Classifying Dota 2 Heroes Based on Play Style and Performance

Transcription

Classifying Dota 2 Heroes Based on Play Style and Performance
Classifying Dota 2 Heroes Based on
Play Style and Performance
Jamie Lowder, Dave Wong, Lynn Gao, James Judd
Motivation
• In team sports a player is usually assigned a position.
• Quarterback
• Center
• Physiological attributes and performance characteristics of players
tend to vary based on position [1, 2, 3, 4].
• Quarterback -> pass completions
• Center -> height
Extension to Dota 2
• In Dota 2, heroes are used for specific positions.
• Assume the findings for traditional team sports extend to eSports.
• Then we can classify heroes by performance and play style.
Dota 2
Rules and basic strategy
Dota 2: Overview
• 10 players
• 2 teams
• 5 players per team
• Each player controls 1 hero
Dota 2
Strategy
Dota 2 Roles
Highest
Gold and Experience
Priority
• Carry
• Weak early. Strong late.
• Solo Lane
• Strong in the midgame.
• Support
Lowest
• Strong early. Weak late.
Data
• Two datasets
1. Professional Data Set
•
124 matches from The International 3
•
1,240 data points
2. Public Data Set
• 12,000 matches from public matchmaking
• Matches are 20-80 minutes long
• ~120,000 data points
Data Labels
• Hero ID
• 108 heroes
• Hero role
1. Carry
2. Solo lane
3. Support
Feature Selection
• 275 features
• Can be broken down into 3 groups
1.
2.
3.
Performance
Items
Skill build
Performance Features
• Per minute:
1. Kills
2. Deaths
3. Assists
4. Gold
5. Experience
6. Last hits
7. Denies
8. Hero damage
9. Tower damage
10. Hero healing
Item Features
• 240 unique items exist
• Each item has unique id
• Array of length 240
• 0 if hero did not have that item
• 1 if hero did have that item
Skill Build Features
• Max hero level is 25
• Each level a hero gets an ability point
• Array of length 25
• Position of ability upgraded at each level
Methods
• Supervised learning
• Logistic Regression
• Random Forests
• Random 90% of data used for training
• Remaining 10% of data used for testing
Results
Professional
Hero ID
Hero Role
Logistic Regression
0.72
0.89
Random Forest
0.80
0.88
Public
Hero ID
Hero Role
Logistic Regression
0.77
0.84
Random Forest
0.72
0.85
References
1.
V. Di Salvo, R. Baron, H. Tschan, F. J. Calderon Montero, N. Bachl, and F. Pigozzi, “Performance characteristics according
to playing position in elite soccer,” International journal of sports medicine, vol. 28, no. 3, p. 222, 2007.
2.
D. C. W. Nicholas, “Anthropometric and Physiological Characteristics of Rugby Union Football Players,” Sports Med, vol.
23, no. 6, pp. 375–396, Jun. 1997.
3.
G. Ziv and D. R. Lidor, “Physical Attributes, Physiological Characteris-tics, On-Court Performances and Nutritional
Strategies of Female and Male Basketball Players,” Sports Med, vol. 39, no. 7, pp. 547–568, Jul. 2009.
4.
G. Ziv and R. Lidor, “Physical characteristics, physiological attributes, and on-court performances of handball players: A
review,” European Journal of Sport Science, vol. 9, no. 6, pp. 375–386, 2009.