Follow-up to: Free-to-Play Games: Three Key Trade-Offs

The central free-to-play mechanic is to ration action and resources via real world time. This leads to two of the three key trade offs. Players are prevented from having fun because they are time restricted, either unable to play or unable to have the resources to play the way they would like, allowing the game to sell a solution to these problems. More perniciously, players become trained to constantly check in with the game in order to claim rewards and keep their resources from becoming idle. This can warp a person’s life more than one would think, changing behavior to allow timely access, and preventing focus on other subjects.

This obsession effect, and the ability of real world time delays to be an interesting resource to include in trade-offs, have also caused these mechanics to seep into non-free games, especially RPGs.

Resource rationing takes the form of timers. The form of the timer does a lot to determine how toxic the rationing will be to the player. There are several knobs one can turn.

The Knobs

Many of these knobs represent related aspects, and thus are closely intertwined, but listing them out still seems useful. In each case, moving towards the first named end will reduce toxicity.

1. Steady versus Sudden: If a resource accumulates over time, does the resource accumulate gradually with no limit, gradually up to a limit, or all at once?
2. Fixed versus Delayed: Does the resource replenish at a fixed time, or at a time after it is used?
3. Slow versus Rapid: How frequently must one check-in to maximize results?
4. Batched versus Disjoint: Are there multiple timers running simultaneously? If so, how hard are they to line up?
5. Tracked versus Lost: How easy is it to track when accumulation is complete?
6. Forgiving versus Punishing: How punishing is it to fail to check in?
7. Isolated versus Cumulative: Are there cumulative rewards for reliably checking in? How important are they?
8. Progressive versus Competitive: Are you in competition with others? If so, what kind?
9. Queued versus Idle: Can you give orders in advance?
10. Explained versus Mysterious: Can you tell what matters?
11. Annoying versus Limiting: Are the things you get from timers the long-term limiting factor in your ability to do things?
12. Incidental versus Central: Is timing the key determinant of your success?

Knob 1: Steady versus Sudden

If you have a resource that replenishes steadily over time, then there is a broad window during which it is efficient to utilize that resource. Even if it has a maximum amount you can store, you can spend it at any point up to and including the moment you have stored up that maximum, with no losses. While you lose potential compound interest gained from whatever the resource might have bought you, and there are usually worries about maximum resource storage, the player can mostly relax most of the time. Even if the resource briefly stops accumulating, this tends to be a relatively light punishment. There is still the worry that spending the resource now is always better than spending it later, as it reduces future risk, and this can line of thinking can still be harmful.

If you have a resource that replenishes all at once, then knob two becomes very important. The risk is that the time of the sudden accumulation becomes something the player worries about missing, and thus schedules around. The player might even sits idle waiting for that time to arrive.

Knob 2: Fixed versus Delayed

If you have a resource that replenishes once per day, at 12:00 midnight Pacific time, you have a full day-long window in which to use that resource. There is no reward or punishment for using it in the morning, afternoon or evening. The player does not feel distracted, and does not have to worry about losing resources when they delay. It does create potential large pressure near the end of the full period, if the resource has yet to be used and cannot be stored.

If you have a resource that replenishes once per day, twenty-four hours after it is used or claimed, then every second you do not use the resource is another second it will be delayed, forever. If you wait until the time when it is convenient for you to use it, you will need to wait until after that time the next day, and generally run the risk of losing an entire cycle whenever things get delayed at all. There is no stable, relaxed equilibrium available.

If the accumulation is delayed, then that raises the importance of several of the other knobs.

Knob 3: Slow versus Rapid

If full accumulation takes a day, then it is easy to see why this will probably not be too disruptive. Check-in need be at most once per day. A steady and slow accumulation is definitely mostly harmless.

If full accumulation takes place every few hours, or less than that, then this can prove extremely disruptive, to the point of costing the player sleep and a constant state of distraction. This is especially true if it is also sudden and the next accumulation is delayed.

