the concepts laid out in Defense of the Ancients.17 These games may feature elements of automation, but they require players to initiate combat and are handled in a way which does not break the character’s connection to the action. The more conventional, tried-and-true combat systems in these games merely compare unfavorably to the better character-action games which came before them, whether 3D brawlers like Devil May Cry or (in the case of Awesomenauts) 2D action games like Metal Slug. But even in this losing situation, those dota games—and the action games they fall short of—demonstrate there is more potential for a great experience as you move away from the automation.

And you have preserved auto-attacking for what, because it “adds depth”? Unlike the free camera, which at least provides players with greater spatial awareness, auto-attacking is nothing more than a mechanical barrier which bears no relation to the aesthetic presentation. It exists for no other purpose than to be restrictive, and to demand more from the player in the absence of more complex concepts. If auto-attacking is “good” because it “adds depth”, and you can’t explain how it makes the feedback loop more immersive, engaging, or elsewise, then it isn’t that good of a concept to begin with. Again: The ideal goal in game design is to come up with skillful elements that are liberating and visually interesting, and the dota approach to auto-attacking is simply not one of those things.

What may seem like streamlining or convenience for the fans of Blizzard RTS games, the Diablo series, or the MMORPG—and an element of skillful input for veteran dota players—only dumbs down the combat for those who are familiar with the action games popularized on consoles and in the arcades. And if you would argue that dota actually improves itself by automating the repetition, you are arguing the repetitive elements in these games are so awful that it is better to forego them entirely.18 When, as countless action games in a wide variety of genres have demonstrated, the goal should not be to simplify or remove the repetitive elements, but to improve them.

Continue to Chapter 6: The Chess Pieces

Discuss this chapter on the forums.