SLIDE 1
The University of Sydney Page 1
Spatio-temporal analysis
- f team sports
The University of Sydney Page 2
of team sports Joachim Gudmundsson The University of Sydney Page 1 - - PowerPoint PPT Presentation
of team sports Joachim Gudmundsson The University of Sydney Page 1 - - PowerPoint PPT Presentation
Spatio-temporal analysis of team sports Joachim Gudmundsson The University of Sydney Page 1 Team sport analysis Talk is partly based on: Joachim Gudmundsson and Michael Horton Spatio-Temporal Analysis of Team Sports ACM Computing Surveys,
SLIDE 2
SLIDE 3
The University of Sydney Page 3
Overview of major approaches
SLIDE 4
The University of Sydney Page 4
Input data
PLAYER NAME TEAM NAME MATCH FIXTURE HALF TIME Player X Position Player Y Position Bacary Sagna Arsenal Arsenal v Bolton First half
- 1745
1897 Bacary Sagna Arsenal Arsenal v Bolton First half 0.1
- 1748
1902 Bacary Sagna Arsenal Arsenal v Bolton First half 0.2
- 1751
1907 Bacary Sagna Arsenal Arsenal v Bolton First half 0.3
- 1754
1913 Bacary Sagna Arsenal Arsenal v Bolton First half 0.4
- 1757
1918 Bacary Sagna Arsenal Arsenal v Bolton First half 0.5
- 1760
1923 Bacary Sagna Arsenal Arsenal v Bolton First half 0.6
- 1763
1929 83.8 Touch DIABY Abou 24
- 13
84.8 Block BASHAM Chris 25
- 12
86.7 Pass MCCANN Gavin 23
- 4
88 Foul GARDNER Ricardo DENILSON 15
- 8
109 Direct Free Kick Pass JAASKELAINEN Jussi 14
- 7
111.2 Header CLICHY Gael
- 26
- 11
113 Touch CLICHY Gael
- 26
- 17
SLIDE 5
The University of Sydney Page 5
Input data
SLIDE 6
The University of Sydney Page 6
History: Sports analysis
– Box scores for baseball started in the 1850s. – Manual notation of football games started in the 1950s. – Moneyball-era in baseball – Similar development in basketball in the last 10 years – Human observations are unreliable. Franks and Miller [1986] showed that expert observers’ recollection of significant match events is as low as 42%. – Automated tracking of sport players started in the early 2000s. – Nowadays a number of automatic tracking systems for football, ice hockey and basketball (not much in rugby, AFL and handball).
SLIDE 7
The University of Sydney Page 7
Outline
– Playing area subdivision – Dominant regions – Applications – Modelling player interaction as social networks – Data mining – Labelling – Identifying formations and plays – Trajectory analysis – Sport-specific trajectory problems
SLIDE 8
The University of Sydney Page 8
Playing area subdivision: Intensity maps
First attempts to analyze trajectory data…
SLIDE 9
The University of Sydney Page 9
Playing area subdivision: Intensity maps
And more…
SLIDE 10
The University of Sydney Page 10
Playing area subdivision: Dominant region
A team’s ability to control space is considered a key factor in the team’s performance. Dominant region [Taki and Hasegawa’99] The dominant region of a player p is the region of the pitch that player p can reach before any other player.
p Reach?
SLIDE 11
The University of Sydney Page 11
DR(p)={x | d(x,p) ≤ d(x,q) for all qp} If d(,) = Euclidean distance then Dominant region = Voronoi diagram [Descartes 1644]
Dominant region [Taki and Hasegawa’99] The dominant region of a player p is the region of the pitch that player p can reach before any other player.
Playing area subdivision: Dominant region
SLIDE 12
The University of Sydney Page 12
Playing area subdivision: Movement model
[Taki and Hasegawa’99] Linear interpolation of acceleration in all directions. [Fujimura and Sugihara’05] Introduced a resistive force to decrease acceleration.
