« Burning Down the House | Main | How Low Can You Go? »

Mar 14, 2005

Comments

1.

Wonderful Job Nick. It looks great.

2.

I apologise in advance for what I figure is going to be a long post...

I notice that this latest report begins with a discussion of my player types model, so here’s a quick response.

>Bartle assumed that your underlying motivations “suppressed” each other.

I didn’t assume it, I asserted it. If you want to knock over my theory, this is therefore a fairly good place to hit it!

>In other words, the more of an Achiever you were, the less of a Socializer, Explorer and Killer you could be, but just because you like ice-cream doesn’t mean you will hate pasta.

This isn't a great analogy. Just because you’re male doesn’t mean you’re not female. Oh wait, it does.

Why is the player types model more like ice-cream/pasta than male/female? You have to say this because you can use the analogy to critique the theory. That's not to say it isn't a valid analogy, but at the moment it looks no different to the male/female one.

>The assumption of polarized motivations is also not supported by the correlations of the
current data set.

The motivations aren’t polarised – you can be an explorer tending towards achievement, for example. You can also be an achiever tending towards socialiser. What’s more, my 8-types model makes a stab at saying what kind of socialiser might become an achiever and what kind of achiever might become a socialiser. I don’t therefore see this result as a nail in the coffin of my player types model (although that’s not to say it won’t turn out that way!).

>Bartle proposed that people who like to chat and make friends are also the people who like to role-play. These are in fact two independent motivations

No I didn’t. Search through the original paper for the paragraph that includes the phrase "I personally favour the view that role-playing is merely a strong framework within which the four types of player still operate".

>While Bartle proposed that Achievers and Griefers were separate Types, they are in fact fairly correlated with each other. The Advancement and Competition subcomponents are correlated at r = .41, p < .001.

I’d expect Advancement and Competition to correlate, because they’re both qualities associated with achievers. What have either to do with griefing, though? Is it that you’re saying any form of competition is griefing? That’s not what I meant in my paper at all (assuming you’re talking about what the paper calls "killers" as being “griefers”).

>Bartle construed Explorer’s as people who enjoyed both exploring the world,gathering information as well as enjoying tinkering with the underlying system and mechanics. These are also in fact two different kinds of people. My earlier attempts to find the Bartle Explorer failed until I tried to look for those two constructs separately. In other words, there is a Discovery subcomponent that revolves around finding and accumulating knowledge that is separate from the Mechanics subcomponent that is interested at unraveling and tinkering with the game mechanics.

This is excellent news from the point of view of my 8-types model, as I have two sub-types of explorer in it that map pretty well onto what you’ve identified here.

>The Immersion subcomponents revolve around story-line, role-playing, fantasy, customization and escapism and are independent of the Socializing motivations.

This is what I’d expect, too. I don’t have a separate player type for people who play to be immersed because for me, everyone who plays for fun plays to be immersed. They have different degrees of awareness of this, but they do pretty well all want to be immersed. I see immersion as an emergent property of progressing through the player types, as I explain in my book (and, perhaps a little more coherently, in the first chapter of Massively Multiplayer Game Development 2). Thus, I wouldn’t lump it together with Achievement and Social as I feel it’s at a meta-level with respect to them. This would probably explain why it comes out as independent of the two in your analysis.

Also in your report, you respond to some of what I wrote in my book about your first Facets paper. Here are my comments on that:

>the survey doesn’t implicitly presume a grouping of statements and that was the more important
goal of the survey

Yes, OK. My criticism of the fact that the motivations are implicit in the questions was intended to suggest that there may be more motivations than you’ve yet found, and that the reasonable match with my player types model (excluding explorers) was because your questions were specifically pitched to look for these types anyway.

>The data showing discrepancies with Bartle’s original Types illustrate how player motivations can’t simply be brainstormed. They must be tested.

This is correct. It’s all well and good my making assertions about player types, and people nodding their heads and saying "yes, this matches my experience", but you can’t beat objective testing. If hard data gets us a better theory, I’m all for it.

>The labeling of the facets is not provided by the factor analysis. Nor are the labels of the
Player Types inherent in any way.

They’re not, but they follow on from the dimensions of the graph pretty well. There’s a structure to the player types model that’s currently missing from the motivations model, and unfortunately no amount of factor analysis will reveal it. When you get a structure, you might find that the motivations (well, sub-motivations – I think the main ones are probably going to stand the test of time) need some relabeling.

Some other things in the report that I’d like to comment on:

>Bartle’s Types puts people in one of 4 boxes and argues that the other 3 boxes say nothing about you.

