ELECTRONIC SPORTS A new phenomena catalyzed by social TVs and a growing social community Mehdi Kaytoue (INSA de Lyon/LIRIS) Loic Cerf, Wagner Meira Jr. (Universidade Federal de Minas Gerais) Chedy Raïssi - INRIA
Digitization of the society Services (the world of e-stuffs) Information, news access Buying, booking Encyclopedia Social relationships/interactions Content access consumption (video, music, pictures) Many examples: forget about the paper, consume digitized Environment? Money! Access anywhere, anytime Smart phones PC Tablets Extended TV services Consoles
Entertainment Movies, music, sport, etc. content and main actors followed on the Web Video games : A flourishing industry Assassin s creed (2012): 40 millions euros budget, expected revenue of 300 millions Angry birds: 600 millions of active players across the world Gaming communities appear Some games require skills to master at Competing spirit of (most of) games: one goal, to win As such, tournaments start to spread
Electronic sport 10 years old in South Korea Spreading to Americas & Europe strongly since 2010 Organizations, league Teams composed of professional gamers, coaches More and more championships More and more software and hardware sponsor to target an audience of 18-35 s males Gamescom 2011, 2012: Dota competitions with $1 millions for the 5- players winning team HuskyStarcraft Youtube channel, E-Sport commentator: 345 millions views, 685k subscribers
5 E-sport is professional
What media to reach a growing audience?
Social TVs: Supply and Demand meet A way for remote viewers (in space and time) to socially interact about a video content Device/network (smartphones, pc, etc) Synchronization Modality (chat, voice) Social reach (family, friends, strangers) Important upheaval of human interactions and socialization Watching TV tends to be more and more active (Crazy society!) Yahoo! and the Nielsen company: 86% of mobiles Internet users (and 92% of the 13-24 s) simultaneously watch TV and use the mobile phone
Following tournaments, leagues
Following gamers directly
Objectives Assess, understand, characterize the phenomena through its social Web community Evaluate, warn about its potential For industrials For researchers : a rich source of datasets, available (for now ) M. Kaytoue, L. Cerf, W. Meira Jr., A. Silva, C. Raïssi. Watch me playing, I am a professional: a first study on video game live streaming. Mining Social Networks Dynamics (MSND@WWW12). M. Kaytoue, L. Cerf, W. Meira Jr. C Raïssi. Witnessing the digitization of sport through social TVs. (Submitted).
Data
Social TV: Entities of interest Content/Object Video content broadcast in live, also called (video) stream Surrounded with a chat tool Content producer Broadcasters, also denoted as streamers Produce the stream Content consumer Watchers, spectators, or viewers Consume the stream Unregistered users: simply watch the stream Registered users: watch, chat, favorites, share
Dataset 1: Channels and their audience An insight to video game channels, their activity, the topics that they cover, and their audience of any user (channel, date, audience, topic, category, description, )
Dataset 2: Users IRC signals An insight to consumer activities, habits, behaviors of registered users (user, channel, date, log in/out) (user, channel, date, message)
Dataset 3: Users favorite channels An insight into watchers interest outside the gaming world (user, channel, channel-category,channel-cumulated-audience,...)
Characterization
Be serious, people really watch? 500000 400000 Chat signals Login signals Logout signals Signal count 300000 200000 100000 0 06/10 13/10 20/10 27/10 03/11
Really, really? On a daily basis? Average Signal Count 30000 25000 20000 15000 10000 Chat signals Login signals Logout signals 5000 0 Sun Mon Tue Wed Thu Fri Sat
Let me guess who produce the content?
Typical Web content producers? 1 1 agregate views 0.8 0.6 0.4 0.2 agregate views 0.8 0.6 0.4 0.2 cumulative (%) 0 0 10 20 30 40 50 60 70 80 90 100 stream rank 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 10 0 10 1 10 2 10 3 10 4 10 5 duration (min) cumulative (%) 0 0 10 20 30 40 50 60 70 80 90 100 streamer rank 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 10 0 10 1 10 2 10 3 10 4 10 5 10 6 duration (min)
Watching, a daily habit? Cumulative (%) 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 1 10 100 Duration (minutes)
What do they watch?
What do they watch?
Language, a frontier? Channels Language Proportion English 89.20 % French 3.28 % Chinese 2.34 % Spanish 1.40 % Japanese 0.93 % Telugu 0.47 % German 0.47 % Russian 0.47 % Polish 0.47 % Watchers Language Proportion English 94.20 % Chinese 1.32 % French 0.85 % Japanese 0.72 % Telugu 0.71 % Portuguese 0.53 % Spanish 0.46 % Polish 0.28 % Russian 0.16 %
Who are they? number of users 1e+06 100000 10000 1000 100 animals science_tech news sports entertainment social gaming 10 1 100 1000 10000 100000 1e+06 per-category ranking
Who are they?
What do they speak about?
Summing up briefly E-Sport: an important phenomena at its start Social TVs: wherenthe supply and the demand meet, dual catalyzers (Twitch.tv is an important E-Sport sponsor) Addicts gather up around social TV to share/interact about video games Young people (with their classical interests ) Watching e-sport, question, enjoy/feel Discovering games before buying them Produced by USA (Europe follows) Daily activity/week-end activity Language barrier
Rich and accessible sources of data Twitch API Prolific community on Twitter/Facebook Specific game data Official rankings Games logs Many opportunities as application for your algorithms Graph mining Recommender systems Understanding the data through a characterization allow to play the role of the expert in the application
30 Example of acquired chat session #steven_bonnell_ii> mutagling joins #steven_bonnell_ii> harvardmethaddict: AF server > EU #steven_bonnell_ii> juno1990ahn joins #steven_bonnell_ii> dadgun603 parts #steven_bonnell_ii> bennyschwein: i cant play while im tired, maybe its the same here #steven_bonnell_ii> bobbyboosted: does minigun stream? im new here i want to see him play #steven_bonnell_ii> ryunoske: Like, I won't be like "Idra def didn't win" Because, Idra actually is really good. But... Idra def didn't win. D: #steven_bonnell_ii> roamy01 parts
Replays A file storing all actions of a single game Mouse click, selection, camera movement etc. Building/unit training, etc. Chat between players 31
32 (Extremely!) Noisy Data Actions are stored, but not their result SC2 engine required to build the states: not available Hard to evaluate states (current money, army, buildings) Mind game (use of fog of war and traps) Mental & physical skills: vision/micro/macro
33 A 1 st experiment with replays 18547 (unique) pro replays available at 30 Sept 2011. Harvested from the 6 main online repositories Sequence mining determine frequent (surprising) (winning) strategies Only based on build-orders <(depot) (barracks) (refinery)..> [WIN] <(hatchery) (spawning pool) (extractor)..> [LOSS] Several applications! Lot of information in replays (time, geo, selections, APM )
34 Replay abstraction Start from a huge noisy list of orders/events Abstract it at different levels Detect, name & interpret groups of events (+ partof relation => a lattice) Helps in strategy characterization, event detection, datamining (e.g. replay clustering), AI design, player profiling, turning point detection, (supervised) Win ratio evaluation (like poker), etc.
Ladder, Leagues & rankings 35
36 Example of LiquiPedia information
Social Sciences 37
Artificial Intelligence 38
Jusque dans les bars 40