During after-match interviews, statements like “we dominated the opponents” or “we suffered their playing style too much” are common ways to express a coach’s perception of game dominance in soccer. However, they do not necessarily reflect the actual match statistics and the performance data. A team’s ball possession, for instance, is often “sterile” as it is mainly performed far from the opponents’ goal. Similarly, shots are doubtless a key performance indicator but they cannot fully express the concept of “game dominance”.

A new possible interpretation of game dominance relies on the concept of invasion index, a metric which quantifies a team’s ability to play close to opponent’s goal. A team A’s invasion index during a game can be computed with a simple algorithm:

  • Split the game into possession phases, i.e., sequence of consecutive events on the ball generated by team A;
  • for each possession phase of team A, take the event with the highest weight. An event’s weight is computed as its “dangerousness”, i.e., the probability to score from the position where the event occurs. The dangerousness of a position is computed according to the schema in Figure 1;
  • compute the sum of all the weights obtained for team A during the game.
Coaching World  IMG01
Figure 1. Dangerousness of each zone of in the opponent’s area. Every zone is associated with a value of dangerousness according to the probability of scoring from that zone.

Wyscout data allow us to compute a team’s invasion index in a very efficient manner: since an event (pass, duel, shot, etc.) is characterized by a timestamp and a spatial location, we can easily split a game into possession phases and assign a weight to every event.

The Invasion Index describes how effective a team’s playing style is. To give an idea of its descriptive power, let us consider an entire season of two teams: Mourinho’s Manchester United (sixth in the last Premier League) and Gasperini’s Atalanta (fourth in the last Serie A). In Figure 2, as the season goes by we plot the difference between the Invasion index produced by Manchester United and its respective opponents. Figure 3 does the same for Atalanta. A positive value indicates that Manchester United (or Atalanta) has a higher invasion index than the opponent. For Manchester United, we find three main trends: a slow start, a growth as the season goes by, and a significant decrease at the end of the season. Atalanta shows a more fluctuating behavior, as witnessed by the higher number of negative peaks during the season.

Bokeh Plot

Figure 2. Evolution as the season goes by of the difference between the invasion index of Manchester United and the invasion index of its opponent. Trend indicates a smoothed average.

Bokeh Plot

Figure 3. Evolution of the difference between the invasion index of Atalanta and its opponents during the season. Trend indicates a smoothed average.

To better characterize a team’s playing style, we also consider the average acceleration, an estimate of how fast a team is on reaching the most dangerous position during a possession phase (Figures 4 and 5). A team’s average acceleration during a game is computed as follows:

  • For each possession phase, take the position of the first event e_1 and the of the event associated with the most dangerous position e_max;
  • compute the time difference delta_t = t(e_max) and t(e_1), where t(e_i) indicates the timestamp of an event;
  • compute the ratio between the invasion index of the possession phase and the square of delta_t.

Looking at the average acceleration, the difference between the playing styles of Manchester United and Atalanta is more evident (see Figures 3 and 4). We find that Atalanta has, in average, higher accelerations than Manchester United, the latter rarely reaching acceleration values higher than 60 (Figures 4 and 5). In the last season, the surprising team managed by Gasperini was able to reach the opponent’s area more frequently and faster than Mourinho’s team.

Bokeh Plot

Figure 4. Violin plot representing the average invasion acceleration for Manchester United. The higher the average invasion acceleration, the faster the team reached the opponent’s area. The width of the violin indicates the frequency of possession phases having a certain average invasion acceleration.

Bokeh Plot

Figure 5. Violin plot representing the average invasion acceleration for Atalanta. The higher the average invasion acceleration, the faster the team reached the opponent’s area. The width of the violin indicates the frequency of possession phases having a certain average invasion acceleration.

The analysis of the invasion index in a single game also provides interesting information about game dominance. One of the most difficult matches for Manchester United in the last season was the match against Chelsea at Stanford Bridge (Figure 6). Chelsea (blue line) started the game overperforming the Red Devils: after 20 minutes Chelsea was already leading 2-0. At the end of the first half, Manchester United started dominating the game, as witnessed by a higher invasion index during the game (Figure 6).

Bokeh Plot

Figure 6. Invasion index obtained during the game between Chelsea (blue line) and Machester United (red line).

The most important game for Atalanta was probably the victory against Napoli, a team well known for having a dominating playing style. Gasperini and his team managed to “invade” the opponents pitch for almost the entire first half, getting the lead of the game: 1-0. Napoli then got the control of the game, without scoring any goal though. Atalanta, on the contrary, scored the second goal.

Bokeh Plot

Figure 7. Invasion index obtained during the game between Napoli (blue line) and Atalanta (red line).

Invasion index is just one of the new metrics to be computed using detailed data available from WyScout data. As an instance, by defining a possession phase as a sequence of interactions among players, we could compute the most “invasive” subsequences of interactions. Then, we can add a further spatial layer, by taking into account zones of the pitch where interactions occur, and analyze how invasive possession phases are created. In summary, the invasion index can be a starting point for developing a different way of understanding dominance in soccer during both an entire game and a single possession phase, on which we can construct more refined features to detail collective and individual soccer performance.


Invasion index has been introduced in the following paper:
D. Link and H. Weber, Using individual ball possession as a performance indicator in soccer, workshop on Large-Scale Sports Analytics, 2015.



Coaching World  logo-SoBigData-DEFINITIVO-300x106

Coaching World  rossi-alessio
Alessio Rossi
Coaching World  cintia-paolo-298x300
Paolo Cintia
Coaching World  pappalardo-luca-298x300
Luca Pappalardo