SLIDE 13
The University of Sydney Page 13
Playing area subdivision: Movement models
Simple way to model?
SLIDE 14
The University of Sydney Page 14
Circle model Ellipse model
Movement model
Playing area subdivision: Movement model
SLIDE 15
The University of Sydney Page 15
A bisector in the ellipse model
Movement model
Playing area subdivision: Movement model
SLIDE 16
The University of Sydney Page 16
Dominant region
Playing area subdivision: Movement model
SLIDE 17
The University of Sydney Page 17
Model: Turning cost + Euclidean distance
Movement model
Playing area subdivision: Movement model
SLIDE 18
The University of Sydney Page 18
1 [Taki & Hasegawa’00]
Movement model
Playing area subdivision: Movement model
SLIDE 19
The University of Sydney Page 19
Movement model
Playing area subdivision: Movement model
[De Berg, Haverkort and Horton’17]
SLIDE 20
The University of Sydney Page 20
Playing area subdivision: Movement model
Open problem 1: Define a motion function that faithfully models player movement and is tractable for computation.
SLIDE 21
The University of Sydney Page 21
Playing area subdivision: Passing evaluation
A player p is open for a pass if there is some direction and (reasonable) speed that the ball can be passed, such that p can intercept the ball before any other player.
SLIDE 22
The University of Sydney Page 22
Playing area subdivision: Passing evaluation
Passability with a fixed pass speed (20m/s).
SLIDE 23
The University of Sydney Page 23
Playing area subdivision: Passing evaluation
The existing models for determining whether a player is
- pen to receive a pass only consider passes made along the
shortest path between passer and receiver and where the ball is moving at constant velocity. Open problem 2: Develop a more realistic model that allows for aerial passes, effects of ball-spin, and variable velocities.
SLIDE 24
The University of Sydney Page 24
Spatial pressure of player
Playing area subdivision: Spatial pressure
[Taki et al. ‘96] Spatial pressure for a player p is related to the fraction P
- f the disk of radius r centred at p that lies within dominant
region of opposing players, i.e. m(1-P)+(1-m)(1-d/D), where d – distance between p and the ball D – distance from p to point furthest from p on pitch m – preset weight
SLIDE 25
The University of Sydney Page 25
Spatial pressure of player
Playing area subdivision: Spatial pressure
SLIDE 26
The University of Sydney Page 26
Spatial pressure of player
Playing area subdivision: Spatial pressure
SLIDE 27
The University of Sydney Page 27
Spatial pressure of player
Playing area subdivision: Spatial pressure
SLIDE 28
The University of Sydney Page 28
Spatial pressure of player
Playing area subdivision: Spatial pressure
SLIDE 29
The University of Sydney Page 29
Spatial pressure of player
Playing area subdivision: Spatial pressure
SLIDE 30
The University of Sydney Page 30
Playing area subdivision: Spatial pressure
The definition of spatial pressure is very simple. Open problem 3: Can a model that incorporates the direction the player is facing or the direction of pressuring opponents be devised and experimentally tested?
SLIDE 31
The University of Sydney Page 31
Modelling team sports as social networks
Understanding the interaction between players is one of the most important and complex problems in sports science. Numerous papers apply social network analysis to team sports. Passing network Transition network
SLIDE 32
The University of Sydney Page 32
Modelling team sports as social networks
Many properties of passing networks have been studied: – Centrality – Degree – Betweenness – Closeness – Eigenvector centrality and Pagerank – Clustering coefficients – Density and heterogeneity – Entropy, topological depth, Price-of-Anarchy
SLIDE 33
The University of Sydney Page 33
Modelling team sports as social networks
[Grund’12] Studied degree centrality on networks generated from 283k passes. Conclusion: High level of centralization decreases team performance. Open problem 4: A systematic study reviewing various centrality and clustering measures against predefined criteria, and on a large dataset would be a useful contribution to the field.