I don’t argue that the other boxes say nothing about you; my paper says nothing on the subject, as far as I recall. In my book, I do suggest they say something about you, in that you can tell what kind of player someone is likely to become if you know what type they are and have been. That graph of yours showing achievement component against social component looks like how I’d envisage the acting/interacting versus player/world distribution to appear. It may even be that these two sets of axes are isomorphic.

>The problem is that subdivision is not the answer. People are never just one thing. The answer is not subdivision but understanding that people don’t fit into boxes.

I think the problem is that you see my types model as putting people into boxes when it doesn’t. In the third paragraph after the preface of my paper, I wrote (concerning the types I was proposing): "Most players leaned at least a little to all four, but each tended to have some particular overall preference". I’m talking about their main motivation, but that isn’t to say it’s their only one. I’d expect them to have others, as they gradually mature and work their way through the various types. I’m just as anti-pigeonholing of people as you are. Those aren’t boxes, those are shadows cast under different lights.

>aren’t members of raid-oriented guilds both Achievers and Socializers?

On the whole, no. They’re either achieving to socialise or socialising to achieve. The former are slightly behind the latter in their hero’s journey, but they’ll become the latter at the very next step.

Sorry if I come over rather critical of your report here, it’s not intended (as I’m sure you didn’t intend to be venomous of me, either!). For the record, I think the Daedalus Project is some of the most important work being done in virtual worlds today, and I’m grateful that you’re spending some of your energies hammering at my player types theory. If we get a better theory as a result, that can only make for better virtual worlds.

Richard

3.

/takes a seat
/gets a -large- order of popcorn

4.

Richard - Thanks for the feedback. Totally understood as constructive rather than critical.

Nick

5.

Richard,

Never apologise for long posts! :)

6.

In other words, the mindset of the player is a better predictor of problematic usage than game mechanics (a la Skinner Box).

That's probably right, but how do you know? Wouldn't the kind of people who would have real problematic usage be the ones to deny it? While the ones that would admit it would be the ones who are open (and therefore more capable of taking care of their problem)?

7.

The "primary motivations" shows that achievement, socializing, and immersion are primary motivators for about 1/3 of the player population each. (Achievement is slightly lower than 1/3, immersion slightly higher.)

The question that pops into my mind is: If I make a world that has no achievement, but only socialization and immersion, do I end up with 2/3 the number of players I would have had if I catered to all 3?

I don't think so. I think there's interaction between the personalities, as Richard Bartle points out in his player-type models.

Could a survey be designed to test this hypothesis? Or, could the survey data be sorted by MMORPG to determine what player-motivations each one caters to, and then try to determine a relationship between how well balanced the catering is and the number of players?

The numbers point out a testable joke-hypothesis... Since there's a very strong correlation between achievement and young males, there should likewise be a correlation between how achiever-oriented a world is and how skimpy the clothing is on female avatars...

8.

Mike said: "The question that pops into my mind is: If I make a world that has no achievement, but only socialization and immersion, do I end up with 2/3 the number of players I would have had if I catered to all 3?"

Don't put people in boxes. Just because a player doesn't have "achievement" as their primary component doesn't mean it's not a close second. The correlations among the 3 main components show that scoring high on one component has no bearing on how you score on the other 2 components.

All players have an achievement score and for each person there's a threshold at which they would leave a game. The reason you end up with far fewer than 2/3 people is because the people with their main motivation as achievement are only the tip of the iceberg. Many players with Socializing or Immersion as their main component care about Achievement too.

Motivations don't suppress each other. Stop thinking boxes. Start thinking complementary scores.

9.

/handOutMorePopcorn

First of all, I agree: thanks to Nick for all his efforts in helping to gather this data -- it's absolutely irreplaceable.

As to specifics:

Nick> there is a Discovery subcomponent that revolves around finding and accumulating knowledge that is separate from the Mechanics subcomponent that is interested at unraveling and tinkering with the game mechanics.
Richard> This is excellent news from the point of view of my 8-types model, as I have two sub-types of explorer in it that map pretty well onto what you’ve identified here.

For what it's worth, these two subclassifications can also map nicely onto two of the Myers-Briggs subcategories of Keirsey's "Rational" (NT, or iNtuitive Thinker) temperament. If the Explorer is an NT in a play-specific context, then the Discoverer fits the "Perceiving" xNTP type very well (the xNTP is more concerned with the organization and design of things than with their functional structure), while the Mechanic seems a good fit for the "Judging" xNTJ type (the xNTJ is more concerned with making a system work than with grokking the design theory behind the system).

