Analysis of World of Warcraft PvP Leaderboards

This is a draft version of the post but I am publishing it now because the bootcamp is ending. I will keep working on it until it is done.

Note to self: don't forget to check final version against project 1 rubric and report template

If you combine your problem statement, executive summary, data dictionary, and conclusions/recommendations, you have an amazing README.md file that quickly aligns your audience to the contents of your project. Don't forget to cite your data sources!

Some recommendations on plotting: * Plots have titles * Plots have axis labels * Plots have appropriate tick labels * All text is legible in a plot * Plots demonstrate meaningful and valid relationships * Plots are interpreted to aid understanding

Introduction

**problem statement**
**executive summary**

data acquisition

connect to db, list tables
SQL query: battlegrounds & players
**data dictionary**
**data sources**

measures of best player

rating
rank
win ratio

rating vs rank
win ratio vs rank
num matches vs win ratio
wins vs num matches
ratings vs rank grouped by leaderboard

histograms, populations

rank histogram
ilvl histogram
achievement histogram

player class bar chart
player class bar chart stacked factions

ratings histogram for all leaderboards

exploring correlations

pair plot
heat map

ilvl vs rank
ilvl vs rating
achievement vs rank
rating histogram
win ratio bar chart by faction

Groupings

ilvl vs rank grouped by leaderboard
ilvl vs rank grouped by player class
win ratio vs rank grouped by player class
win ratio vs rank grouped by faction

mean rating bar chart by class

mean rank bar chart by class for all leaderboards

Descriptive and Inferential Statistics

measures of central tendency, spread, and shape/skewness
For each variable in your data, summarize the underlying distributions
Be sure to back up these summaries with statistics.

plot and interpret boxplots

summary, conclusions, etc

best class to play?
best PvP type to play? (2v2 3v3 battlegrounds)
recommendations for climbing leaderboards
does the hoarde receive preferential treatment from blizzard or not?

I recently scraped the PvP leaderboards from the World of Warcraft website and wrote them out to a SQLite database. Let's plot some charts of the data and see if anything interesting turns up.

There are three leaderboards: the 2v2 Arena, the 3v3 Arena, and the 10x10 Battleground. Each leaderboard lists the top 1000 players by rating. (I think these are Elo ratings.) Anyway, I saved the leaderboards to a SQL database as three separate tables. I also scraped a minimal amount of data from the profile pages of each of the characters on the leaderboards and saved the profile data to a fourth table.

import sqlite3
import altair as alt
import pandas as pd

query = '''
SELECT *
FROM sqlite_master
WHERE type='table'
'''
con = sqlite3.connect('data/wow.db')
pd.read_sql(query, con)

	type	name	tbl_name	rootpage	sql
0	table	arena_2v2	arena_2v2	2	CREATE TABLE "arena_2v2" (\n"rank" INTEGER,\n ...
1	table	arena_3v3	arena_3v3	32	CREATE TABLE "arena_3v3" (\n"rank" INTEGER,\n ...
2	table	battlegrounds	battlegrounds	62	CREATE TABLE "battlegrounds" (\n"rank" INTEGER...
3	table	players	players	93	CREATE TABLE "players" (\n"name" TEXT,\n "tit...

query = '''
    SELECT name, rank, rating, wins, losses, achievement, ilvl
    FROM battlegrounds
    LEFT JOIN players
    ON battlegrounds.url = players.URL
'''
con = sqlite3.connect('data/wow.db')
df = pd.read_sql(query, con)
con.close()
df

	name	rank	rating	wins	losses	achievement	ilvl
0	Crdefender	1	2163	81	4	18325.0	461.0
1	Lifeswaplol	1	2163	80	3	9070.0	464.0
2	Wolf	3	2153	77	4	18205.0	466.0
3	Intricate	4	2135	72	4	10565.0	462.0
4	Jøkes	5	2125	69	2	15570.0	462.0
...	...	...	...	...	...	...	...
995	None	995	1443	9	14	NaN	NaN
996	None	995	1443	10	12	NaN	NaN
997	None	995	1443	8	3	NaN	NaN
998	None	995	1443	12	14	NaN	NaN
999	None	1000	1442	15	12	NaN	NaN

