Data-driven
evaluation of soccer players performance
Paolo Cintia - University of Pisa
{'eventName': 8,
'eventSec': 8.221464, 'id': 217097515,
'matchId': 2576132, 'matchPeriod': '1H', 'playerId': 8306,
'positions': [{'x': 42, 'y': 14}, {'x': 74, 'y': 33}],
'subEventName': 83,
'tags': [{'id': 1801}], 'teamId': 3158}
pass
high pass
accurate
1700 events per match (in average)
identifiers
Ranking soccer players
● Multidimensional and flexible
existing metrics are based just on passes or shots (e.g., FS, PSV)
Characteristics of the method
PlayeRank
● Dynamic and role-aware
compare apples to apples
● Validated
existing metrics are validated just against goals or assists (!)
soccer logs
Feature weighting Role detector
learning
Player Rating
Players Rankings Individual performance
extraction
feature weights
Role detector
Learning Rating
batch online
𝜇 (a)
(c) (b)
b1
b2 c1
(d) Ranking
d1d2 c2
Feature weighting
Role classification
Rating computation
performance rating of u in game g
taking into account
the number of goals
18 competitions
30 million events
20K matches
21K players
Experiments
Best players in the dataset
Evaluation of PlayeRank
1. We randomly extract pairs of players from the rankings 2. Ask professional scouts who’s the best for each pair 3. Ask PlayeRank who’s the best for each pair
4. Compute the agreement between the answers Compare with state-of-the-art:
● FC → Duch et al., Quantifying the Performance of Individual Players in a Team Activity, PLoS One
● PSV → Brooks et al., Developing a Data-Driven Player Ranking in Soccer using Predictive Model Weights, SIGKDD
Evaluation of PlayeRank
Evaluation of PlayeRank
Evaluation of PlayeRank
Evolution of players
Evolution of players
Patterns of performance
Versatility of players
Versatility of players
https://arxiv.org/abs/1802.04987