Month: May 2011

We frequently pick what to do in-game based on the rewards available rather than on what would be most fun. Stumbling on Happiness suggests that this might not be a horrible idea in terms of remembered enjoyment.

The two things you remember most about an experience are the most extreme event and the last one. Take the biggest emotional peak or trough, take your last thoughts, and those will stick with you long after the details have faded. If the best part of a movie is the ending, you will likely remember it fondly, even if an inventory of the moments finds it lacking.

This suggests the game design wisdom of giving out candy at the end of every quest/dungeon. If the experience ends with your receiving a shiny, you will remember it more fondly. (This also argues against making looting take less time, because you want the player to dwell on that shiny moment at the end.) This is an evolutionarily powerful meme that designers do not even need to pursue intentionally; all things being equal, games that give out candy at the end of each unit will be more popular just because human brains place emotional weight on that.

This also suggests the game design folly of risking disappointment. When faced with a loot slot machine, human brains will tend to value the high of winning more than the expected value suggests, but you are still having quite a few people end the dungeon with disappointment. Maybe they will take the boss kill as “ending on a high note,” or they will be happy for the loot roll winner, but that loot roll at the end will not be a high note for most. This suggests the rise of tokenization as a strong meme, because everyone gets a unit of candy.

It should also suggest that the common model of wiping on bosses for days/weeks before passing them is a horrible design. Multiple tries in a night could still mean ending on a win (high note), but every night that ends in a wipe is a raid full of disappointed people, except for those who take solace in “we’re making good progress.” Perhaps this falls under the loot slot machine principle, whereby the occasional wins are valued more than the frequent losses.

: Zubon

There has been much MMO blog chatter lately on optimization, “I just want to have fun,” what you owe your team/raid, exploring mechanics, etc. I wish to add 2.5 small postulates, upon which I hope we can agree.

If one option is better than another by a non-negligible margin in all cases or nearly so, it is a poor decision to choose the inferior option, and it is fair to criticize someone for doing so. This is true even if there are role-playing or decorative reasons for the choice. As an extreme example, it is perfectly valid to refuse to group with someone who is role-playing an ascetic mendicant who refuses to use equipment or someone who wants to raid in level 10 equipment because it looks nice.

A corollary is that if one option is clearly superior in the circumstance in question, one that is readily available and has no meaningful drawbacks, you are welcome to criticize someone for choosing the inferior option. Don’t be a jerk about it, but this guy is being a jerk by playing his class with only two buttons and not even the right two buttons.

The second postulate is that it is usually poor game design to have stochastically dominant options over non-idiotic ranges of comparison. If 1 2 1 3 1 4 really is the best damage rotation in 90+% of fights, you have reduced most of the game’s complexity to eight characters. If there is no point at which a Warrior should trade off Strength for Dexterity, even at a 10-to-1 ratio, your attribute system is not terribly interesting. These problems are doubly so if some aspect keeps the stochastic dominance from being apparent, such that someone might not notice the tiny DoT icons which tell you that you should never add numbers other than 1-4 to that rotation.

: Zubon

I have a friend who is a stochastic dominatrix.

My latest reading on “what is wrong with the human brain” is Daniel Gilbert’s Stumbling on Happiness. The titular theme is that human beings are really bad at predicting what will make us happy. We imagine future events but leave out important details. You have the essence of the experience in mind but forget all the details that could mitigate it, leading us to over-estimate how good the good will be and how bad the bad will be.

The author’s recommended solution, which he does not expect most to take, is to use surrogates. On average, you will better predict how much you will enjoy X if I tell you how much some random person enjoyed X than if I tell you what X is. The odds are better that your brain will miss something important than that this random person is much different from you on an important variable. It is also known that the average person thinks he is different from average for most important variables, so this means you.

Apply this for a moment. You have been following that MMO’s development for years, listening to developers, getting all the details, and on that basis you will probably have a worse estimate of how much you will enjoy it than if you just tried to get a sense of how the playerbase feels as it launches, worse than if you just asked a random player/tester how much fun s/he had. If you consider the pre-game anticipation part of the fun, go to, but consider it a separate form of entertainment rather than useful information-gathering.

Of course, as an MMO player, you are in a niche market. You can improve your enjoyment estimates by narrowing the range of potential players over whom you are seeking a random or average estimate. But remember to stick to that principle or at least keep track of how often you get burned when you veer from it: if you usually like what Ravious likes, and Ravious says he likes/hates X, you can rationally make a buying decision based solely on that. I am happy to take movie recommendations from a few reviewers who I know to have similar tastes, and Roger Ebert will usually know before I do whether I will enjoy a film.

: Zubon

I grant that Bhagpuss is a unique snowflake whose preferences will differ on the important variables.