Nick> The data showing discrepancies with Bartle’s original Types illustrate how player motivations can’t simply be brainstormed. They must be tested.
Richard> This is correct. ... you can’t beat objective testing. If hard data gets us a better theory, I’m all for it.

I can't agree with this strongly enough. In all my research on personality models, virtually every bad model was one that started with a theory and went looking for data. The few that seemed to actually work all started with raw data and formed theories to explain and predict observed patterns in that data.

Nick> In other words, the more of an Achiever you were, the less of a Socializer, Explorer and Killer you could be, but just because you like ice-cream doesn’t mean you will hate pasta.
Nick> The assumption of polarized motivations is also not supported by the correlations of the
current data set.
Nick> Bartle’s Types puts people in one of 4 boxes and argues that the other 3 boxes say nothing about you.
Nick> The problem is that subdivision is not the answer. People are never just one thing. The answer is not subdivision but understanding that people don’t fit into boxes.
Nick> Motivations don't suppress each other. Stop thinking boxes. Start thinking complementary scores.

We really, really need to get past this.

I can't and wouldn't try to speak for Richard, but I'm pretty confident I feel the same way he does on this subject: No one is trying to put people in boxes.

There is nothing inherently wrong with a model of human personality that describes people as having one of N qualities that are utterly and completely mutually exclusive. There is nothing inherently wrong with a model that describes people has having M of N qualities (where M ranges from 0 to N) that have no correlation to each other at all. Either of these models can have useful explanatory and predictive power *if* that model is composed of a theory that is a good fit for the available data, and if those data reflect reality to a sufficiently high degree.

In other words, Nick, if your approach works, that's good -- we all benefit from considering it. But if an approach that looks at people as tending to express one primary motivation over others also works (in the sense of effectively explaining and predicting behavior), then why must that approach be attacked as "putting people in boxes?" If it can be shown to work, and in no way limits the utility of your preferred approach, why try to kill it?

Here's my statement of my own Keirseian version of Richard's original four player types: A person is mostly one type of player at a time, because a person has mostly one primary kind of motivation (temperament) at a time. He may demonstrate behaviors more common to other temperaments when his options are constrained, but given the choice, he will behave like a Rational in most situations.

Furthermore, it's my personal theory that for each temperament, there are two "auxiliary" temperaments and one "opposite" temperament. A Rational will mostly do Rational things, and will sometimes do Idealist and Guardian things, but will generally avoid doing Artisan things. An Idealist will mostly do Idealist things, and will sometimes do Artisan and Rational things, but will generally avoid doing Guardian things. And so on.

In the model I use, there is some order in behavioral preferences; preferences are not fully independent. A preference for one kind of behavior defines to some extent one's preferences for other kinds of behavior. To me, this is not "putting people in boxes" -- this is a useful theory about how human beings actually behave based on empirical observation no less real than your own (if less rigorous).

My preferred model does not invalidate yours by its mere existence. Your model doesn't invalidate mine. If yours has utility, I'm pleased -- but why must your model be the *only* one?

--Flatfingers

10.

Flatfingers said: "No one is trying to put people in boxes."

But then he said: "There is nothing inherently wrong with a model of human personality that describes people as having one of N qualities that are utterly and completely mutually exclusive."

Mutually exclusive categories is like putting people in boxes. Richard doesn't argue for mutual exclusivity though.

Flatfingers' main point seemed to be: "Either of these models can have useful explanatory and predictive power *if* that model is composed of a theory that is a good fit for the available data, and if those data reflect reality to a sufficiently high degree."

The available data doesn't fit the "mutually exclusive" model. Underlying components are not negatively correlated. The data doesn't show any of the components to mutually exclusive. To be fair, as Richard describes above, he doesn't promote a mutually exclusive model either.

Flatfingers also claimed that: "A person is mostly one type of player at a time, because a person has mostly one primary kind of motivation (temperament) at a time."

That's exactly the phenomena the data doesn't show. We would expect a negative correlation among the main components if that were true. That isn't the case.

Flatfingers concluded that: "My preferred model does not invalidate yours by its mere existence. Your model doesn't invalidate mine. If yours has utility, I'm pleased -- but why must your model be the *only* one?"

The models that we're talking about have testable hypotheses. If players do have primary motivations that are dependent on each other, we would expect motivation factors to be negatively correlated. The data doesn't support this hypothesis. The data shows that motivation components are largely independent of each other and that motivations do not suppress each other.

Models do not invalidate each other. It's about how well they can predict and describe the actual data. I have shown you a lot of data that supports the hypotheses of one model which do not support the hypotheses of another model. Instead of telling me what your model assumes, show me data that supports your model.

11.

This is a wonderful model, really interesting, but one thing leaps out at me.

You're assuming here that there's a sharp distinction between analysing the game mechanics (your Mechanics subcomponent) and exploring the game world (your Discovery subcomponent). I wouldn't necessarily disagree, but I don't think the distinction is where you put it - in my (admittedly anecdotal) experience, exploring the mechanics without bothering to crunch the numbers, or caring about quantitative analysis, is very common.

I know I was very surprised at my low Mechanics score, because I've always really enjoyed figuring out which factors contribute to what, and in what relative proportions - but the questions you ask for that are very specific about precise numbers and numerical optimization.

Quoting Flatfingers above: "For what it's worth, these two subclassifications can also map nicely onto two of the Myers-Briggs subcategories of Keirsey's "Rational" (NT, or iNtuitive Thinker) temperament. If the Explorer is an NT in a play-specific context, then the Discoverer fits the "Perceiving" xNTP type very well (the xNTP is more concerned with the organization and design of things than with their functional structure), while the Mechanic seems a good fit for the "Judging" xNTJ type (the xNTJ is more concerned with making a system work than with grokking the design theory behind the system)."

Is it because I are INTP?

12.

Nick Yee wrote - Motivations don't suppress each other. Stop thinking boxes. Start thinking complementary scores.

To put the question another way:

I don't know if you have enough data to calculate this, but... How do different MMORPGs rate in regards to the 3 motivations? (For example: Is the average WoW player higher in achievement than EQII? Or higher in socialization? Or escapism? How about DAOC? AC? Etc.)

Is there any correlation between how prevelent (or balanced) the motivations of the players are and the success of the MMORPG?

13.

Nick Yee>Models do not invalidate each other. It's about how well they can predict and describe the actual data. I have shown you a lot of data that supports the hypotheses of one model which do not support the hypotheses of another model.

Maybe we're looking at different data, but it seems to me that your data does support my model (or, to be absolutely strict about it, doesn't deny my model).

Also, you have to be careful about using the word "model". Yours is a data model, yes, but it's not a theoretical model: it doesn't describe why, it simply describes that. In my model, I can make some attempt to describe why people who are currently playing predominantly as one type are doing so, and predict on the basis of that where they're probably heading next. I don't have any hard data to support it, though, so it stands or falls on its ability to work in practice. Your model says that people have three motivations for playing virtual worlds, and you have data to back it up, but you have no explanation (yet) as to why they may have these three motivations, and you don't know that there aren't other motivations you haven't yet discovered.

Now you've isolated your 3 facets, the next thing to do is to try and build a theory that explains the data. Otherwise, all you're saying is that people like adventuring, socialising and immersion, and we knew that anyway.

Richard

14.

Nick> Flatfingers also claimed that: "A person is mostly one type of player at a time, because a person has mostly one primary kind of motivation (temperament) at a time."
Nick> That's exactly the phenomena the data doesn't show. We would expect a negative correlation among the main components if that were true. That isn't the case.

You mean, among the main components that you've designed your survey questions to measure?

Why are you unwilling to accept that different questions (such as Myers-Briggs questions) could highlight different aspects of human behavior, aspects that aren't fully independent? Do you really believe that such a result threatens any of the valuable work you've done?

Nick> I have shown you a lot of data that supports the hypotheses of one model which do not support the hypotheses of another model. Instead of telling me what your model assumes, show me data that supports your model.

Fair enough.

The source data underlying the four-temperament model were acquired through the many years of designing, refining, and applying the Myers-Briggs Type Indicator (MTBI). These results were described in detail -- including some statistical analyses -- in _Gifts Differing_ by Isabel Briggs Myers and Peter B. Myers. Given the nature of this work, I'd think that anyone interested in understanding the nature of personality (including that of gamers) would want to examine this book closely, and I recommend it to you and anyone else interested in this subject. You don't have to accept its conclusions, of course; it's just the best source of the behavioral choice data you requested.

As for how the four-temperament model is constructed from this data, again I can't do better than to refer you to a book, in this case David Keirsey's _Please Understand Me II_. And again, you're under no obligation to agree with that author -- I'm just assuming that you're willing to consider a wide range of theories on the subject of general human behavior before coming to a conclusion as to which have utility and which don't.

Finally, I offer you my own experience in studying personality. It's anecdotal because I haven't worked as hard as you have to acquire hard data, but after years of people-watching and considering various theories (some of which have proven to be junk), I'm satisfied that people do have specific behavioral preferences that exclude other kinds of preferences. You don't have to accept this, either, naturally, but I think it's fair to offer it as a single data point.

