smarter than the average bear

November 23, 2009

Well only just… I’ve got some results from my attempt to divide up the feral druid population into cats and bears. We started from the fact that there is no “form” tag in the armoury XML – no direct way to count the thing we want to count. The only way to get an insight into this is to find a proxy for each of the forms – something that is in the data which can be used to separate the sheep from the goats, if you’ll pardon the mixed metaphor.

Talents seem to be the obvious choice, so long as there is one talent that bears will take and cats not and another talent that is vice versa. Glyphs are the other possibility. Whatever we choose just has to be i) something that players are highly likely to take  and ii) something that is orthogonal; something that definitely points in one direction for bears and another for cats.

But the basic problem is that there are a lot of um… how to put this politely… there are a lot of left-of-centre specs out there. Talents and glyphs are both less orthogonal than I was hoping for – many specs look a bit bearish and a bit cattish at the same time. And there is a big group that takes none of the talents or glyphs that we want to use.

That’s why I decided not to make the queries very complex – adding more talents or glyphs into the selection criteria¬† just increases the number of toons that fall into the grey area. Also I’ve counted specs and not toons since the original question was related to the number of druids specced for tanking.

Thanks to the commenters who made suggestions on possible talents and glyphs that might fit these criteria. I’ve run two queries against the data.

The first query counts feral druids who have Natural Reaction versus those who have Predatory Instincts. A druid with some points in Natural Reaction and none at all in Predatory Instincts might be a bear; t’other way round for cats. Those with points in neither are marked as “unknown”; those with some points in both are the “could be either” group.

The second query counts druids who have a Glyph of Maul versus those who have either a Glyph of Shred and/or a Glyph of Rip. Equipping Maul but not Shred or Rip indicates bear; Shred or Rip but no Maul indicates cat. Again we have groups with a mix of these glyphs, and, unfortunately, a huge group with none of them.

Anyway this is what we’ve got:

(Patch 3.2.2 data; sample size 16327 level 80 feral druids with 28970 specs).

Talent-based spec count:

  • Bear: 30%
  • Cat : 33%
  • Could be either: 5%
  • Unknown: 31%

Glyph-based spec count:

  • Bear: 18%
  • Cat: 9%
  • Could be either: 15%
  • Unknown: 58%

Frankly I’m still not sure how valid these numbers are, but I hope they provide a bit of insight. The talent-based count may at least provide a low-water-mark indication of the number of bearish specs in there.

About these ads

28 Responses to “smarter than the average bear”

  1. Nelson Says:

    Is one possible conclusion that most druids play both cat and bear?

  2. Andy Says:

    So…first off, I really like what you’re doing and I’ve been following you for a pretty long time, but, what’s up with all the “Unknown”.

    I understand what it means, but why did you make the post when 58% of the glyphs lead you to an unknown on the role and 31% of the talents lead you to an unknown role…seems kinda silly.

    I think Natural Reaction/Predatory Instinct and Maul/Shred was the right choice…maybe people out there have really really weird spec/glyph combinations? Feral Druids are probably the hardest to split up too.

    Again, HUGE fan of the site…just wondering if I was missing something.

  3. fiz Says:

    The people contained in the “huge group with none of them” are likely safe to ignore :)

  4. zardoz Says:

    Hey, it’s better to have datamined and lost than never to have datamined at all… This was just one of my bright ideas that didn’t quite come off.

  5. Alphaman Says:

    What about 3/3 in “Survival of the Fittest” for mining tanking-specced ferals? This is what makes bears uncrittable since leather no longer has defense. Add 3/3 in “Thick Hide” to this as well? What bear wouldn’t want to max his armor?

    Kittys are not as clear-cut. Maybe the lack of SotF and/or lack of “Thick Hide”?

    Druids may just tend to play around with their specs more since we have so many playable options. This might also help explain the lack of glyphs.

  6. zardoz Says:

    Alas, adding more talents – especially Thick Hide – just produces a bigger grey area.

  7. marv Says:

    What about looking at gear?
    Since blizzard pigeonholed Druids to share neck, rings, and trinkets with the plate tanks, these slots generally look distinctly different between bears and kitties.

    Looking for defense/dodge(/block/parry) on say, at least 3 out of the 5 slots should be a pretty good indicator that the character is wearing a bear set.

    While it is very possible that a predominately cat druid is simply logged out in a bear set, the same error could have happen with determining a main spec through talents/glyph what with dual spec in play.

  8. zardoz Says:

    Yes, gear is worth a try. Let me see what I can come up with.

  9. Alaron Says:

    Ah, just came across this. Thanks for working on this, Zar, though this isn’t going to be an easy answer. I’d recommend (and I don’t know how possible this is) building a query that checks for multiple points of differentiation, and assigns each a positive or negative value based on cat/bear applicability. For example, if they’ve talented Infected Wounds, +1 (probably bear); if they’ve talented Predatory Instincts, -1 (probably cat), etc. Add the values, and specs above/below a certain threshold are definitely bears/cats; those in the middle are either hybrids or confused.

    Cats glyph Shred/Rip/Savage Roar. They wear jewelry with crit/haste/AP/ArPen. Their trinkets have haste/crit/AP/ArPen procs. They talent into Predatory Instincts/Natural Shapeshifter(resto). They enchant bracers with Greater Assault (+50 AP), Striking (+38 AP), or Fur Lining: Attack Power for LW (+130 AP)

    Bears glyph three of of Maul/Frenzied Regeneration/Survival Instincts/Growl, though this is less strict since these glyphs are more situational. They wear jewelry with dodge/defense rating. Their trinkets have armor/dodge/hp stats or procs. Unfortunately, new bears will probably use cat jewelry and trinkets, which are much more common as leveling pieces. They talent Thick Hide/Natural Reaction/Infected Wounds/Protector of the Pack. They enchant Bracers with Major Stamina (+40 sta) or Fur Lining: Stamina for LW (+102 sta).

  10. zardoz Says:

    Alaron: thanks for those thoughtful comments. There is no technical obstacle to doing what you suggest. Where the problems arise is deciding what threshold levels to use to put toons in each camp.

    I’ve been experimenting with “score”-type queries for a number of things, not just this question. They always turn out to be problematic. What you generally get is a surprisingly smooth distribution of players across every score level! Now, that statement is a bit of an exaggeration, but you see my point.

    Often you do get some “clumping”. It’s pretty clear that maybe 30% of feral druids have at least one spec designed for some sort of bearish tanking role. But that doesn’t mean that 70% don’t… There is always a big group – maybe another 30% or more – that ends up with a middle score.

    To me that grey area means we’re not going to get a good enough answer on this one, but it’s an interesting problem and I’m still mucking around with different ways of attacking it.

  11. coldbear Says:

    Survival of the Fittest is a MUST-HAVE TALENT for both Cat and Bear. Yes, it is that good for both specs. Alphaman above is regrettably completely off-base and spreading misinformation by posting what he did.

    Your methodology is sound, though I could see you possibly working with the Protector of the Pack talent which is a bit further down in the tree – to avoid hybrid pvp specs that try to go semi-deep in feral and also spend more than 15-16 pts in Resto or Balance.

    What we’ve learned from this is that many druids out there do not bother to learn how to spec and glyph properly, or are not played enough for their owners to bother much about it.

    There’s plenty of druids out there who still cling to the old BC ways of being able to dps and tank with the same spec, which is why you’re seeing all this crazy grey area stuff – those druids will be somewhat effective at both, but certainly not good enough to compete with other tanks or dps in heroic TOGC25 as of 3.2.2.

    Again, thanks for the great work.

    And people, before you start talking about talent this or talent that, please first reference reputable sources such as
    Toskk
    Rawr
    Simulation Craft
    Feral By Night
    Cat DPS for Dummies

    Regarding gear – many ferals carry at least two if not three completely different gearsets nowadays, so might log out in pvp gear.

  12. zardoz Says:

    Thanks. I did run a query with Protector of the Pack in the mix. It didn’t really reduce the “grey area” all that much.

  13. K Says:

    Thanks for the effort Zardoz.

    I think you’ll find problems because cat/bear hybrid specs are actually quite feasible in Wow. Whether it’s deliberate or just poor choices, a spec is probably still viable for a lot of content. Even specs or glyphs that any educated bear/cat would say is totally wrong can still be “good enough” for someone that isn’t min/maximizing.

    As an embarassing, but graphic example, I forgot to change specs and tanked Koralon in a cat spec this weekend. It was a mistake, I felt stupid and apologised, my spec was completely wrong for tanking. But it still worked.

    If a balance or resto druid wants a feral spec that they can use to do dailies or tank 5-mans they can do it without being precise with their spec or glyphs.

  14. Alphaman Says:

    Thank you, Coldbear for the correction (and for not flaming me too badly while doing so:-).

    In rethinking it, it makes perfect sense that all ferals would want SotF and since I don’t go kitty much except for dailies, I totally missed that.

