Guild Wars 2

I’ve noticed that there seems to be some forums that are unhappier than others, so I conducted an investigation to see whether or not this was true.

So, what did I find out? I’ll put a summary up first but there’s a lot more information below for those who are interested!
But please note that forums were ranked on a 1-5 scale, with 1 being rather unhappy and 5 being rather happy. So a score of 3.2 means that forum poster was slightly happier than neutral.

And by the way, I am just finishing my second year of a Statistics degree so I do (more-or-less) know what I’m doing :-)

SUMMARY

The overall forum happiness score is 2.76 (slightly unhappy).

The Profession forum was the happiest.
The Features forum was the unhappiest.

The happiest Profession was the Thief. The least happy profession were the Necromancers and Rangers.
The happiest Support forum was the Mac Support, the least happy was the Account Issues.
The happiest Races were the Charr and Sylvari. The least happy race was the Human.
WvW was slightly happier than SPvP.
The happiest Feature forum was the Lore, the least happy Feature forum was the Personal Story.

The happiest forum sub-section was: the Thief.
The unhappiest forum sub-section overall: tie on Personal Story and the Black Lion Trading Post.

METHOD

To collect my data I used a random number generator to randomly pick 25 samples from each Forum Sub-section. I only sampled the FIRST post in a thread, i.e. the person who created that thread. I would then judge the happiness of the creator of the thread and rank it on a scale from 1 to 5 (1 being rather unhappy, 3 neutral, and 5 being rather happy). This was largely done based on my judgement, so other people might have had different views.
To allocate the happiness score I would:
1. Consider the overall tone of the thread
2. Look for word indicators
3. Look for faces (happy, sad)

Forums that scored a 1: generally included words and phrases such as “hate” “really frustrating” kittenwtf” and “makes me want to quit right now”. Also included people that issued orders in an angry manner (“I demand it be fixed RIGHT NOW”).
Forums that scored a 2: these people were not angry, but were moderately upset. Word indicators such as “annoying” “it’s sad that…”, and sad faces indicated a 2 score.

Forums that scored a 3: these people were neutral, and their post had no words or tone indicating an emotional investment. These were generally simple questions (“how do I do this” or “what does this mean”). I also allocated 3’s to people who said “I like this…” but also “I don’t like that…”

Forums that scored a 4: these people were moderately happy. Common features in a 4 post was people posting smiley faces, including words such as “I really like…” 4’s were also allocated to posts that existed purely to try and help people (guides, etc). 4’s were also posts that set a happy tone by saying “Hey guys!” and “Thanks so much!”

Forums that scored a 5: generally involved people praising the game or Arena Net, or involving words like “I love…”

STATISTICS MATTERS
Skip past this if you’re not interested ;-)

1. What is a p-value?
In our case, the p-value describes the probability of our mean happiness score data popping out, assuming that all the forums have the same happiness score. In order to have a statistically significant result, the p-value has to be equal to or less than 0.05, meaning that our data is only expected to pop out less than or equal to 5% of the time (therefore highly unlikely).

2. The Independence Assumption
The independence assumption is the big kahuna in linear models. It states that your observations cannot influence each other. In our case, it means that one forum post cannot influence other forum posts. As you can probably see, this may not be the case. A person might read a forum thread, and then under its influence decide to write a new forum thread. On top of this, forum threads are linked through time, and influenced by changing matters in the game, so a forum thread 10 pages ago might be under quite different influences that a forum thread on page 1. In fact, the more I think about the independence assumption the more worried I am about actually fitting a linear model, haha. I might be violating a vital law of Statistics here…
PLEASE DO NOT PUT ANY WEIGHT ON MY ANOVA RESULTS DUE TO MY DOUBT ABOUT THE INDEPENDENCE ASSUMPTION.

3. The Equality of Variance Assumption
This is the second-most important assumption. Remember those box-and-whisker plots you did back at school? Well, each forum section has a box and whisker plot associated with it, and you put all those plots side by side. The independence assumption basically requires that those plots aren’t too crazy; with some being way higher than the others, or some way wider than the others, etc. I did not have any problems with this assumption in any of my analyses.

