ESPORTS FANS ANALYSISJIAYANG LIOctober 25, 2015

LEAGUE OF LEGENDS

AGENDA

PREDICTION SEGMENTATIONBACKGROUND

APPENDIXGENDER

BACKGROUND

SUMMARY

Objective:Analyze general gamers and esports fans to make recommendation on esports marketing strategy and resource allocation.

Analysis:1. Hypothesis Test & Prediction Analysis2. Segmentation Analysis.3. Explore Gender Difference

SUMMARY

VARIABLES

Gamer ID Gender Weekly Online Duration 2015

Average Annual RP PurchasingIP Earned 2015

Registration Time

Rank Proportion

History Best Ranked TierAverage KDA

Clicks Video Clicks Article

TotalClicks

Variable explanations are in the Appendix

VARRIABLES

PREDICTION ANALYSIS

Define esports fans by his total clicks on esports videos and articles.

Methods:Statistic, T-Test, Chi-Square Test

Esports Fans: total clicks > average

General Gamers:total clicks < average

Test the difference between esports fans and general gamers

INDICATOR

Results:Esports fans are performing differently from general gamers on 4 indicators: “rank proportion”, “weekly online duration”, “average RP purchasing” and “registration time”.

RESULTS

21%

Long Time Registration Time > 4 years

Moderate Time2 years < Registration Time < 4 years

Short TimeRegistration Time < 2 years

Esports Fans General Gamer

Two groups are performing differently on “Registration Time”

“Registration Time”

65%

60%

55%

50%

45%

40%

35%

30%

25%

20%

15%

10%

5%

0

24%

55%

20%

37%

43%

P-value < 0.05 Indicator

EXAMPLE

50%

45%

40%

35%

30%

25%

20%

15%

10%

5%

0

4%

15%17%

25%

Challenger Diamond Platinum

Esports Fans General Gamer

“History Best Tier”

Two groups are performing similarly on “History Best Tier”

13%

26%

2%

19%

15%

30%

8%

26%

Gold

Silver Bronze

EXAMPLE

P-value > 0.05 Indicator

Build Predictive Model

MODELING 1) Which game behaviors are significantly correlated to esports behaviors.

2) How these game behaviors influence gamers’ clicks on esports material.

To Find:

Methods:Regression ModelingModel Validation

Rank Proportion

Tota

l Clic

ks

Optimal Model:Total Clicks = 111.18 R + 0.29 D

*R: rank proportion*D: weekly online duration 2015

MODELING

The more frequently a gamer plays rank, the more he pays attention to esports

1.

The more time a gamer spends on LOL weekly, the more he pays attention to esports

INSIGHTS

2.

Average RP Purchasing is not correlated to whether he is an esports fan 3.

4. Most non-esports fans’ average RP purchasing are low

The x-axis is gamer ID and the y-axis is average annual RP purchasing

The color represents whether he is an esports fan (blue = Yes, orange = No)

INSIGHTS

Registration time not correlated to whether he is an esports fan 5.

6. Most esports fans are new gamers

The x-axis is gamer ID and the y-axis is registration time

The color represents whether he is an esports fan (blue = Yes, orange = No)

INSIGHTS

Pick gamers who play rank frequently or play LOL for longer than weekly average time as survey samples. Without further analysis, they are most likely to be esports fans.

Sampling Method:

Research on insight 6. If research shows long-time gamers are tired of LCK dominating, we should hold more cross-league tournaments to improve communication in tactics and market more on non-league genres like All Star.

Engagement:

Insight 4 may indicate a causal relationship between esports and RP purchase. Conduct A/B testing to evaluate our contribution on ARPU and find out the reason behind to take advantage of it.

ARPU:

RECOMMENDATION

Prediction Analysis could be applied to various aspects.

such as:Predict channel performance among Game Log-in interface, Youtube and TwitchTV. Allocate budget and resources accordingly.

In reality, we need to consider more factors than this projects, such as: How long they spend on each platform after the click (bounce rate) Their behaviors after watching videos (share to social media)

EXTENSION

SEGMENTATION ANALYSIS

Group A

Group B

Group C

6%

7%

28%

Divide esports fans into groups according to their similarities on behaviors. (process in Apendix)

>

There are 3 groups in esports fans, takes a total 41% of sample gamers

>

Method:K-Mean Clustering, Cluster Plot

SEGMENTATION

Population Proportion

GROUP A

GROUP B

GROUP C

GROUP A:

Most Valuable Segment

GROUP B: GROUP C:

PROFILES

They are still learning about the game and they are curious about esports. They watch lots of videos to improve skills and spend much money in game too.

Losing Segment

They are master gamers. They no longer need to spend money in game. They still watch videos but they are losing passion to esports.

Bottleneck Segment

They play LOL a lot but lack of tactics. They are familiar with esports but didn’t get much fun yet

GROUP A:

Most Valuable Segment

GROUP B: GROUP C:

STRATEGY

Made special videos on pro gamers and teams to deepen their interest.

Cooperate with popular commentators online to market esports via entry-level unofficial channels

Losing Segment

Stimulate their intersts with new changes.(ex. encourage pros to deepen champion pools in game)

Help weak leagues to grow to make the pro games more exciting

Bottleneck Segment

Hold special “Recap Show” after Live, with detailed champion & tactics commentary to help them understand the fun part of esports

Segmentation could be applied to various aspects.

Such as:Segmentation on Riot social media followers on different platformsSegmentation on esports merchandise buyers

Use Segmentation before Prediction Analysis to get a more accurate prediction model.

EXTENSION

GENDER

GENDERTest whether male gamers and female gamers have different taste on choosing esports materials: videos or articles

Methods:

T-Test

Article Clicks

Female GamerMale Gamer

Video ClicksThere is no significant difference between male gamers and female gamers on the taste of choosing videos or articles

RESULT

Gender Test could also test other taste differences

Such as differences on:

Favorite teams/pros, commonly used social media platformtaste on esports merchandise, trigger to become esports fans

There are many demographic groups we can analyze too, such as: different leagues, countries, age groups.

EXTENSION

APPENDIX

Data Analytics All the data and analysis are processed by R

Code Script & Output:

http://rpubs.com/jli101/120496

Variables

Gamer ID Gamer Account Name (unique)

Gender Gamer Gender

Weekly Online_Duration_2015Each gamer’s average weekly log-in duration in 2015. A good indicator for whether he is an active user in 2015.

Average_Annual_RP_Purchasing

Average annual dollar spend on purchasing RP. A good indicator for purchasing power and ARPU. Also minimize the aggregate effect due to registration duration

IP_Earned_2015Total IP earned in 2015. A good indicator for whether he is an active user in 2015 and game skills.

Registration_TimeFor how many years he has been registrated as a summoner in League of Legend

Rank_Proportion The proportion of playing rank out of all the games he plays

Hitory_Best_Ranked_Tier The best ranked tier he has ever get. A good indicator for game skills

Average KDA Average KDA. A good indicator for game skills

Clicks_VideoHow many times he clicks on the esports videos on LOL welcome interface during the "Worlds"

Clicks_ArticleHow many times he clicks on the esports articles of LOL welcome interface during the "Worlds"

Total_ClicksSum of clicks_video and clicks_article. A good indicator for whether he is a heavy esports fans

Clustering Plot

THANK YOUFor Your Watching and Everything