1000 rows × 7 columns

df['num_matches'] =  df['wins'] + df['losses']
df['win_ratio'] = df['wins'] / df['num_matches']

alt.Chart(df).mark_point().encode(x='rank', y='rating')

alt.Chart(df).mark_point().encode(x='rank', y='ilvl')

alt.Chart(df).mark_point().encode(alt.X('rating:Q', scale=alt.Scale(zero=False)),
                                  alt.Y('ilvl:Q', scale=alt.Scale(zero=False)))

alt.Chart(df).mark_point().encode(x='rank', y='achievement')

alt.Chart(df).mark_point().encode(x='rank', y='win_ratio')

alt.Chart(df).mark_point().encode(alt.X('win_ratio:Q', scale=alt.Scale(zero=False)), y='num_matches')

alt.Chart(df).mark_point().encode(alt.Y('win_ratio:Q', scale=alt.Scale(zero=False)), x='num_matches')

alt.Chart(df).mark_point().encode(x='num_matches', y='wins')

alt.Chart(df).mark_bar().encode(alt.X("rating:Q", bin=True), y='count()')

alt.Chart(df).mark_bar().encode(alt.X("rank:Q", bin=True), y='count()')

alt.Chart(df).mark_bar().encode(alt.X("ilvl:Q", bin=True), y='count()')

alt.Chart(df).mark_bar().encode(alt.X("achievement:Q", bin=True), y='count()')

alt.Chart(df).mark_circle().encode(
    alt.X(alt.repeat("column"), type='quantitative', scale=alt.Scale(zero=False)),
    alt.Y(alt.repeat("row"), type='quantitative', scale=alt.Scale(zero=False))
).properties(
    width=100,
    height=100
).repeat(
    row=['rank', 'rating', 'win_ratio', 'ilvl', 'achievement'],
    column=['rank', 'rating', 'win_ratio', 'ilvl', 'achievement']
)

query = '''
SELECT board, rank, rating, name, title, realm, class, details,
    faction, wins, losses, achievement, ilvl, players.url
FROM (
    SELECT *, '2v2 arenas' as board
    FROM arena_2v2
    UNION
    SELECT *, '3v3 arenas' as board
    FROM arena_3v3
    UNION
    SELECT *, 'battlegrounds' as board
    FROM battlegrounds
) leaderboards
JOIN players
WHERE players.url = leaderboards.url
'''
con = sqlite3.connect('data/wow.db')
df_boards = pd.read_sql(query, con)
con.close()
df_boards

	board	rank	rating	name	title	realm	class	details	faction	wins	losses	achievement	ilvl	url
0	2v2 arenas	4	2514	Niarb	Notorious Gladiator	Laughing Skull	Paladin	120 Human Holy Paladin	ALLIANCE	98	42	13215.0	469.0	https://worldofwarcraft.com/en-us/character/us...
1	2v2 arenas	2	2591	Thugonomiczz	Wrathful Gladiator	Stormrage	Warlock	120 Dwarf Destruction Warlock	ALLIANCE	95	9	15410.0	473.0	https://worldofwarcraft.com/en-us/character/us...
2	2v2 arenas	15	2411	Nessper	Notorious Gladiator	Bleeding Hollow	Rogue	120 Blood Elf Assassination Rogue	HORDE	145	61	19680.0	469.0	https://worldofwarcraft.com/en-us/character/us...
3	2v2 arenas	33	2372	Tickix	None	Bleeding Hollow	Priest	120 Undead Discipline Priest	HORDE	119	60	11430.0	470.0	https://worldofwarcraft.com/en-us/character/us...
4	3v3 arenas	229	2058	Tickix	None	Bleeding Hollow	Priest	120 Undead Discipline Priest	HORDE	67	39	11430.0	470.0	https://worldofwarcraft.com/en-us/character/us...
...	...	...	...	...	...	...	...	...	...	...	...	...	...	...
1969	battlegrounds	626	1538	Landsharkzx	Grand Marshal	Bleeding Hollow	Druid	120 Night Elf Guardian Druid	ALLIANCE	10	5	12000.0	448.0	https://worldofwarcraft.com/en-us/character/us...
1970	battlegrounds	503	1568	Kelzaru	Arena Master	Emerald Dream	Druid	120 Zandalari Troll Balance Druid	HORDE	15	11	12565.0	455.0	https://worldofwarcraft.com/en-us/character/us...
1971	battlegrounds	634	1536	Miamigodxo	Firelord	Bleeding Hollow	Rogue	120 Night Elf Assassination Rogue	ALLIANCE	22	24	10805.0	448.0	https://worldofwarcraft.com/en-us/character/us...
1972	battlegrounds	422	1598	Abrastal	Highlord	Gorgonnash	Paladin	120 Human Holy Paladin	ALLIANCE	9	6	20055.0	463.0	https://worldofwarcraft.com/en-us/character/us...
1973	battlegrounds	422	1598	Intdruid	the Fabulous	Kel'Thuzad	Druid	120 Night Elf Guardian Druid	ALLIANCE	9	6	19620.0	445.0	https://worldofwarcraft.com/en-us/character/us...