4. The Normality Assumption
This assumption requires that your data fits the shape of a bell-curve reasonably well. Some of my data did not conform with this one very well, but it is the least important assumption, thanks to the wonderful thing called the Central Limit Theorem (which I won’t go into but just know that it’s wonderful).

DATA FROM SUPPORT FORUMS

Support Forums included: Account Issues, Tech Support, Mac Support, Game Bugs, Forum Web Bugs, Suggestions, Players Helping Players and Discussion.

Overall Mean Happiness Score: 2.82
Forum Ranking (Happiest to Unhappiest):
1. Mac Support: 3.26
2. Forum Web Bugs: 3.04
3. Suggestions: 2.83
4. Tech Support: 2.83
5. Game Bugs: 2.73
6. Players Helping Players: 2.69
7. General Discussion: 2.65
8. Account Issues: 2.56

Statistically significant evidence:
1. We have strong evidence to believe that there are differences in HS in the Support Forums (p-value=0.022).
2. We have strong evidence to believe that Mac Support has a higher HS than Account Issues (p-value=0.020).
Mac Support has a mean HS between 0.06 and 1.33 higher than the Account Issues forum.

DATA FROM FEATURES FORUMS

Features forums included: Crafting, Dynamic Events, Lore, Personal Story, Dungeons, Audio, and the Black Lion Trading Post.

Overall Mean Happiness Score: 2.64
Forum Ranking (happiest to unhappiest)
1. Lore: 3.16
2. Audio: 3.12
3. Crafting: 2.95
4. Dynamic Events: 2.45
5. Dungeons: 2.42
6. Black Lion Trading Post: 2.12
7. Personal Story 2.12

Statistically significant evidence:
1. We have very strong evidence that differences exist in the features forums happiness scores (p-value?0).
2. We have evidence to believe that Audio has a higher mean HS than the Black Lion Trading Post (p-value =0.028).
We estimated that Audio has a HS between 0.05 and 1.70 higher than the Black Lion Trading Post.
3. We have strong evidence to believe that the Audio forum has a higher mean HS than the Personal Story forum (p-value=0.006).
We estimate that Audio has a HS between 0.18 and 1.82 HS higher than the Personal Story forum.
4. We have evidence that the Lore forum has a higher HS than the Black Lion Trading Post forum (p-value=0.018).
We estimate that the Lore forum has a HS between 0.09 and 1.73 higher than the Black Lion Trading Post forum.
5. We have evidence to believe that the Crafting forum has a higher HS than the Personal Story forum (p-value=0.044).
We estimate that the Crafting forum has a HS between 0.01 and 1.65 higher than the Personal Story forum.
6. We have strong evidence that the Lore form has a higher HS than the Personal Story Forum (p-value=0.004).
We estimate that the Lore forum has a HS between 0.22 and 1.86 higher than the Personal Story Forum.

DATA FROM PVP FORUMS

These forums contained the Structured PvP and WvW sub-sections.

Overall Mean Happiness Score: 2.72
Forum Ranking:
1. WvW: 2.79
2. SPvP: 2.67

(I didn’t even bother running a model as it was obvious there wouldn’t be anything going on here).

DATA FROM RACE FORUMS

These forums contained the Asura, Human, Charr, Sylvari and Norn.

Overall Mean Happiness Score: 2.79
Forum Ranking:
1. Charr & Sylvari: 2.875
2. Norn & Asura: 2.75
3. Human: 2.71

(Again I didn’t bother testing out a model since it was apparent there would be no significant differences).

DATA FROM PROFESSION FORUMS

These forums were about the Elementalist, Engineer, Guardian, Mesmer, Necromancer, Ranger, Thief and Warrior.

Overall Mean Happiness Score: 2.85
Forum Ranking:
1. Thief: 3.29
2. Engineer: 3.12
3. Elementalist: 3.04
4. Guardian: 3.00
5. Warrior: 2.75
6. Mesmer: 2.66
7. Necromancer & Ranger: 2.50

Statistically Significant Evidence:
1. We have very strong evidence to believe there are differences in happiness score for the different professions (p-value=0.00).
2. We have strong evidence that Thieves have a higher mean HS than both Necromancers and Rangers (p-value=0.01).
We estimate that Thieves have a HS between 0.11 and 1.48 higher than Necromancers and Rangers.