The cost of potentially missing a full accumulation can have high emotional resonance. If you are given resources continuously, every moment that this is taken away can be seen as losing out on those resources. If you are going to continue to participate anyway, it can feel extremely bad to let this happen, and seems to be able to trigger the loss aversion circuits in our brains.

Knob 4: Batched versus Disjoint

If your timing triggers are batched, you can respond to each batch as one action.

If your timing triggers are disjoint, you cannot do so without sacrificing efficiency, and the effective tax on your attention and time is much higher.

Suppose your widgets renew every two hours, and your whatsits renew every three hours. You need to check in at hours 2, 3, 4 and 6 every six hour period. You get some benefit of being able to align the second three-hour window with the third two-hour window. If these are delayed triggers, than they might become slightly disjoint, and the intervening time might be wasted literally looking at a countdown timer. The alternative is to not wait around, but if you do that, then the two become fully disjoint.

Consider a game where there are a bunch of different timers and queues – for example, you might have a construction timer, a research timer, an army training timer, a special reward timer and an army task timer. If these always lined up, you might be able to check in (let’s say) only three times per day with minimal loss, and each time you’d have a bunch of fun things to do. Instead, you feel bad if you do not check in fifteen times per day, and each time you have only one thing to do.

Juggling does have an upside, if you have control over the length of the timers. If you have the ability to choose between different things to build, research or train each of which takes a different amount of time, then you can trade off what you want most per unit time against what will allow you to check back in at the same time? Similar planning problems involving when you sleep, work or are otherwise not going to check in can also be interesting, or they can risk actually disrupting your plans in bad ways. Or both.

Knob 5: Tracked versus Lost

If accumulation is easy to track, that is far less distracting. If you know that accumulation will be sufficient to take action, or will be at maximum levels, or otherwise what you want, at exactly 10:34 AM today, then you can set an alarm, or you can remember to check in at about that time, or you can have your phone notify you, ideally having the game send a notification at that time. The better and easier you can find out and track the right time to do things, the less attention you have to pay to make sure that you’ll be there at the right time, so less time and attention is wasted. It is important that games give you this information in forms that are easy to understand and to track.

If accumulation is difficult to track, things can be far worse. An example of this is if the timers are misleading, forcing you to adjust them and to not be able to wait for your notifications. A prominent game I explored allows the player to speed up any timer by spending resources, with the last ten minutes of this process being free. Thus, if something has nine minutes to go, you can click on it to complete it now, so it is essentially finished but for your noticing, and the resource is effectively sitting idle. This ten minute window means that every timer in the game is off, and every notification of this type comes too late, forcing upon the player constant paranoia. It was a relief when, later in the game, average timer length expanded enough that ten minutes was no longer something worth worrying about.

Knob 6: Forgiving versus Punishing

What happens if you fail to check-in at the requested time? Do barbarians burn down your castle? Does your voyage through the stars run out of anti-matter, forcing an abandonment of all rewards or a payment of precious premium currency? Or does your progression only pause briefly until you are convinced to genuflect in the game’s direction in the form of a few clicks?

If you can lose a lot of progress, or even a lot of your resources, by not paying enough attention, that is a very big deal and highly disruptive. You need to have a system where that almost never happens, or the game likely becomes unplayable. If all that happens is your accumulation temporarily halts, the other trade-offs can make this somewhat painful, but it mostly seems fine to play and note care much about full maximization.

It’s pretty real-world terrible to hold these types of threats over people’s heads. I once read a website that described what machinations one should go through each evening players of a free-to-play game should go through, as non-paying players, in order to be able to go to sleep at night. 

It is hard to imagine what is being offered in exchange for that, that would make the right response anything but the uninstall button.

Knob 7: Isolated versus Cumulative

If each timer is isolated, then missing one timer and its associated rewards does not impact future timers. Each timer is its own opportunity. The marginal penalty for missing one is small.

If the timers are linked, then the important rewards are usually based on consistently checking in for most or all of the timers. The marginal benefit of meeting each deadline goes up as you meet more of them, and missing even one often sends you back to square one to start over.