Nick> Mutually exclusive categories is like putting people in boxes.

One last try to address this "box" charge....

I'm a utilitarian when it comes to trying to understand the human condition. If I see a model that seems to do a good job of explaining and predicting behavior, then I'll use and endorse that model... but if I then come across a new model that works better, I'll use and endorse that one. I might even hang on to the old one for some situations for which it has better explanatory and predictive power.

So I have no *intention* of "putting people in boxes." I'm neither interested nor disinterested in saying that any individual can have one and only one motivation. I just go with what works. If a multi-valued, fully independent model (such as yours) works, then I'll use it. If on the other hand I see a model that finds some dependency between preferred styles of behavior, and that model can demonstrate to my satisfaction that it works, then I'll use it.

The latter happens to be the case. I've studied a number of personality models with a skeptical eye, and found that Keirsey's model (which I personally believe is a superset of Richard's original four-type model) offers satisfactory explanatory and predictive effectiveness. It doesn't set out to stuff people into boxes -- it simply recognizes that there are regular clusterings in behavioral styles.

If your data don't show this, that doesn't mean there's anything wrong with your data. You may simply be measuring other aspects of human personality.

But neither do your results prove (as you seem to think they do) that the information I've seen or my interpretation of it are wrong, either theoretically or on a practical basis. If you don't like some model, fine; don't use it. But don't assume that because your data don't seem to support that model, it is false and must be beaten down. Human nature is diverse; there's room for more than one good model of human behavior.

So:

1. No one here is trying to put people in boxes. That is, no one I'm aware of has any deep-seated need to try to pigeonhole people, nor is anyone allowing such a ulterior motive to color their research claims. If that's your meaning, then it's not an accurate statement and you need to stop making it.

2. No one I've seen writing here on this subject is presenting any model that claims that individuals can only ever express one motivation, period. No one is pushing a "mutually exclusive" model. If such a model were presented, and if it demonstrated effectiveness, then I'd take it seriously (and so should you), but no one has proposed any such thing. In fact, Richard has expressly described his most current model as a kind of map that (among other things) shows the path of a journey from one behavioral style to another. And I have explicitly said that the model I prefer suggests that individuals, while tending to prefer one style out of four common preferences, seem to also express behaviors common to two "neighbor" temperaments while avoiding behaviors common to an "opposite" temperament. In other words, I'm coming right out and saying that most people aren't limited solely to one temperament. Individuals have behavioral preferences, but a preference is not a box -- if that's what you mean, then you need to stop making that assertion because it is not correct.

There is no box.

--Flatfingers

15.

Sam K.> I was very surprised at my low Mechanics score, because I've always really enjoyed figuring out which factors contribute to what, and in what relative proportions - but the questions you ask for that are very specific about precise numbers and numerical optimization. ... Is it because I are INTP?

Speaking as a fellow INTP and as someone else who finds chasing little numbers to be mind-numbingly tedious... I'm inclined to say "yes." :-)

I like seeing how the pieces fit together (or not) at a high level, but the idea of running hundreds of experiments to generate very specific details about low-level internals leaves me cold. Other people are good with details, and I've learned to be content to leave that sort of thing to them to concentrate on what I feel I can offer, which is an ability to "see" the overall structure of a thing, and to use that ability to design complex systems.

Turns out that's a typical INTP preference, but the ENTP "Inventors" share this interest in being more interested in high-level design than in detailing every last specification.

All of which is moot to someone who doesn't buy into the whole Myers-Briggs thing, of course....

--Flatfingers

16.

Flatfingers wrote: "Why are you unwilling to accept that different questions (such as Myers-Briggs questions) could highlight different aspects of human behavior, aspects that aren't fully independent?"

I think this is the heart of our disagreement. The MBTI uses forced-choice dichotomous response options. The bimodal nature of the data is an artifact of the assessment tool.

This is the reason why all modern personality assessment tools used by personality psychologists in academia don't use this approach. They mainly use trait-based scales which always show a normal distribution.

The MBTI assumes that the underlying scales are bi-modal. Those traits (such as Introversion) almost always fall along a normal distribution. The MBTI is forcing dichotomies where they don't really appear naturally. It's taking a continuum on a normal distribution (such as IE or JP) and saying there's just mainly 2 kinds of people.

I think our disagreement doesn't lie with Richard's model as much as where we stand on the validity of the MBTI. There's a reason why the MBTI isn't used in academia.

------------------------------------------------------
Richard wrote: "Maybe we're looking at different data, but it seems to me that your data does support my model (or, to be absolutely strict about it, doesn't deny my model) ... Your model says that people have three motivations for playing virtual worlds, and you have data to back it up, but you have no explanation (yet) as to why they may have these three motivations, and you don't know that there aren't other motivations you haven't yet discovered."

I think we mean different things when we use the word "model". I don't think we disagree on a taxonomy level - that there is some combination of Achievement, Socializing and Immersion going on.

I think we disagree on:
1) whether those underlying motivations suppress each other.
2) whether most people have clear primary motivations or not (and whether it makes sense to have labels such as "Achiever")
3) whether there are two underlying orthogonal axes that capture all player motivations (for example, you argue that role-playing supercedes the model, whereas I argue it's another equivalent factor).

Fundamentally, I think the core disagreement is whether it's possible for a person to be have high Achievement and Socializing motivations at the same time. In your model, motivations are zero-sum. It's not possible for someone to score high on both at the same time (because this would be different points on the graph). The data that my model is based on says the opposite. There are players who score high on both, low on both, or one high one low. The two scores are not dependent on each other.

I also don't buy the data/explanation problem you point out. Your model basically side-stepped the "why do people have these motivations" by postulating 2 hypothetical axes. But why do people have those 2 axes; are they the only axes?

I also don't believe that having those 2 axes (or 3 now) means you've mapped out all possible motivations. For example, role-playing doesn't fit into those 2 axes. In fact, postulating higher-level abstractions might simply constrain your ability to find motivations.

The original intent of the data-gathering was - "do player motivations cluster in factors the way they are proposed in Richard's model?" But I think my goal was always to get to a point where we could assess those motivations and look at how they interacted with other things (demographics, in-game preferences). I think our perspectives here are shaded by our backgrounds. For me, the utility of a model lies in its ability to generate measurable constructs and provide a foundation for further empirical exploration. I think it's more important for you to have theoretically clean models and data itself is less important.

But overall, I don't think our main disagreements are at the taxonomy level. I think they are on the level of how motivations interact with each other, and perhaps more importantly, what are valid ways of building models (with data or with intuition).

17.

Nick> The MBTI uses forced-choice dichotomous response options. The bimodal nature of the data is an artifact of the assessment tool. ... The MBTI assumes that the underlying scales are bi-modal. Those traits (such as Introversion) almost always fall along a normal distribution. The MBTI is forcing dichotomies where they don't really appear naturally.

Ah -- I understand your position now, and it makes good sense. Thanks for the explanation.

I'll give this some more thought.

Side note: My wife works for a good-sized market research company. When I asked last night what a truly random (i.e., not self-selected) 60-question survey of gamers over the Internet would cost, she quoted me a rough estimate of about $10,000. That's a little more than I'm looking to spend just to acquire my own data set, so it's good to know that you're on the job. Carry on!

--Flatfingers

18.

Mike Rozak asked: "How do different MMORPGs rate in regards to the 3 motivations? (For example: Is the average WoW player higher in achievement than EQII? Or higher in socialization? Or escapism? How about DAOC? AC? Etc.) Is there any correlation between how prevelent (or balanced) the motivations of the players are and the success of the MMORPG?"

I had large enough samples from WoW, EQ, EQ2, SWG and DAoC to run the comparison. For the following comparisons, Low is < -.3, Med is around 0, and High = > .3 on a z-score distribution. Essentially, 0 is the mean, and Low is low compared with the mean.

EQ: Low Ach, Low Imm, High Soc
EQ2: Low Ach, Med Imm, Med Soc
SWG: Med Ach, High Imm, Med Soc
DAoC: Med Ach, Med Imm, High Soc
WoW: Med Ach, Med Imm, Med Soc

The data seems weird at first. But here's my interpretation of it. The reason why EQ and DAoC have high Soc is because the only players left in it are the players who are staying for the social network. People looking for Achievement or Immersion have moved on.

Now, one might have expected WoW to score high on all 3 because of its success in attracting very different kinds of players. But a game that scores high on a score is essentially driving away players who have a low score in that component. The reason why WoW works is because it's attracting hard-core and casual or non-achievers equally. And that's why the Achievement score comes out around the mean. The game appeals to a broad spectrum of motivations.

I wonder if this is the key take-away: A healthy game system is one that has something for everyone. And I mean that in the sense that it has something for players along the spectrums of Achievement, Social and Immersion. There's something in the game that caters to casual achievement as well as hard-core achievement. Tools that allow ease of grouping and character skills that allow for soloing.

Imagine a game that forced you to group and only had hard-core achievement elements. You would drive away players who score low on teamwork or low on achievement. But here's the catch - you would also drive away the players who score high on teamwork and low on achievement. This is because players don't operate on single preferences. In essence, you might drive away the very players you are catering to ...

19.

Nick Yee>I think we disagree on:
>1) whether those underlying motivations suppress each other.

The way I view this, it's not so much that they supress one another as that one will tend to dominate. They can be complementary, too: someone may enjoy socialising, but actually be socialising so that they can achieve. Your facets model would pick them up as being both socialising and achieving, seemingly independently. My model has an explanation as to why they are socialising (ie. for socialising's sake or for some ulterior motive) but it has no data to support this other than anecdotal evidence.

>2) whether most people have clear primary motivations or not (and whether it makes sense to have labels such as "Achiever")

OK, we do disagree on this. I feel quite strongly that players do tend to have a primary motivation at any stage in their playing career. They may not always play that way, but they'll mainly play that way.

I think we might also disagree with the labels on your facets. I'm uncomfotable with the idea that what I classify as a griefer you would classify as someone acting under an achievement motivation.

>3) whether there are two underlying orthogonal axes that capture all player motivations (for example, you argue that role-playing supercedes the model, whereas I argue it's another equivalent factor).

It depends on the type of role-playing. Soft role-playing (ie. that practiced by most players) I would say fitted into the model: players are role-playing in order to achieve, socialise, grief, explore etc.. Hard role-playing, where the character is fixed and only the player changes, is more problematical, and I suspect that it doesn't fit so well (because it's not a hero's journey thing). You're not going to find many such role-players in the virtual worlds you're sampling, though.

>Fundamentally, I think the core disagreement is whether it's possible for a person to be have high Achievement and Socializing motivations at the same time.

Why do you call them motivations? Asking "how important are the following things to you" isn't asking what their motivations are; it's asking for a partial ordering of significance. An achiever might say that chatting with other players is "tremendously enjoyable" because it helps them co-operate to get more XP; you pick up the former but not the latter.

In terms of your motivations, yes, it is possible for people to have achievement and socialisation motivations at the same time, because you're testing for surface motivations only. You know they like these things you've classified as being achiever-signifiers or whatever, but you don't know that this is what motivates them. My model attempts to look deeper than that. Players can have two of what you call motivations, but be employing one in the service of the other. My model tries to address this, rather than the basic motivations.

>Your model basically side-stepped the "why do people have these motivations" by postulating 2 hypothetical axes. But why do people have those 2 axes; are they the only axes?

They're not the only axes: trivially, I could add a male/female axis that would add a new dimension. They're the axes that are useful for game designers, though, which is who my model is for.

As for why those particular axes, yes, the original paper did side-step this. In my book, though, I map the player types onto a hero's journey model, and that does explain why they have the types they do.

Richard

20.

Nick Yee wrote - The data seems weird at first. But here's my interpretation of it. The reason why EQ and DAoC have high Soc is because the only players left in it are the players who are staying for the social network. People looking for Achievement or Immersion have moved on.

Thanks. The information provides some clue as to what's going on. I wonder if it means that as VW's age they should improve their socialization features...

I also noticed that SWG has a high immersion, which could be because of its world-like nature.

21.

Nick Yee>The reason why EQ and DAoC have high Soc is because the only players left in it are the players who are staying for the social network.

Don't you have usable data going back to a time when they weren't quite so new games? Also, can't you tell from your data how long these people have been playing?

Richard

22.

EQ: Low Ach, Low Imm, High Soc
EQ2: Low Ach, Med Imm, Med Soc
SWG: Med Ach, High Imm, Med Soc
DAoC: Med Ach, Med Imm, High Soc
WoW: Med Ach, Med Imm, Med Soc

Could you be talked into breaking that down by sub-component? :)

23.

If I am reading what you said correctly, is having a High OR a Low in a game a disadvantage in the marketplace?

24.

Raph wrote - If I am reading what you said correctly, is having a High OR a Low in a game a disadvantage in the marketplace?

My read is that Nick Yee's data shows a roughly gaussian curve for distribution, or at least a curve with one peak. This means having a VW that's high or low limits the market size.

Mass-market games would probably aim for dead center because that's where the populations are.

However, the more games aiming for dead center, the smaller piece of the pie for each. A major VW aiming slightly askew might be able to "own" its own pie, albiet a smaller one that the one all the centrist worlds are squabbling over.

25.

Mike, that was my take on it, too.

But: Thanks. The information provides some clue as to what's going on. I wonder if it means that as VW's age they should improve their socialization features...

...that's raises some interesting questions, too. Not only "Should older games focus a bit more on socialization?", but maybe older games should focus a bit less on achievement.