Hope you found it as interesting as I did! :-)

Not sure what conclusions to draw from this (why are people happy with ele but not ranger when ele is not viable and ranger is very powerful and easy…?) but it is interesting anyway! Especially GD being so far down…

Interesting statistics! I would have thought the Dungeon forum section would have been the most unhappy considering how much they’ve been complaining about some broken mechanics.

Also surprised about Ranger. I have a lot of fun with mine but I also don’t pvp.

Karina.9871:

Hope you found it as interesting as I did! :-)

I did! I’m a data junkie and an engineer with a background in signal processing (so statistics lite).

I do think that there are issues with independence among the posts though, and I probably would have omitted the bugs/problems forums entirely (especially because you were looking at thread initiators and only the first post). The population of thread initiators in those forums is heavily skewed toward unhappy.

As far as untangling how unhappy people beget more unhappy people on forums, I think you’d need a team of stats and psych students to tackle it. Maybe some anthropology majors too. Might be a really interesting project though.

Yes I was surprised with ranger too! But the majority of the unhappy posts there came from people complaining about pet mechanics/AI.

And yes, Pandemoniac I agree with that. It’s just a little project I undertook… and by no means should be relied on! If I had a stats and psychology team with me it would be awesome ;-)

And this is what I love about MMO communities

Interesting, though either your scale is pretty lenient or GW2’s forums are overwhelmingly positive compared to other MMOs (just looking at it and giving it an eye test, GW2’s forum seems as negative as any other).

I dunno, I’m kind of expecting one of these days for an MMO developer to just not have ways for its playerbase to gather and discuss the game, because negativity in the community tends to feed on itself.

That’ll be a sad, sad day, but if I’m looking at it from a business perspective, I can’t help but wonder why developers allow players to convince each other that the game sucks via the forums.

Yeah, something that I noticed was that although the first post was more-often-than-not positive (and I only sampled the first post leading to what you think is a lenient scale perhaps), the thread tend to deteriorate as it went on. Might be cause for another investigation in future…

Quantification of subjective emotions.

I love it.

Does that happy face mean raise this forum’s happiness score? If so…

:)

Haha! Definitely a 5 rating for that post Vorch

Corian.4068:

Interesting, though either your scale is pretty lenient or GW2’s forums are overwhelmingly positive compared to other MMOs (just looking at it and giving it an eye test, GW2’s forum seems as negative as any other).

Actually I disagree. I think the forums are actually more positive than other forums I’ve participated in for games with similarly sized player bases. Of course Diablo 3 has tainted my perception of “negative” forever, but in general I think that there is a higher proportion of threads with fairly calm discussion in them (even if it’s not complimentary to the game itself) than raging hateful threads that I see other places.

Of course that is probably more due to the crack team of commando moderators we have around here than anything else

Karina’s results pretty much matched my perception, although I did find the negativity in the ranger forum surprising. However, there is a fairly vocal group that want to play rangers without pets though, so maybe I shouldn’t be too surprised.

I think game forums contain generally a lot more unhappy people than happy ones.
Most happy people are playing the game and only come here when they meet some issue, so they are counted as unhappy even if they are not.
Removing the forums related to bugs/issues is mandatory to get a more realistic result.

Furthermore, some nationalities don’t post in the official forums much.
I.E. I read an Italian forum where the userbase is extremely happy about the game and it counts roughly 2000 people.
You can count at most 2-3 complainers, but they are not even remotely comparable to the complainers here in terms of maturity; they complain bugs etc, not demand a wow clone.

There is also to consider some people use forums in general as a way to rant and vent frustration without consequences.

Due to the aforementioned facts, I don’t consider forum “happiness” to be of any value.

You need a FAR larger sample size. I can guarantee the Elementalist board is FAR unhappier than the Ranger board!

Fair enough guys. I would love to have had more data, but as it was taking 25 samples from each of the 30 sections amounted to 750 observations! So considering that I think I did quite well :P

And yes perhaps I should have removed the bugs forum etc. It probably didn’t have very much influence in the grand scheme of things though.