Extreme versions of this stretch rewards over periods of weeks or months, or require intense levels of activity on specified event days or weekends. You have to continuously spend time, then if you don’t focus in when they tell you to, most of what you have worked for is lost. The central ideas are the concept of having daily rewards that require not missing timers and then a daily reward for completing all the daily rewards, a login bonus or other reward for doing actions on consecutive days with no misses, and then a ‘mastery track’ that gives you increasing marginal returns for consistently getting all those daily rewards.

Knob 8: Progressive versus Competitive

If you are free to progress at your own pace in an essentially static world, the game will let you play in a way that is compatible with your life. You do not have to worry that other players will defeat you by making the sacrifices you won’t. All that happens if you don’t go crazy is that the game gets slower and harder. That might not even be a bad thing.

If you are in a race against time because you face off against other players in a competition to power up the fastest, all starting from roughly the same point as a new shard or region is created, then you are at constant risk of falling behind if you do not maximize your time and money spends, especially if the timers are as central to progression as they usually are in free to play games.

Games that are by their nature competitive races can offer unique and rich experiences. The risk is that when there is not clear separation between the resources of the game and the resources of your life, where does it end? The central reason the mechanics described in Meditations on Moloch of sacrificing everything to keep up in zero-sum competitions doesn’t actually eat the world – the reason that, to quote an upcoming post that hopefully begins an important sequence, Moloch Isn’t Winning, is that such competitions are almost always asymmetrical and complex, with winning not being a direct function of effort and resources, and with the value of winning not being that centrally important.

Games that centrally feature timers are usually intentionally designed to topple this barrier. Reward is explicitly locked to effort, in the form of money spent and time devoted, with skill and strategy intentionally crippled in importance beyond a basic level of competence. Thus, all the various complexities that make it not worth sacrificing one’s children to Moloch are removed from the equation.

A game’s explicit structure of winning and losing also takes away the second most important defense, which is that it’s usually not as big a deal to not win such competitions as one might make it out to be. Games are about winning.

Of course, as Robin Hanson would respond, games are not about winning. Which is also true and important, but doesn’t change the true and important, and more relevant in conext, fact that games are about winning.

(Leaving a marker here for myself both to eventually fully explain this particular flip, and the general case viewpoint on how to approach when X isn’t really about Y but is also of course totally, totally about Y.)

Knob 9: Queued versus Idle

Queues, where you could take action in advance and tell the game what to do when the time came, would mitigate many of these issues. Build queues for towns or buildings in even turn-based strategy games can be the difference between such games being great versus being so tedious as to be almost unplayable. Many mobile games would be much better if, when you tried to do something but lacked the necessary resources, you could click a button that said “do this when resources are available.” Other similar tricks could be used in other situations.

Idle however is the universal rule. You need to tell every resource what to do after it becomes available, or it will sit idle until you do.

It’s easy to see why idle is chosen over queues. The whole point of the system, as we’ll discuss after the list is finished, is to force frequent check-in and obsession to facilitate habit formation, addiction and obsession. This potential solution is solving the player’s problem, not the game’s problem.

Knob 10: Explained versus Mysterious

This is a variation of Tracked versus Lost. If you don’t even know when checking in will be rewarded, then you certainly can’t track it. With Lost, you have a hard time remembering or recording exactly when something will happen, but you likely know more or less what will happen and about when it will happen. If things are sufficiently mysterious, you can feel an obligation to continuously check in case something happens, without any theory as to what or when or how. Eventually it becomes a compulsion without a justification.

Knob 11: Annoying versus Limiting

Knob 12: Incidental versus Central

These last two knobs ask how much the game is fundamentally about its timers.

Annoying timers mean that your life will be worse off if you miss them or mismanage them, but over the long term the number of check-ins is not the determining limiting factor of progress. You’ll end up in the same place. For limiting, that means that the timers give you the main source of the resource or resources that are the limiting factor for your progression. If there are severely limiting factors, it is quite possible for most other resources and accomplishments to not matter.

Incidental versus central compares the generic value of what the timers give you to the generic value of what is otherwise available. In the extreme case, the key limiting resource is only available via timers, or even timers with checking in as the only requirement, and the rest of the game does not actually exist. Instead you end up in the situation described in a comment on my previous post, by Villam:

I will not mention the name of the game here. Anyway, it was the type of game where you build stuff, collect resources, and research new stuff; with many things to unlock. In the game there were three important resources, let’s call them X, Y, and Z. By making better or worse decisions, you could make more or less of the resources X and Y; and I spent some time optimizing for that.

With resource Z, however, the basic way to get it was to play the game regularly. If you logged in at least N times a day, you got M points of resource Z per day; you couldn’t get more for playing longer, but you would get less for taking breaks longer than 1/N of the day. In addition to this, there were also some other ways to get resource Z, but this extra amount was always smaller than the amount you got for merely playing the game regularly. There was no smart strategy to at least double the income of Z. So, whether you did smart or stupid things had a visible impact on X and Y, but almost no impact on Z.

Of course the resource Z was the one that actually mattered, in long term. Your progress on the tech tree sometimes required X and Y, but always required Z. And, of course, the higher steps on the almost-linear tech tree required more of the resource Z.

A popular variation of this is to make your progress impact the rate at which Z is collected when you check in. It then matters how efficiently you navigate the early stages in order to level yourself up, because it increases your timer rewards, and that is the primary thing to optimize for.

Of course, all of that presupposes that your goal is to make your numbers go up as high as possible over a very long term of calendar time in a world in which the purpose of numbers going up is… to make them then go slightly higher than that, at ever slower rates. Not exactly the most enticing proposition, when stated that way.

Ask this question: Would you do better long term if you did nothing today but check-in for the timers, or if you did everything else but didn’t check-in with the timers today?

As The Knob Turns: Toxicity versus Compulsion

Games with timers are using real world time as a managed resource. As noted, this can be an interesting design space and set of optimization problems.

I see three categories of toxicity trade-offs regarding timers.

There are timers that require toxicity because they are putting real world attention into the game as a resource constraint.

There are games that are not trying to do that but which have to pay costs to avoid imposing costs.

Then there are most of the games in the genre, which are mostly using Skinner box tactics to create habits and compulsive behaviors. Toxicity is not a bug, it is a feature. It is the hill that they climb. It is the killer app. Their goal is to turn all twelve knobs as far to the right as possible without players taking too much notice.

The first case is sympathetic. I do believe that it is real and legitimate, and not merely a cover for the third case, but its presence in its good form is rare. An interesting choice requires trade-offs, so the desire to minimize toxicity and real world cost in exchange for in-game benefits becomes the real game. Done right, that’s cool, and you can’t have those real world stakes without at least some real world costs. There is a real trade-off. My best advice for a designer in these situations is to ensure that there are always reasonable solutions that impose only reasonable real world costs, and benefits of further attention have to decline rapidly beyond roughly the commit the player is intentionally making. It is also important that optimization works, with better approaches being faster and requiring less attention and check-in than poor approaches.

The second case is also sympathetic, and in its full form is common. In addition to real world clock time being an interesting resource to trade-off, letting players sample the game at some rate, while imposing costs for moving faster, is a relatively friendly business model. You would prefer to avoid distracting the player outside of time intentionally dedicated to your game, and you would prefer to have that dedicated time focused on interesting game play decisions rather than engaging in micromanagement. Queues are helpful here, as are many of the other knobs, which can safely be turned far to the left.

The third case is, unfortunately, the central case. Toxicity is turned on its head and embraced as the core game feature.

From the players’ perspective, one must figure out how to navigate these dangers, and whether there is a path to doing so without the game becoming net negative. When deciding this, one needs to keep in mind that these systems are designed to hijack your brain, developing habits and compulsions that may be difficult to break. Thus, if things turn out to be bad, that badness is designed to prevent you from realizing this or from being able to execute on disengagement. One cannot do the calculation assuming one can think clearly in the future.

Last time, I recommended looking at how often you were required to ping a game, and how punishing it was to fail to do so, as a way to estimate a game’s toxicity level. This remains a good simple heuristic. Looking at the knobs above allows you to flesh out this evaluation. When looking at the games I have been surveying, this made me realize that I had been fooled by at least one game into thinking it was far less toxic and far less relatively toxic than it was, and that I had to adjust modes of further play to reduce its toxicity level.

From the designer’s perspective, whether or not you have ethical concerns, the question is how to balance the costs you impose on the player, the negative reactions the player will have to those costs, and the upside from getting players addicted and in the habit of playing your game every day and in any spare moment. Here is where we see that in each of the twelve cases, turning the knob to the right increases the rate at which habit and compulsion are imposed upon the player. Thus, the limiting factor will by default be what players will accept without running away screaming or otherwise realizing the game is not their friend.

This leads into the question of how you measure that. I have been thinking a lot about what happens when you have very strong ability to measure specific short term outcomes, but much less ability to measure other longer term outcomes, and there are implicit hidden variables.

Consider the following toy model of a free-to-play game player.

The player operates based on hidden variables. They have ongoing levels of (related) things like Fun, Compulsion, Habit, Willingness to Pay, Willingness to Recommend, Annoyance, Goodwill, Trust, Social Ties, Sense of Accomplishment, and so on.

The game cannot measure these directly. Instead, the game gets to measure things like hours played, what causes players to log out or keep playing in the moment, when players spend money, when players uninstall, how players manage their resources, review star ratings, and so on.

Any given decision will get made largely based on optimizing those short term outcomes. Short term outcomes, especially time played and money spent, will loom large. It will be tough to justify sacrificing those outcomes to lay longer term foundations that may or may not exist or matter, or might even go in the opposite direction.

This is especially true because the correlations are strong between the observed short term outcomes and expected long term outcomes. There are even strong causal mechanisms explaining why good observed results now lead directly to good results later. Players who spend more now will spend more later. Players who play more now form the habits and compulsions to spend more later, as they ‘break the seal’ and identify themselves as people who spend money in this way. Interacting with friends or joining a guild leads to more social engagement.

Combine that with the ease of optimizing between known options via A/B testing in order to hill climb on the details that lead to superior metrics, and you know that the special limited time offer will be the exact correct size and color and price and duration and timing and so on that maximizes short term profits.

Measuring the formation of habits and compulsions is harder than figuring out what color causes more sales in the next hour, but it still offers solid metrics. You should see the trend lines moving in the appropriate ways almost right away, or the effect is not there. The downside costs all this imposes can be divided into two categories. There are the times that you flip a switch in someone’s head that says ‘this game is super toxic and I need to run away screaming right now.’ You learn to avoid doing that. However, there is also the gradual accumulation of annoyances and loss of trust and goodwill that comes from not giving players a good experience, or from imposing steady hard to notice costs upon them, or by being too repetitive, or what not. This sneaks up on you slowly, and any given decision is usually going to have a small impact that is not going to be directly measurable, but that will all add up over time.

These and other related Goodhart’s Law problems explain a lot of what I see as clear failures to optimize player experience, and experience in other services and places across the internet and all of our civilization, in ways I plan to explore more, but I will stop here to avoid too much scope creep.

The biggest question one wants to ask is, to what extent are you in case one versus case two versus case three. If you are in the first two cases, toxicity can be better or worse, but you should not expect toxicity levels to continue to rise, or for them to be especially deceptively high. If you are in the third case, you should assume that things are already worse than you think and designed to get steadily worse than that.

Thus, one should look for knobs that are intentionally turned to the right. If there are mechanics that seem designed to force additional check-in times, to punish failure to check-in, to force you to form reliable habits or lose out on the bulk of rewards, and so on, that should be a big red flag. Combining this with looking at expected rates of check-in should give a good picture. Going down the line on all the knobs should give a better picture still.

The next step in this exploration is to look at particular game mechanics in detail.