By John Considine
Ian Graham’s How to Win the Premier League is a fine book for the way it outlines the difficulties with data and its analysis. The subtitle of the book is “The inside story of football’s data revolution”. I found the story of the evolution, rather than the revolution, more interesting. It is neatly illustrated by a section called “The Ridgewell Problem”.
One of Graham’s earlier models rated Liam Ridgewell as one of the best defenders in Europe. Yet, Ridgewell was on a Premier League team that ranked 19 of 20 when it came to expected goals conceded. Graham says that “In common with other team sports, football has found it most difficult to measure and analyse defensive ability. The problem we had was that we were unsure of the defenders’ positions when tackles didn’t happen and a dangerous chance was conceded.” The italics are Graham’s. How do you measure something that didn’t happen? Let us park this for the moment.
Economists encounter something similar. It is called the broken window fallacy. Our gut instinct tells us that when a kid kicks a ball through a window it is a bad outcome. However, if the kid employs a consultancy firm then the consultants will explain how the kid is serving the economy and should get preferential fiscal treatment. They will explain that the expenditure of money on the replacement glass, and the person who installs it, are only the first round of a chain that will boost national income. They will explain how the glass-shop owner and the window installer will spend the money they got and this will generate income for others. There will be a multiple of the original expenditure circulating. These consultants are like the data analysts that are measuring what happened rather than what did not happen. We see the tackles just like we see the expenditure of the window owner on a replacement piece of glass. We do not see what he would have spent the money on otherwise. We do not see a goal if it was prevented.
I was reminded of this last Sunday by an excellent piece of defending in hurling. It can be viewed here from 11:50 to 11:30 on the footage. One of the best, if not the best, attackers in the game collected the ball in his hand. This meant that one of the attacker’s two permitted catches was taken. The attacker turned left towards goal and considered shooting but the defender was close enough to block any potential shot. Therefore, the attacker elected to put the ball on his hurley and run towards goal. The defender had positioned himself so that the attacker took the outside, and longer, route to goal. There were two further benefits for the defender. The attacker would be forced to strike off his relatively weaker side and from a narrower angle.
As they raced towards goal, the attacker tried to open up the angle a little. This required greater ball security so the attacker used his final permitted catch as he turned in. This reduced the attacker’s advantage as he could only cover four steps with the ball in hand. Realising he could not open the angle, the attacker tossed the ball back onto his stick and attempted to push away the defender. But the defender had left enough of a gap that the handoff was only partly successful. In addition, the defender had used the narrowing of the space to push the attacked the other direction. The defender also clipped the attacker with his hurley in the way that a jockey twists his whip towards a horse’s neck to remind him of his presence. He was unlikely to incur the wrath of the officials for excess use of the hurley. At this stage, the attacker still had the ball but he had only one hand on the hurley that he was holding short to ensure ball control. He could no longer catch the ball. As a result, he tossed the ball upwards and struck it one-handed from a narrow angle and off his relatively weaker side. The ball was struck off target (although the goalkeeper made sure at the cost of a 65m free shot). Why did Aaron Gillane not score? Maybe Sean O'Donoghue had something to do with it.
At 12:27 the commentator says, “Brilliant, from goalkeeper and forward”. Even excellent defending that is seen is not often appreciated. We might be a while away from being able to put the previous two paragraphs into data usable by statistical models. Long live the Evolution.
One of Graham’s earlier models rated Liam Ridgewell as one of the best defenders in Europe. Yet, Ridgewell was on a Premier League team that ranked 19 of 20 when it came to expected goals conceded. Graham says that “In common with other team sports, football has found it most difficult to measure and analyse defensive ability. The problem we had was that we were unsure of the defenders’ positions when tackles didn’t happen and a dangerous chance was conceded.” The italics are Graham’s. How do you measure something that didn’t happen? Let us park this for the moment.
Economists encounter something similar. It is called the broken window fallacy. Our gut instinct tells us that when a kid kicks a ball through a window it is a bad outcome. However, if the kid employs a consultancy firm then the consultants will explain how the kid is serving the economy and should get preferential fiscal treatment. They will explain that the expenditure of money on the replacement glass, and the person who installs it, are only the first round of a chain that will boost national income. They will explain how the glass-shop owner and the window installer will spend the money they got and this will generate income for others. There will be a multiple of the original expenditure circulating. These consultants are like the data analysts that are measuring what happened rather than what did not happen. We see the tackles just like we see the expenditure of the window owner on a replacement piece of glass. We do not see what he would have spent the money on otherwise. We do not see a goal if it was prevented.
I was reminded of this last Sunday by an excellent piece of defending in hurling. It can be viewed here from 11:50 to 11:30 on the footage. One of the best, if not the best, attackers in the game collected the ball in his hand. This meant that one of the attacker’s two permitted catches was taken. The attacker turned left towards goal and considered shooting but the defender was close enough to block any potential shot. Therefore, the attacker elected to put the ball on his hurley and run towards goal. The defender had positioned himself so that the attacker took the outside, and longer, route to goal. There were two further benefits for the defender. The attacker would be forced to strike off his relatively weaker side and from a narrower angle.
As they raced towards goal, the attacker tried to open up the angle a little. This required greater ball security so the attacker used his final permitted catch as he turned in. This reduced the attacker’s advantage as he could only cover four steps with the ball in hand. Realising he could not open the angle, the attacker tossed the ball back onto his stick and attempted to push away the defender. But the defender had left enough of a gap that the handoff was only partly successful. In addition, the defender had used the narrowing of the space to push the attacked the other direction. The defender also clipped the attacker with his hurley in the way that a jockey twists his whip towards a horse’s neck to remind him of his presence. He was unlikely to incur the wrath of the officials for excess use of the hurley. At this stage, the attacker still had the ball but he had only one hand on the hurley that he was holding short to ensure ball control. He could no longer catch the ball. As a result, he tossed the ball upwards and struck it one-handed from a narrow angle and off his relatively weaker side. The ball was struck off target (although the goalkeeper made sure at the cost of a 65m free shot). Why did Aaron Gillane not score? Maybe Sean O'Donoghue had something to do with it.
At 12:27 the commentator says, “Brilliant, from goalkeeper and forward”. Even excellent defending that is seen is not often appreciated. We might be a while away from being able to put the previous two paragraphs into data usable by statistical models. Long live the Evolution.
RSS Feed