1974 rows × 14 columns

alt.Chart(df_boards).mark_point().encode(alt.X('rank:Q', scale=alt.Scale(zero=False)),
                                  alt.Y('rating:Q', scale=alt.Scale(zero=False)),
                                 color='board')

alt.Chart(df_boards).mark_point(opacity=0.75).encode(alt.X('rank:Q', scale=alt.Scale(zero=False)),
                                  alt.Y('ilvl:Q', scale=alt.Scale(zero=False)),
                                 color='board')

alt.Chart(df_boards).mark_point(opacity=0.75, clip=True).encode(alt.X('rank:Q', scale=alt.Scale(zero=False)),
                                  alt.Y('ilvl:Q', scale=alt.Scale(domain=(420, 480))),
                                 color='class')

df_boards['num_matches'] =  df_boards['wins'] + df_boards['losses']
df_boards['win_ratio'] = df_boards['wins'] / df_boards['num_matches']

alt.Chart(df_boards).mark_circle().encode(alt.X('rank:Q', scale=alt.Scale(zero=False)),
                                  alt.Y('win_ratio:Q', scale=alt.Scale(zero=False)), color='class')

alt.Chart(df_boards).mark_circle().encode(alt.X('rank:Q', scale=alt.Scale(zero=False)),
                                  alt.Y('win_ratio:Q', scale=alt.Scale(zero=False)),
                                 color='faction')

alt.Chart(df_boards).mark_bar().encode(x='class', y='count()')

alt.Chart(df_boards).mark_bar().encode(
    x='class',
    y='count()',
    color='faction'
)
# how to get these unstacked? column='faction' makes two plots

alt.Chart(df_boards).mark_bar().encode(
    x='faction',
    y='win_ratio',
)

bar = alt.Chart(df_boards).mark_bar().encode(
    alt.Y('mean(rating)', scale=alt.Scale(zero=False)),
    x='class'
)
rule = alt.Chart(df_boards).mark_rule(color='red').encode(
    y='mean(rating)'
)
(bar + rule).properties(width=500)

df_boards['rank2'] = 1000 - df_boards['rank']

alt.Chart(df_boards).mark_bar().encode(
    alt.X('class'),
    alt.Y('mean(rank2)', scale=alt.Scale(zero=False)),
    alt.Color('board'),
    alt.Column('board')
)

alt.Chart(df_boards).transform_fold(
    ['2v2 arenas', '3v3 arenas', 'battlegrounds'],
    as_=['Leaderboard', '# Characters']
).mark_area(
    opacity=0.5,
    interpolate='step'
).encode(
    alt.X('rating:Q', bin=alt.Bin(maxbins=100)),
    alt.Y('count()', stack=None),
    alt.Color('board:N')
)

mlehotay