[POE] Building Better Builds: Forum Posts, Part i

I've become interested in making my Path of Exile builds better.

One way to do this is to see how other players are creating their builds. Since this is a pretty big topic, I'll first focus on passive skill tree choices by players.  I'm particularly interested in these because many players post guides to their builds - including passive skill trees - on the forums. (Data time!)

I grabbed the 600 most recently updated threads in each class forum and extracted build links on the first page of each of these threads. Because PoE's passive skill tree is constantly changing, I removed builds that are no longer valid with the current skilltree or were less than 80 points. Because sometimes authors post multiple versions of a build, I selected only one build from each thread (the build closest to 100 points). This left me with 914 valid builds.

	Duelist	Marauder	Ranger	Scion	Shadow	Templar	Witch
valid builds	101.0	106.0	144.0	123.0	166.0	112.0	162.0
all builds	707.0	540.0	657.0	732.0	633.0	591.0	615.0
average posts	68.8	79.3	47.7	68.2	49.1	58.8	80.6
average views	25071.7	33006.9	17866.2	18892.1	14617.9	15087.8	32962.1

Curiously, Templar has the least build options to chose from, combined with relatively few posts / views for each thread. I'm curious whether this will lead the developers to consider rebuilding the templar starting area.

Defensive specialization: 
Whether in a hardcore or softcore league, one of the biggest challenges in Path of Exile is simply staying alive. For this section, I'll focus on player choices to get life, energy shield, armor, and evasion. When evaluating life- or energy- based builds, I omitted any build with less than 50 of that stat (because, well, death).

A few observations:

• Life-based builds tend to get around 200 life, with marauder builds getting slightly more life than the other classes. I'm actually surprised at how many builds get less than 200 life, which I considered an absolute necessity. 
• Rangers and shadows tend to focus on evasion, while duelists and marauders focus on armor. 
• Of all the classes, only witch has multiple builds getting a 'reasonable' amount of energy shield for CI / low-life. 

There is, of course, a potential problem with this data. The wide range of life / energy shield values suggests some of these builds simply aren't viable. For example, a large number of builds from new or inexperienced players might tend to undervalue defensive stats. To correct for this, I weighted the data by the number of posts in the thread (on the assumption that high-quality builds are more likely to get attention).

While weighting the data did remove a large number of the low-defense builds, the basic story is unchanged: players tend to get at least 200 life, energy-shield based builds are uncommon, and strength-focused classes get armor while dex-focused classes get evasion. It is notable that a few high-attention builds (like cyclone marauder) tend to dominate the data.

Weapon specialization:
Turning to offense, here I looked at whether a build grabbed a damage specialization node (+damage with a weapon type / spell).

A few features that stand out (to me):

• Wow, caster options are popular (even omitting elemental damage builds).
• Ranger is overwhelmingly the most popular choice for bow builds. The lack of a bow-templar makes me consider running one.
• One-handed builds are more popular than two-handed builds for all classes.
• Swords are slightly more popular than axes, despite the focus on soultaker in the current meta, though this is potentially due to buzzsaw builds. 
• Specializing into wands / staves are remarkably uncommon options (2.6% and 1.4% of builds, respectively). Staves make sense - other than Pillar builds, not many characters go for a staff (they're also devilishly hard to roll). But wands? I seem to remember everyone trying wander early in the league.

Thoughts:
I'm frankly a bit surprised at how few defenses many builds grab, particularly given the common refrain that 200% life is a bare minimum for PoE. I'm equally surprised at how many builds choose to specialize in a weapon, given how that limits purchasing options at endgame.

There are plenty of places I'm considering taking the analysis. I'm definitely going to look at keystones. I'm also interested in some general descriptive-stats for the passive tree (shortest paths, nearest clusters) and how this influences the build meta.

So many fun data adventures to be had.