Most of the challenge of games comes from the pressure of not losing. And so, it warrants spending more time thinking about the purpose of losing from a game design perspective. Since loss conditions interact with system complexity and mastery in non-obvious ways, I’ll describe a combination of loss conditions that I think is the best for maximising player engagement.
Some games do not have loss conditions: either the player reaches the goal and wins, or they stop playing (for example The Witness, cannot be lost, even if individual panels can). But without some kind of restriction, reaching the goal is often a trivial task that only measures time spent. To avoid this problem, most games include some kind of constraint that makes reaching a goal or gaining a higher score a challenge.
Symmetrical two-player games have the most elegant loss condition: if one player wins the other loses. This symmetrical design can also be used for single player games by replacing one player with an AI. However, designing an AI that can pose a challenge to beginner, intermediate, and expert players alike is an arduous task, and often outside the scope of games design (but see Minos Strategos’ solution). Other games, like solitaire, have an asymmetrical design that pits the player against the system. But without an AI to battle against and determine the loss condition, the designer must impose one themselves.
Timers are the simplest loss conditions to add to a game - they end the match after set amount of time or number of turns, if the player hasn’t already won. They are flexible and can be tuned to provide beginners with space to explore the system, or tuned to prune back all but the best possible sequences of actions. As a timer becomes less forgiving, it forces the player to become increasingly efficient at traversing the system.
But while timers are effective at demanding efficiency, they are insufficient for achieving other design objectives. Pac-man with a short timer instead of ghosts is little more than a relaxed version of the travelling salesman problem. Players are challenged to simply plot out the an efficient-enough route through all the pills. The player could calculate their ideal path in advance and simply autopilot their way through (see This article about issues with calculation). This is clearly not the type of coin-popping gameplay the designers wanted to encourage. Instead the designers used four ghosts that randomly move around the maze to create a faster-paced, more dynamic system. The player is forced to avoid these ghosts since they steal a life when they catch the player, and once the player is out of lives the match ends. These ghosts drastically change the ideal path through the maze from something flat and easily solved to something dynamic. These changes reduce the player’s ability to reliably plan ahead and encourage short-term decision making and probabilistic thinking - all of which are desirable properties for a fast-paced action game.
Quantifying player failure using independent loss conditions
Lives lost measure how many times the player has failed to conquer certain challenges on their way to the goal of level completion. In general this type of loss condition allows us as designers to prune the range of options that a player can use to reach their goal by introducing another criteria by which to value the results of the player’s actions. In Pac-man, ghosts prevent players from calculating the most efficient path once and then executing that plan. Players much prefer a dynamic experience--one that requires them to react to new information throughout gameplay. In an FPS we may punish a player that stands still, because we know moving and shooting a target is harder and more interesting. In a hack ‘n’ slash (or beat ‘em up) game we may punish players that simply mash the attack button, because we know that, while plowing mindlessly through enemies is enjoyable, balancing the timing of attack and defense is far more rewarding. These ways of measuring what we define as failure for our game allow us to incentivize players to engage with the most fun and rewarding parts of the game at the expense of boring and trivial approaches. I call these “independant loss conditions”, since unlike timers they can be defined in the absence of any positive goal. They simply measure how many times the player failed some challenge.
Hit point (HP) mechanics are an alternative to lives that articulate the degree of failure in a fine-grained way without causing state to reset. For example, in a fighting game you can be hit many times before you lose a round and different actions, such as blocking special attacks, take off different amounts of HP. As a designer this level of control in failure measurement gives us a lot of freedom in determining just how badly a player can afford to do. But, please keep in mind, the more variety and detail you add to your failure measurement, the more complexity your players will have to learn. Be sure that each addition is worth the complexity cost.
Saving independence with interdependence
The main problem with independent loss conditions is that they are very flat and simplistic, they simply incentivise the player to avoid certain actions. The reason for this flatness is their independence from the goal. As far as the goal is concerned, it doesn’t matter whether you die once or a thousand times, it only matters that you reach it. Similarly these loss conditions don’t care whether you are about to reach the goal or are miles away, they only care that you failed so much that you cannot be allowed to continue. This means we are essentially asking the player to work towards two independent goals: one prescribes actions they should be doing and one prescribes actions they should not be doing. If you are going to double the number of things the player must pay attention to, you should try to maximise the benefit you get from adding such a feature. Fortunately it is simple to make loss conditions less flat and more dynamic: explicitly link them to the goal.
We can design incentives for the player to take the risk of approaching the loss condition by offering them a reward that advances them towards the goal. For example, in Crypt of the Necrodancer, blood weapons do a lot of extra damage when you are exactly 1 hit from dying. This provides a huge advantage since they can dispatch most enemies instantly, but it is very risky since you can die just as quickly. In a system with an independent loss condition, actions move the player closer to winning or move them closer to losing, but with an interdependent loss condition, actions take them closer to losing and closer to winning at the same time. By making the goal and the loss condition interdependent, we generate risk-reward trade-offs that force the player to make difficult decisions that shift the course of the entire match.