Also perhaps a new game should focus more on achievement for a while, then shift to being more balanced as it ages, then shift to being more social-oriented when it's older still.

Well I guess it's all the same question: Should a game cater to different player motivations based on the age of the game?

26.

Jeff> Also perhaps a new game should focus more on achievement for a while, then shift to being more balanced as it ages, then shift to being more social-oriented when it's older still.

Jeff, I'd like to understand this suggestion better.

Do you mean that features should lead player demographics -- that developers should assume that the primary motivations of their players will change in well-defined ways over time, and that developers should plan to push out new kinds of features in advance of changing demographics?

Or do you mean that these are the kinds of motivation changes that happen naturally in VWs, and that when the player base's motivations change according to this sequence, the game's feature set should be changed in response?

In other words, does catering to player motivations mean leading players, or following them?

--Flatfingers

27.

Perhaps socialization features are more enduring than content. We all talk about given players the ability to create content for others, but perhaps the focus should be who's around.

Using the bar/club example, there are always new bars and clubs, but there are also those 20-year old corner pubs that are an institution by themselves? Is it because of the age of the pub or because the pub function as a stable meeting point.

I think the volatile center, over time, will move to a more stable position somewhere in the map (call it long-term niche). My live-action RPG has been around 10+ years. We were a spin-off of another game and there are other spin-offs of our game, but we are the steady institution.

28.

Comments about aging MMORPGs... Here are my guesses about a MMORPG's lifespan:

When a MMORPG is brand new it attracts early adopters who want the latest and greatest. Usually this includes graphics, but it could include other new features such as "castle sieges" or "instanced dungeons". I suspect the early adopters will leave after 6-12 months and go to the next latest and greatest MMORPG.

The next crowd to come in are people who (a) want to wait for the MMORPG to stabalize and mature, (b) don't have the latest hardware, and/or (c) want to visit MMORPGs that have withstood the test of time. (If a MMORPG still has a large player base 12 months after release, it must be good, whereas one whose population quickly plummets isn't worth visitng.)

The last stage are those players looking for a bargain. They visit a MMORPG 3+ years after its launch, when it starts dropping prices.


My guestimate about how these fit into Nick Yee's datapoints:

1) A player with strong social ties is likely to stay around longer, so they may last into the next stage. This would cause later stages to have more socializers.

2) Early adopters are more likely to be male, 20-something. (Teenagers with wealthy parents willing to buy the latest computer and $15/month are also a possibility.) Thus, they're more likely to be achievers. Conversely, later adopters are less likely to be achievers.

3) The low-cost players are more likely to be those people that don't play enough to make monthly payments worth their while, or who are teenagers without a credit card and/or $15/month. (Note: GuildWars' pay-once model is likely to attract a disproporionate number of teenagers because it has no monthly payment.) This would imply that low-cost VWs re-attract achievers, since teenagers are more achiever oriented. If so, where does the "don't play very often crowd" fit? Are they more likely to be achievers, immersion, or socializers?

29.
Flatfingers wrote: In other words, does catering to player motivations mean leading players, or following them?

Pretty much what Mike Rozak said: I suspect early adopters to be more achievement oriented. Whether they say "this game is great!" and so attract other players probably depends on the game appealing more to them than it does to the secondary market of people they're talking into joining the game.

However, they master the game sooner than most and leave it for the next big thing. The people remaining are less focused on achievement than they were: So should the game change to accommodate its new playerbase (getting them to stick around longer than they would otherwise, that is)? Perhaps doing that will even get some of the achievers to stick around longer than they would otherwise: They want to leave for the next big new game, but all their friends are now here and don't want to leave (especially not for a game which is more appealing to the achievers and so less appealing to them).

MMORPGS already follow a model something like this: when they first launch they're more difficult than they are years or even months post-launch. Sometimes this is due to things beyond developer's control (there are more fansites with all the answers later, more fellow-players to answer questions in-game later, more twinking in-game and out later on, etc.), but there's also a tendency I think to "ease up" on the players as time goes by: I suspect as a response to player demand.

So maybe it shouldn't be done as a response sort of thing, but planned from the git-go into the development with a recognition that the difficulty level should change as part of the game's life cycle.

As opposed to pretending that we're developing an MMO to last for a billion years, even though we know that's not something that will happen, nor something that we even want to happen.

For commercial MMOs, I mean.

30.

As opposed to pretending that we're developing an MMO to last for a billion years, even though we know that's not something that will happen, nor something that we even want to happen.

Why not? Tolkien has an expiration date?

31.

Why not? Tolkien has an expiration date?

We're more like Sports Illustrated than LotR.

The comments to this entry are closed.