talent spotting

October 4, 2008

The next thing I want to test is how to deconstruct the myriad of talent builds for each class. Presentation is the challenge; lists quickly grow to eye-glazing length as each of the different x/y/z builds gets dragged out of the data. For a start, I’m only going to attempt to analyze level 70 builds. But even there, there are lots of subtle differences – not just in the x:y:z proportions, but also in the points spend at the per-talent level (that is, not everybody invests all 5 points in a 5 point talent).

One thing is clear: not every player goes straight to WoWWiki and picks one of the cookie-cutter builds. There really are a quite a lot of beautiful and unique snowflakes in Azeroth.

The current plan is to build a set of tables that drill down through the more popular builds per class. It seems to me that at least three levels of analysis are required. The first level just identifies the most popular builds across the three trees. Here I’m expecting that there will be a reasonable amount of clustering. For example in the priest class we find the following:

Discipline

Holy

Shadow

41,20,0

3.5%

20,41,0

14.4%

14,0,47

12.7%

42,19,0

3.2%

23,38,0

9.5%

13,0,48

4.3%

43,18,0

2.6%

21,40,0

4.4%

20,0,41

2.7%

47,14,0

1.6%

19,42,0

3.3%

17,0,44

2.1%

48,13,0

1.0%

16,45,0

2.5%

12,0,49

1.4%

45,16,0

0.8%

18,43,0

2.2%

11,0,50

1.4%

28,33,0

2.2%

18,0,43

1.3%

17,44,0

0.6%

0,0,61

1.1%

24,37,0

0.6%

10,0,51

1.1%

26,35,0

0.5%

19,0,42

1.0%

15,0,46

0.8%

(Note that the percentages reference the total number of level 70 priests.) As you can see, build popularity is reasonably concentrated; by putting in a cut-off at, say, 1% popularity, or even a bit higher, the table can be kept small and readable.

The fun starts when we try to work out which specific talents are popular in the build. Not every 20/41/0 priest has spent their points in exactly the same way although, again, a relatively small set of the most popular spreads can be identified. This leads to analysis level two: one table per popular build which shows where the points in that build have been distributed. So, sticking with that example… For the 20/41/0 priest we get something like this:

Talent

Tier Tree

% Take-up

Unbreakable Will

1

Discipline

97%

Wand Specialization

1

Discipline

5%

Improved Power Word: Fortitude

2

Discipline

94%

Improved Power Word: Shield

2

Discipline

34%

Martyrdom

2

Discipline

8%

Silent Resolve

2

Discipline

88%

Absolution

3

Discipline

21%

Inner Focus

3

Discipline

94%

Meditation

3

Discipline

99%

Improved Inner Fire

4

Discipline

2%

Improved Mana Burn

4

Discipline

4%

Mental Agility

4

Discipline

94%

Healing Focus

1

Holy

89%

Holy Specialization

1

Holy

91%

Improved Renew

1

Holy

99%

Divine Fury

2

Holy

95%

Spell Warding

2

Holy

11%

Blessed Recovery

3

Holy

8%

Holy Nova

3

Holy

23%

Inspiration

3

Holy

87%

Holy Reach

4

Holy

75%

Improved Healing

4

Holy

99%

Searing Light

4

Holy

4%

Healing Prayers

5

Holy

39%

Spirit of Redemption

5

Holy

94%

Spiritual Guidance

5

Holy

98%

Spiritual Healing

6

Holy

100%

Surge of Light

6

Holy

1%

Blessed Resilience

7

Holy

4%

Holy Concentration

7

Holy

92%

Lightwell

7

Holy

15%

Empowered Healing

8

Holy

100%

Circle of Healing

9

Holy

99%

This table doesn’t just list talent names; things are more complicated than that. There is a variety of talent spend choices. That is the goal of the “% take up” column. This number shows the proportion of 20/41/0 priests who have invested at least some points in each talent. Again, it becomes clear that there are a lot of left-of-centre choices that only a few people actually make. For example, 5% of priests using this build think that Wand Specialization is an important Discipline tier-one talent. Most clearly don’t.

But there are others where opinion seems to be more divided. 34% think that Improved Power Word: Shield is worth spending some points on. Significant minority opinions like that are worth identifying. Here again, we could just drop the far-outliers; that would reduce the complexity in the tables.

However there is still one level of analysis to go. If we look at the above table we see that in tier one of the Discipline tree, 5% like their Wands and 97% want their Unbreakable Will. So that’s 102% of all 20/41/0 priests covered… phew! The answer is that the points spend across the two talents isn’t clear yet. Some priests have had an each-way bet – putting some points into each of the tier one talents. That’s why the totals don’t add up to 100%.

So we need more columns. What we need is some indication of the most popular points spend in each talent. Some talents are basically cr@p but people invest some points in them, either to get to another talent that they really want, or to sink points so they can move onto the next tier. Others are worth every point. That’s the difference that we want to capture.

Here I must admit to running out of inspiration (for the time being at least). I’m getting the feeling that these reports are going to end up too detailed to really comprehend. For the time being I’m proposing to add one other column to the above table – the average points spend in each talent. For example, the average 20/41/0 priest puts only 4 points into Unbreakable Will, a 5 point talent. So, we end up with a table that will look something like this:

Talent

Tier

Tree

% Take-up

Average spend

Unbreakable Will

1

Discipline

97%

4 of 5

Wand Specialization

1

Discipline

5%

4 of 5

Improved Power Word: Fortitude

2

Discipline

94%

1 of 2

Improved Power Word: Shield

2

Discipline

34%

2 of 3

Martyrdom

2

Discipline

8%

1 of 2

Silent Resolve

2

Discipline

88%

3 of 5

Absolution

3

Discipline

21%

1 of 3

Inner Focus

3

Discipline

94%

1 of 1

Meditation

3

Discipline

99%

2 of 3

Improved Inner Fire

4

Discipline

2%

3 of 3

Improved Mana Burn

4

Discipline

4%

2 of 2

Mental Agility

4

Discipline

94%

4 of 5

Healing Focus

1

Holy

89%

1 of 2

Holy Specialization

1

Holy

91%

4 of 5

Improved Renew

1

Holy

99%

2 of 3

Divine Fury

2

Holy

95%

4 of 5

Spell Warding

2

Holy

11%

4 of 5

Blessed Recovery

3

Holy

8%

2 of 3

Holy Nova

3

Holy

23%

1 of 1

Inspiration

3

Holy

87%

2 of 3

Holy Reach

4

Holy

75%

1 of 2

Improved Healing

4

Holy

99%

2 of 3

Searing Light

4

Holy

4%

1 of 2

Healing Prayers

5

Holy

39%

1 of 2

Spirit of Redemption

5

Holy

94%

1 of 1

Spiritual Guidance

5

Holy

98%

4 of 5

Spiritual Healing

6

Holy

100%

4 of 5

Surge of Light

6

Holy

1%

2 of 2

Blessed Resilience

7

Holy

4%

2 of 3

Holy Concentration

7

Holy

92%

2 of 3

Lightwell

7

Holy

15%

1 of 1

Empowered Healing

8

Holy

100%

4 of 5

Circle of Healing

9

Holy

99%

1 of 1

But I’m not totally convinced about that last column. “Average points spend” doesn’t quite cut it for me. If there is one choice that is reasonably popular and a lot of “noise” choices which are just left-field ones (and that pattern of behaviour would match the pattern we see in the rest of the data) then an average will not necessarily pick that up. Needs more thought here…

Advertisements

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

%d bloggers like this: