Corinthians and Flamengo face each other this Tuesday, at 9:30 pm, for the first leg of the Copa Libertadores quarter-finals in a match with strong potential to be decided from an aerial move: not counting penalties and direct fouls, from the last ten goals scored by Corinthians, six came from high-ball plays, and Flamengo thus conceded seven of the last ten goals. On the other hand, Flamengo lifted the ball to score eight of the last ten goals, and Corinthians conceded half of the last ten goals that way. It’s a huge potential for the aerial game to decide.
+ See the Libertadores table
In partnership with economist Bruno Imaizumi, we analyzed 88,629 shots registered by Espião Statístico in 3,619 Brasileirões games since the 2013 edition, which serve as a parameter to measure the current productivity of teams based on the goal expectation (xG), an internationally consolidated metric ( see the methodology at the end of the text).
Based on this metric, even as a visitor, Flamengo is the favorite to win the match, with a 44% probability of taking the three points against 28% in favor of Corinthians and 28% for a draw. Flamengo has the most accurate attack in Libertadores, with a goal scored every 4.6 completions and an average of 13.3 completions per game. There were 106 shots and 23 goals scored. Flamengo’s offensive pressure earned him an average of 3.6 corners per match.
— Photo: Statistical Spy
Even when considering only the four games played as a visitor in the competition, Flamengo made 37 shots and scored 8 goals, one every 4.6 attempts.
In Libertadores, Corinthians has the third most resistant defense against shots, with an average of one goal conceded for every 21.3 contrary conclusions and only 10.6 shots conceded per match. No other Brazilian defense was so resistant to pressure in this year’s edition of Libertadores.
Flamengo, in comparison, is only the 14th resistance of the Libertadores, with a goal conceded for every 11.3 completions and an average of 10.0 shots conceded per game.
It will be a challenge for Corinthians. If in the Brasileirão it has the most effective attack in the competition with a goal scored every 8.3 attempts, in Libertadores it only achieved the sixth lowest offensive efficiency when compared to the 32 teams that started the competition: Corinthians needed an average of 15.8 shots for each goal scored with an average of 9.0 attempts per game. If Flamengo got 3.6 corners per match, Corinthians had an average of 3.0.
See below the performance of the two teams in the last 60 days and how they compare to the 20 teams that compete in Serie A of the Brazilian Championship. Corinthians did not concede a goal in seven of the eight games they played as home team this two-month period, and Flamengo only failed to score in three of the eight games they played as a visitor.
— Photo: Statistical Spy
Favoritism presents the potential that each team carries comparing the performance in the last 60 days as home or away in all competitions and in the last six games, regardless of home. Also considered are the defensive and offensive performances of the teams in the aerial and low game. The calculations referring to the influence of high balls and the exchange of low passes between goals scored and conceded only consider the characteristics of goals scored in plays. Olympic goals, penalty kicks and direct free kicks do not count towards determining aerial or ground influence as they are kicks made directly into the goal.
We present the statistical probabilities based on the parameters of the “Expected Goals” or “Goals Expectation” (xG) model, a metric consolidated in the data analysis that has as a reference 88,629 shots registered by the Espião Statístico in 3,619 Brasileirões games since the edition of 2013. We consider the distance and angle of the shot, as well as characteristics related to the origin of the play (for example, whether it came from a cross, direct foul or a steal), the body part used, whether the shot was made first, the difference in the market value of the teams in each season, the playing time and the difference in the score at the time of each shot.
A player’s performance is compared to the average for his position, whether he is attacker, midfielder, midfielder, full-back or defender, and we consider what would be expected of the shot if done with the “good foot” (the right for right-handers, the left for left-handers) and “bad foot” (the opposite). The ambidextrous, which kick approximately the same number of times with each foot, were identified.
Out of every hundred shots from the half-moon, for example, only seven turn into a goal. So, a half moon shot has a goal expectation (xG) of about 0.07. Each position on the field has a different expectation of a shot turning into a goal, which increases if it is a counterattack because there are fewer opponents to avoid the completion of the play. Each score is added up over the course of the match to arrive at a team’s total xG in each game.
The model used in the analysis follows a statistical distribution called Bivariate Poisson, which calculates the probabilities of events (in this case, the goals of each team) to happen within a certain time interval (the game). To arrive at the predictions about the chances of each team finishing the championship in each position, the Monte Carlo method was used, which basically relies on simulations to generate results. For each game not yet played, we run ten thousand simulations.
*The Statistical Spy team is formed by: Guilherme Maniaudet, Guilherme Marçal, João Guerra, Leandro Silva, Leonardo Martins, Roberto Maleson and Valmir Storti.