Now that we have loss conditions that are trade-offs, the player should think about decisions as either increasing or decreasing their chances of ultimately reaching the goal. And because the player doesn’t have perfect information about the future, they should think probabilistically. So they ask ‘what is the chance of this action helping me win, versus the chance of it causing me to lose’. We can also think about HP and lives as a resource that a player can leverage to increase their chance of winning.
When viewed in this framework we can see that the cost of this resource (how close the player is to losing), varies throughout the match. Its value increases the closer they are to losing. The cost of losing their last bit of HP is infinite, while the cost of losing their first piece is negligible. And so, the cost of each successive mistake, and each trade-off is higher than the last. But this resource also becomes less valuable the closer they get to winning, since it does not matter how far from losing they are when they won, just that they won. This means players will be incentivised to spend almost all of their HP as they near victory. For example, in an FPS if you have full hp and are trying to finish off the last boss, you’ll be willing to stand still and eat a chunk of damage to maximise your chance of landing that next rocket in their weak point. This is a desirable property for a game system: having interesting trade-offs that vary in value throughout a match since they draw the player closer to winning and to losing. But the cost structures that constitute the trade-offs cause a kind of degeneration as costs near zero or infinity--suddenly the incentives that worked so well cease to be considerations.
One common way of eliminating this kind of degeneracy is to make the goal and the loss condition more deeply interdependent: approaching the goal should also increase the risk of losing. This is usually done with a difficulty curve (for example a multi-phase boss battle, or the difficulty spike of the final waves of a horde mode). This has the positive outcome of ensuring that tension persists even in the final moments before victory. While these approaches can feel-heavy handed, more systemic solutions may be possible. For example, scoring points could summon additional enemies, with their difficulty tied to the amount scored. Just take care designing and balancing complex feedback loops like these.
Open ended play undermines win-loss trade-offs
As players approach the highest levels of mastery of a game they know almost perfectly how to avoid losing. In some games, they become so good they can indefinitely hold off losing, until fatigue, rather than poor decision making does them in. Take for example high score chasing in Asteroids. Asteroids is practically solved, so getting a new high score is mostly a matter of endurance, which is not an interesting skill to test. If we think about the cost structure of any game that lacks timer, prioritising survival above all else can potentially result in infinite time/turns to pursue the goal. This kind of overly conservative play is something we should prevent, since it is a tedious waste of time. It is a mistake to think we can resolve this issue simply by tweaking the cost-benefit structure of actions as they relate to winning and losing. Instead we must be certain that the even the very best players will lose if they take too long, and the best way to do this is to use some variation of a timer.
Using all of our tools
Placing a hard turn or time limit on a game stops expert players from playing maximally cautiously. Instead they must treat their proximity to losing as a resource and think about how they can get the most return on investment over the course of the game. A well-designed system using timers and interdependent loss conditions asks the player to maximise the efficiency with which they convert their resource (proximity to losing) into their chance of winning. As stated previously timers encourage players to be efficient, and efficiency is a very desirable trait to encourage at the upper echelons of mastery of a game. I would argue that all games that include strategic decisions should have some system that acts as a timer, simply for the fact that it encourages efficiency, let alone that it also counteracts conservative play.
Now of course traditional timers aren’t the only way to bring matches to a close, we can also use systems that act like timers, so called soft timers. A soft timer would be something like constantly increasing the speed that tetrominoes fall in Tetris, or increasing the number of asteroids in Asteroids after each wave (but far beyond the levels reached in those games). Eventually if difficulty is raised high enough, the rules of the game will not allow the player to survive, no matter how skilled they are. This kind of soft timer has the nice property of being woven into the difficulty curve of the game (see Imbroglio for a good example). However, I would be hesitant to recommend it unless the designer is confident that the skills being tested are valuable, the difficulty becomes insurmountable, and the sources of randomness in the game don’t interact too unpredictably with match length.
Loss conditions allow designers to limit the pool of player actions and strategies to those which we think are the most interesting. While independant loss conditions allow designers to cut the player off from the least fun parts of a system, we need to make them interdependent with the win condition in order to justify the amount of complexity they add to a game. In such a system, a player mistakes are actions that take them closer to losing with little or no upside, while trade-offs take them closer to winning and losing. Timers, whether soft or hard, encourage the player to be as efficient as possible as they become more strict. As such, they should be included in any game, where mastery and strategic decision making are central. When used together timers and interdependent loss conditions encourage a special type of efficiency where the player must trade-off the risk and reward of approaching the loss state across the entire course of the game. By ensuring that the value of approaching the loss state varies somewhat unpredictably throughout the match, we ensure that there’s always an interesting decision for the player to make.
I’d like to thank Evizaer, Hopenager, Keith, Vivafringe and Redless for the articles and discussion that led to this and Hopenager and especially Evizaer for feedback and advice on the article.