deletions | additions       diff --git a/subsubsection_Position_based_metric_Goal__.tex b/subsubsection_Position_based_metric_Goal__.tex  index ae367a6..8c5d4e3 100644  --- a/subsubsection_Position_based_metric_Goal__.tex  +++ b/subsubsection_Position_based_metric_Goal__.tex 
...
\subsubsection{Position-based metric}  When $C_{penalty}$ is low $(C_{penalty}=1.3)$ almost 75\% of the clusters contain only one player.  Goal keepers appear almost always isolated in two different clusters excepting during an action embedding or close to a shot. They are the last players to join the last cluster when $C_{penalty}$ is increased.  During the first minutes of the play we get less clusters embedding more players (2 clusters with 6 players) which might be due, in part, to the disorganization of the team at the beginning.  Those results are encouraging. It seems MI-based metric with position is relevant for the cues about a match we were looking for. for adjusting accurately $C_{penalty}$.