Toast
Established Member
I translated this intresting article from a Dutch site for you. Enjoy. Original article here.
Part II has now been included in this post.
****
Although Arsène Wenger has been at the helm of the club as Arsenal manager for 20 years, a subtle change of course is happening behind the scences at the Gunners. The American school of 'to measure is to know' is to give Arsenal an advantage against its wealthy competitors. Reality turns out to be more difficult that the theory.
In March this year Stan Kroenke gave an insight into his philosophy of sportsmanagement. 'I've always been a fan of the Moneyball-model. Billy Bean is a fan of our manager Arsène Wenger, who is a qualified economist. He has always maintained that analytical look. Just like our other teams we have at Arsenal a statistical view.
[Due to his American franchises] taking sport-related decisions is normal for Kroenke. After taking over at Arsenal he immediately presses for investment in an advance stats analysis team. Dat leads to the acquisition of StatDNA in October 2012. From a cost point of view the statistics are gathered in Cambodi.
[While the Americans praise and celebrate the use of statistics] Arsenal do not support this narrative. Ivan Gazidis never names StatDNA in the annual financial reports, instead referring to AOH-USA LLC. Nevertheless, Gazidis claims that the buying of StatDNA was crucial to Arsenal's competitiveness. The analysts advise on various issues. From scouting and identifying talent to game analyses and tactical insights.
In a recently published book player-agent Jon Smith explains how intensively the analysts are involved with Arsenal's incoming transfers. 'After Kroenke's takeover the club created indicators for players. If a footballer does not meet all the criteria, then on this basis the maximum amount of money to be spend on this player is determined. This is what Wenger means when he talks about player value. Arsenal use statistics to back-up observations about transfer targets. Or to raise questions marks over potential acquisitions. In some cases this leads to no bid being made for a scouted player, but the club mainly wants to use the date to take away doubts when, for example, poaching a player from a rival for big money'.
The statistics are then not a replacement of the traditional methods, but a supplement of them. If the manager, scouts and the data-analysts are all confident in the added value of a transfer target, then the chance of that player flopping is smaller then when one of the parties has doubts. By adding numbers as a source of extra information, Arsenal hopes to reduce the risk of making a bad decision. At the MIT Sloan Sport Analytics Conference, Kroenke compares it to retail. In retail a vast amount of data on cosumer preferences has been unearthed over the past couple of decades. This helps with making more efficient decisions, but the intuition of people of the trade is still vital. Good managers find the optimal balance between these different sources of information. It is the same in the sporting world, Kroenke thinks.
****
Part II
In the 2015/16 season Arsène Wenger starts actively integrating the insights of the analysist into his style of play. Over the years data has led to new insights into the effectiveness of certain offensive strategies in football. Particularly shots from distance and crosses are a thorn in the side of the statisticians.
Let's start with shots from distance. As a football supporter it's easy to recall dozens of examples of goals from thunderous shots from outside the box. The hopeless shots that never came anywhere near the goal are forgotten a lot quicker. Statistics don't suffer from this selective memory. By counting hundreds of thousands of attempts from the past, an estimate can be made of the probability that a shot from a certain distance and angle will lead to a goal. The jargon for this is Expected Goals. This statistic is further refined by adding extra factors, for example, whether the attempt is made with the foot or the head. Using this method it is possible to calculate the probability of a shot from distance resulting in a goal. For the average attempt from outside the box the general rule is that it leads to a goal in just over 3% of the attempts. For an attempt from within the box this goes up to 15%. Naturally, in both situations the rule 'the closer to the goal, the better' goes.
On the basis of these figures one can use game theory to search for an optimal strategy. Let’s say that a player can take a shot from 20 meters from goal, then history tells us that this shot will be successful in roughly 3% of cases. Take 33 shots like this in one game and you may expect to get one goal. It therefore seems more prudent to keep combining in the hopes of creating a qualitatively better chance. Better one chance of 30% then 5 of 3%. It’s essentially what Barcelona have done for years: combination play right up until the goal line.
Then crosses. Supporters in England seem to love nothing more than a high ball into the box. Such a ball will always lead to danger, is the thought. The statistics show something else. One in fifty crosses in the Premier League results in a goal. Here too there are exceptions to the rule: high balls from the corner of the box or after a quick counter are much more dangerous that the classic cross from the by-line. Returning the ball to the axis will in the long term probably result in better chances then crosses will, is the argument.
Wenger has been convinced by these figures. In press conferences he refers regularly to Expected Goals and new signings say that Wenger instructs them to be more selective in when to take a shot. The average distance of a shot at Arsenal in 2015/16 was 16 meters. A record since such things have been recorded in the Premier League. The number of attempts from distance decreased from the previous season from 234 to 156. A similar development happened with crosses, which decreased from 576 to 503.
Defensively, the reverse trend was true. Arsenal tempted the opponent to shoot from distance more often and to put in more crosses.
The results confirmed to expectations. Arsenal saw its attempts in the box increase and created more big chances than any other team in the Premier League. Conversely, the Gunners’ opponents had fewer chances in the box and Arsenal allowed fewer big chances as well. In almost all statistical models Arsenal came out on top as the best club in England.