    May need to check out that “Cat DPS for Dummies”…


  15. [...] I’ve pointed this out previously to people curious about WoW’s in-game demographics, but Armory Data Mining is really a fantastic but underappreciated site. To be specific, there are actually two sites of interest here — the actual Armory Data Mining and its related blog. Zardoz, the creator and maintainer of both, uses the former to collect and update statistics on class, race, and sex popularity in WoW (in addition to reports on class battleground performance and profession popularity), while the blog is often used to look at smaller issues or questions like the effort to distinguish between bear and cat specs through the Armory. [...]

  16. Kevin S. Says:

    More interesting than the fact that so many are bears or cats is that so many either don’t know how to do one right, or are trying to do both.

    Perhaps you could try this again only for druids with some hardmode gear (or some other criteria to designate them as ‘serious’) and run it again to see if it divides better.

  17. Myrddin Emrys Says:

    This is old, but I’d like to offer my opinion: Bayesian weighting. This is the exact domain for which the Bayes algorithm was created.

    Take each notable characteristic of the class like talents and glyphs. Add in indicator gear slots like set pieces, trinkets, jewelry, and relics. Then go through and manually classify a bunch of druids… and run Bayes algorithm over the rest and watch it TELL you what the best indicators for bear vs cat are.

  18. Aanye Says:

    Seems like there’s a lot of disagreement about what’s one spec, what’s another. Perhaps you should let the numbers explain the numbers.

    Take a query of all druids that are dual-spec, and that have BOTH specs deep feral. Then make a note of the talents that are in one spec but not another. It could reveal a “tell” between one spec or the other. I would be willing to bet that most dual-ferals are more likely to know what they’re doing than druids that have only a single feral spec.

    Might also consider, for the purposes of investigation, only people with a minimum average item level. Those with an average iLvl of 235-240 or higher are more likely to be specced right (i.e. properly enough to get the gear they’re wearing).

    Still, I think that a lot of feral druids will simply remain “hybrid”. Whether it’s a utility grinding spec, or a lot of people just don’t get feral, I don’t know.

  19. Darush Says:

    Have you tried some more sophisticated clustering techniques? There are very powerful techniques out there which you might be able to use.

  20. Caelean Says:

    Sorry for the late comment on this, but I swap out Glyph of Maul for Glyph of Growl on many raid bosses.

    How about having 2/3 major glyph slots with Growl, Maul, Frenzied Regen, and Survival Instincts?

    Mangle doesn’t indicate either, IMO.

  21. zardoz Says:

    Thanks for all the fresh comments. A couple of general points:

    1. I’m pretty confident that the basic problem is that the specs just don’t strongly tend to cluster around bear/cat poles in the way that I’d hoped. By accident or design, Blizz has produced a system that prevents any simple reverse-engineering attempts.

    2. Adding more factors in, like achievements or raiding behaviour, is worth looking at. I’ll see what I can do.

    3. More powerful stats techniques would help but, unfortunately I don’t have a background in statistics, and can’t easily go beyond the basics. I’m aware of the Bayes theorem but only at the level of the wikipedia article. Anybody who does have these skills and wants to either suggest some things I could read, or to play around with the data themselves, is welcome to leave a comment here and I’ll get in touch by email.

  22. Darush Says:

    Wikipedia for “Clustering” is a good introduction. Furthermore, The Elements of Statistical Learning is an excellent book on the subject that can be downloaded for free:

    http://www-stat.stanford.edu/~tibs/ElemStatLearn/download.html

    Alternatively, is the raw data available somewhere online? (I tried http://armorydatamining.appspot.com/, got a 503 Over Quota error) If it is, I will be more than happy to give it a go myself.

  23. zardoz Says:

    Whoa, our comments crossed in cyberspace. I’ll send you an email now!

  24. Nin Says:

    Another possibility you might want to consider is to look at the raw stats of the character? Assuming you can filter out those with low levels of gear (eg below 232/226 average) then those with Stam > Agi are more like to be bears, and those with Agi > Stam are more like to be cats? Just a thought for a quick query that might provide something to compare your other results with.

  25. Narkondas Says:

    You might want to try to contact the people behind http://elitistarmory.com – they seem to have a pretty good grasp at differentiating between tank and DPS players. Perhaps they would be willing to share their selection criterias.

  26. zardoz Says:

    interesting… thanks


  27. [...] collected statistics for Armory Data Mining, he tried his hardest to figure out a way to count which feral players were decisively bears and which were definitely cats. A guest blogger named Darush later revisited the issue using some [...]


  28. [...] collected statistics for Armory Data Mining, he tried his hardest to figure out a way to count which feral players were decisively bears and which were definitely cats. A guest blogger named Darush later revisited the issue using some [...]


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

Follow

Get every new post delivered to your Inbox.

%d bloggers like this: