Understanding the value of a football player is a challenging problem. Player valuation is not only critical for scouting, bidding and negotiation processes but also attracts a large media and fan interest. Due to the complexities which arise from the fact that player pool is distributed over hundreds of different leagues and many different playing positions, many clubs hire domain experts (often retired professional players) in order to evaluate the value of potential players. We argue that such human-based scouting has several drawbacks including high cost, inability to scale to thousands of active players and inevitable subjective biases. In this paper we present a methodology for data-driven player market value estimation which tackles these drawbacks. To examine the quality of the proposed methodology and demonstrate that our data-driven valuation outperforms widely used transfermarkt.com market value estimates in predicting the team performance.
[1]
Dino Pedreschi,et al.
The harsh rule of the goals: Data-driven performance indicators for football teams
,
2015,
2015 IEEE International Conference on Data Science and Advanced Analytics (DSAA).
[2]
Lars Magnus Hvattum,et al.
Using ELO ratings for match result prediction in association football
,
2010
.
[3]
Nagarajan Natarajan,et al.
Learning with Noisy Labels
,
2013,
NIPS.
[4]
Aritra Ghosh,et al.
Making risk minimization tolerant to label noise
,
2014,
Neurocomputing.
[5]
Mohamed Hefeeda,et al.
Estimating the Maximal Speed of Soccer Players on Scale
,
2015,
MLSA@PKDD/ECML.
[6]
A. Elo.
The rating of chessplayers, past and present
,
1978
.
[7]
M. Glickman.
The Glicko system
,
2011
.
[8]
Bill Gerrard,et al.
Ticket Prices, Concessions and Attendance at Professional Sporting Events
,
2007
.