It could’ve been a beautiful film scenario. A 66-year old French manager makes his style of play more efficient thanks to American data analysis, resulting in the first league title in 12 years. In reality, however, Arsenal never even came close to the title and finished 10 points behind shock-winners Leicester City.
Stranger still is that Arsenal scored fewer goals, despite the uptake in high quality chances. Arsenal’s total goal tally decreased by six compared to the previous season, whereas on the basis of Expected Goals, an increase was expected. Moreover, the expected decrease in the number of goals conceded failed to materialise as well.
It is a perplexing paradox that Arsène Wenger cannot explain either. ‘The big difference between this season and the previous one is the relation between the number of chances we created in home games and the number of goals we scored from them.’, he said in April on the Arsenal website. ‘Last year we scored the expected 2 goals per game, now 2.5. But last season our goal tally was 114% of our Expected Goals, and now it is only 50%.
Three explanations have been put forward for the discrepancy between the underlying statistics and Arsenal’s real performance.
From a statistical perspective reference is made to the phenomenon called regression to the average. Over the long term a team’s performance will be roughly even to its intrinsic quality. Over a shorter period, good or bad fortune can cause create positive and negative outliers. According to this theory Wenger is simply unlucky, and missed out on the title because the season is only 38 games. In a sport like football, in which relatively few goals are scored, the team with the best quality chances simply does not always win. In the very long term this coincidental effect will be negated, but a football season is simply not long enough for it.
An alternative hypothesis is that the Arsenal players are simply very bad finishers. Generally speaking, professional footballers tend to score more or less their expected number of goals. In previous years this also held true for the Arsenal players, which makes the explanation that they forgot how to finish improbable. The injury crisis might possibly be an explanation for the dip in finishing. A lack of fitness, according to this theory, can mean the difference between a certain goal and an inexplicable miss.
The third idea is that Arsenal has become too predictable due to its stats-modified strategy. If your opponent knows you will neither cross nor shoot from distance they will not have to think about such threats when defending. Less variation leads to fewer surprises and, ultimately, to fewer goals. Moreover, it simply takes longer to make it into the box, giving defences more time to organise themselves.
It is not clear what Arsenal have internally pointed out as the cause. What is clear is that the Gunners have started shooting from distance slightly more (but still less than in 2014/15), with two stunning goals from new signing Granit Xhaka as a result.
In the game against Chelsea the true strength of this Arsenal team will be on show. Better decisions thanks to the use of data are nice, but ultimately Wenger will need performance on the pitch to keep the support of the critical Arsenal supporters.
Part II has now been included in this post.
****
Although Arsène Wenger has been at the helm of the club as Arsenal manager for 20 years, a subtle change of course is happening behind the scences at the Gunners. The American school of 'to measure is to know' is to give Arsenal an advantage against its wealthy competitors. Reality turns out to be more difficult that the theory.
In March this year Stan Kroenke gave an insight into his philosophy of sportsmanagement. 'I've always been a fan of the Moneyball-model. Billy Bean is a fan of our manager Arsène Wenger, who is a qualified economist. He has always maintained that analytical look. Just like our other teams we have at Arsenal a statistical view.
[Due to his American franchises] taking sport-related decisions is normal for Kroenke. After taking over at Arsenal he immediately presses for investment in an advance stats analysis team. Dat leads to the acquisition of StatDNA in October 2012. From a cost point of view the statistics are gathered in Cambodi.
[While the Americans praise and celebrate the use of statistics] Arsenal do not support this narrative. Ivan Gazidis never names StatDNA in the annual financial reports, instead referring to AOH-USA LLC. Nevertheless, Gazidis claims that the buying of StatDNA was crucial to Arsenal's competitiveness. The analysts advise on various issues. From scouting and identifying talent to game analyses and tactical insights.
In a recently published book player-agent Jon Smith explains how intensively the analysts are involved with Arsenal's incoming transfers. 'After Kroenke's takeover the club created indicators for players. If a footballer does not meet all the criteria, then on this basis the maximum amount of money to be spend on this player is determined. This is what Wenger means when he talks about player value. Arsenal use statistics to back-up observations about transfer targets. Or to raise questions marks over potential acquisitions. In some cases this leads to no bid being made for a scouted player, but the club mainly wants to use the date to take away doubts when, for example, poaching a player from a rival for big money'.
The statistics are then not a replacement of the traditional methods, but a supplement of them. If the manager, scouts and the data-analysts are all confident in the added value of a transfer target, then the chance of that player flopping is smaller then when one of the parties has doubts. By adding numbers as a source of extra information, Arsenal hopes to reduce the risk of making a bad decision. At the MIT Sloan Sport Analytics Conference, Kroenke compares it to retail. In retail a vast amount of data on cosumer preferences has been unearthed over the past couple of decades. This helps with making more efficient decisions, but the intuition of people of the trade is still vital. Good managers find the optimal balance between these different sources of information. It is the same in the sporting world, Kroenke thinks.