SLIDE 34
The University of Sydney Page 34
Modelling team sports as social networks
[Balkundi and Harrison’06] Density-performance hypothesis. More passes will make a team stronger. Open problem 5: The density-performance hypothesis suggests an interesting metric
- f team performance. Can this hypothesis be tested scientifically?
SLIDE 35
The University of Sydney Page 35
Data mining: Labelling events
– Evaluate passes (good/bad) [Horton et al.’15] – Identify teams (based on formation) [Bialkowski et al.’14] – Predict rebounds (offensive/defensive team) [Maheswaran et al’12]
SLIDE 36
The University of Sydney Page 36
Data mining: Labelling passes
Examples of features:
- Area of receiving player’s dominant region
- The net change in the area of receiving player’s dominant region
- Total area of the team’s dominant region
- The net change of the total area of the team’s dominant region
- Passer Pressure
- Receiver Pressure
- Passer-Receiver Pressure Net Change
SLIDE 37
The University of Sydney Page 37
Data mining: Labelling passes
- Extracted feature vectors from 2932 passes from four matches
- Pass examples were labelled by humans watching video of match
- Class imbalance:
- SVN classifier: Accuracy 90.8% which is similar to a human observer
- Features based on dominating region are among the most
important
Class
- Rel. frequency
Count Good 0.066 193 OK 0.789 2314 Bad 0.145 425 [Horton et al.’15]
SLIDE 38
The University of Sydney Page 38
Data mining: Labelling passes
Our algorithms can with high accuracy give the following information:
– Number of “good”, “ok” or “bad” passes made by a player. – The number of high risk vs low risk passes a player makes. – A player’s ability to “execute” a pass.
SLIDE 39
The University of Sydney Page 39
Data mining: Role assignment to players
Role swapping has been shown to be an effective attacking tactic. (Left defender swaps position with left midfielder during play) Given the position of the players and a “formation” which role has each player? Assignment problem (minimize sum).
4-4-2
SLIDE 40
The University of Sydney Page 40
Data mining: Role assignment to players
What if we have many different “formations”?
4-4-2 3-5-2 4-5-1 1-3-3-1-2
SLIDE 41
The University of Sydney Page 41
Data mining: Identifying plays
Given the movement of the players and a “predefined play” which role has each player?
SLIDE 42
The University of Sydney Page 42
Data mining: Identifying plays
What if we have many predefined “plays”?
SLIDE 43
The University of Sydney Page 43
Currently not used much in team sports analysis. – Hard to work with – Not many available tools
Trajectory analysis: Team sport perspective
SLIDE 44
The University of Sydney Page 44
Given a set T={T1, …, Tm} of trajectories. Typical queries: – Given a query trajectory Q, report the “nearest” subtrajectory of a trajectory in T. [Restricted in time? Restricted to subset of trajectories?]
Trajectory analysis: Team sport perspective
T1 Q
[Driemel and Har-Peled’13, De Berg et al.’13, G and Smid’15]
SLIDE 45
The University of Sydney Page 45
– Given a set of query trajectories Q={Q1, … , Qk}, report the “nearest” set of k subtrajectories of k different trajectories in T. [Subtrajectories must be during same time interval. Restricted to subset of trajectories?]
Trajectory analysis: Team sport perspective
T1 T2 T3
Q1 Q2
SLIDE 46
The University of Sydney Page 46
– Subtrajectory clustering of large sets of trajectories? Current approaches are very slow. – Distance measure between trajectories?
Trajectory analysis: Team sport perspective
SLIDE 47
The University of Sydney Page 47
– Clustering of multiple subtrajectories occurring in the same time interval?
Trajectory analysis: Team sport perspective
SLIDE 48
The University of Sydney Page 48
One season in Premier League generates roughly 1 billion points. General questions: – Can we sample the data? – Can we use core sets for some simple query problems? – Can we construct data structures that supports adding more data, without having to recompute them? – Can we construct multi-purpose data structures?
Trajectory analysis: Team sport perspective
SLIDE 49
The University of Sydney Page 49