****
Part II
In the 2015/16 season Arsène Wenger starts actively integrating the insights of the analysist into his style of play. Over the years data has led to new insights into the effectiveness of certain offensive strategies in football. Particularly shots from distance and crosses are a thorn in the side of the statisticians.
Let's start with shots from distance. As a football supporter it's easy to recall dozens of examples of goals from thunderous shots from outside the box. The hopeless shots that never came anywhere near the goal are forgotten a lot quicker. Statistics don't suffer from this selective memory. By counting hundreds of thousands of attempts from the past, an estimate can be made of the probability that a shot from a certain distance and angle will lead to a goal. The jargon for this is Expected Goals. This statistic is further refined by adding extra factors, for example, whether the attempt is made with the foot or the head. Using this method it is possible to calculate the probability of a shot from distance resulting in a goal. For the average attempt from outside the box the general rule is that it leads to a goal in just over 3% of the attempts. For an attempt from within the box this goes up to 15%. Naturally, in both situations the rule 'the closer to the goal, the better' goes.
On the basis of these figures one can use game theory to search for an optimal strategy. Let’s say that a player can take a shot from 20 meters from goal, then history tells us that this shot will be successful in roughly 3% of cases. Take 33 shots like this in one game and you may expect to get one goal. It therefore seems more prudent to keep combining in the hopes of creating a qualitatively better chance. Better one chance of 30% then 5 of 3%. It’s essentially what Barcelona have done for years: combination play right up until the goal line.
Then crosses. Supporters in England seem to love nothing more than a high ball into the box. Such a ball will always lead to danger, is the thought. The statistics show something else. One in fifty crosses in the Premier League results in a goal. Here too there are exceptions to the rule: high balls from the corner of the box or after a quick counter are much more dangerous that the classic cross from the by-line. Returning the ball to the axis will in the long term probably result in better chances then crosses will, is the argument.
Wenger has been convinced by these figures. In press conferences he refers regularly to Expected Goals and new signings say that Wenger instructs them to be more selective in when to take a shot. The average distance of a shot at Arsenal in 2015/16 was 16 meters. A record since such things have been recorded in the Premier League. The number of attempts from distance decreased from the previous season from 234 to 156. A similar development happened with crosses, which decreased from 576 to 503.
Defensively, the reverse trend was true. Arsenal tempted the opponent to shoot from distance more often and to put in more crosses.
The results confirmed to expectations. Arsenal saw its attempts in the box increase and created more big chances than any other team in the Premier League. Conversely, the Gunners’ opponents had fewer chances in the box and Arsenal allowed fewer big chances as well. In almost all statistical models Arsenal came out on top as the best club in England.
It could’ve been a beautiful film scenario. A 66-year old French manager makes his style of play more efficient thanks to American data analysis, resulting in the first league title in 12 years. In reality, however, Arsenal never even came close to the title and finished 10 points behind shock-winners Leicester City.
Stranger still is that Arsenal scored fewer goals, despite the uptake in high quality chances. Arsenal’s total goal tally decreased by six compared to the previous season, whereas on the basis of Expected Goals, an increase was expected. Moreover, the expected decrease in the number of goals conceded failed to materialise as well.
It is a perplexing paradox that Arsène Wenger cannot explain either. ‘The big difference between this season and the previous one is the relation between the number of chances we created in home games and the number of goals we scored from them.’, he said in April on the Arsenal website. ‘Last year we scored the expected 2 goals per game, now 2.5. But last season our goal tally was 114% of our Expected Goals, and now it is only 50%.
Three explanations have been put forward for the discrepancy between the underlying statistics and Arsenal’s real performance.
From a statistical perspective reference is made to the phenomenon called regression to the average. Over the long term a team’s performance will be roughly even to its intrinsic quality. Over a shorter period, good or bad fortune can cause create positive and negative outliers. According to this theory Wenger is simply unlucky, and missed out on the title because the season is only 38 games. In a sport like football, in which relatively few goals are scored, the team with the best quality chances simply does not always win. In the very long term this coincidental effect will be negated, but a football season is simply not long enough for it.
An alternative hypothesis is that the Arsenal players are simply very bad finishers. Generally speaking, professional footballers tend to score more or less their expected number of goals. In previous years this also held true for the Arsenal players, which makes the explanation that they forgot how to finish improbable. The injury crisis might possibly be an explanation for the dip in finishing. A lack of fitness, according to this theory, can mean the difference between a certain goal and an inexplicable miss.
The third idea is that Arsenal has become too predictable due to its stats-modified strategy. If your opponent knows you will neither cross nor shoot from distance they will not have to think about such threats when defending. Less variation leads to fewer surprises and, ultimately, to fewer goals. Moreover, it simply takes longer to make it into the box, giving defences more time to organise themselves.
It is not clear what Arsenal have internally pointed out as the cause. What is clear is that the Gunners have started shooting from distance slightly more (but still less than in 2014/15), with two stunning goals from new signing Granit Xhaka as a result.
In the game against Chelsea the true strength of this Arsenal team will be on show. Better decisions thanks to the use of data are nice, but ultimately Wenger will need performance on the pitch to keep the support of the critical Arsenal supporters.
Last edited: