Paris Saint-Germain v Olympique Marseille

220px-New_PSGMy first post in a while, so though I’d go a bit tactical. 150px-Olympique_de_Marseille_logo.svgConsidered an earlier season title battle, this promised to be an important game for both sides even with Marseilles changing their manager earlier in the season.

PSG Line-Ups 2

Paris Saint-Germain

PSG started the match with 4-3-3 consisting of Cavani, Di Maria and Ibrahimovic as the front 3. Cavani and Di Maria would exploit the pockets in and around the full-back and the centre-backs of Marseille. In addition, Ibrahimovic tried to come and receive the ball from the midfield players. Thiago Motta was the deepest midfield player dropping inbetween the centre-backs to receive the ball and attempt to play forward. Verratti acted as another defensive midfield when needed and then supported Matuidi who performed as the most attack minded of the 3. Both Aurier and Maxwell bombed on high up the pitch allowing Thiago Motta to drop in and receive. As well, Di Maria and Cavani could go in off the line looking for the ball.

For the benefits of the system, it has its drawbacks. Primarily, as a result of the three luxury players in the front 3. Cavani was the showed to be the main defensive assistance. But positionally clearly did not know what he was doing, so did not help much. As a result, this allowed Marseille to have some joy in the spaces between the Wide Forwards and the Full Backs. Or if the Full Backs had pushed on, even behind them. The midfield 3 provided good cover but were rarely tested by Marseille. At times, PSG were defending with 6, but because Marseille could not keep the ball. They failed to exploit the opportunities presented to them.

Olympique Marseille

Marseille on the other hand decided to start with a 4-2-3-1. This ideally meant that Cabella could act as a number 10 for the Batshuayi. Barrada and Alessandrini would provide some width or come in off the line for the full backs to overlap. Diarra and Silva would collect theMarseille Line-Up ball from the defence and then help break the lines to the more attacking players. This in principle was how it should have worked, but PSG stopped them playing out of the back. They then were forced to play more direct passes into the forward players. The full backs did offer much going forward so failed to exploit the spaces provided by PSG. As a result of these problems, Marseille struggled on the ball and could not form coherent attacks. Although the goal came from this system, it was from a deep cross and had to be perfectly executed.

In defence, problems came with the midfield. Cabella didn’t perform like an attacking midfield, he was more like a second striker. This then created a 4-4-2 when defending and PSG outnumbered Marseille in the middle of the park. The midfielders couldn’t lock on and so PSG passed the ball in and around the midfield. Also as the Batshuayi and Cabella did not press the PSG Centre Backs effectively, so it was easy to get out of the defence to the midfield. If Batshuayi and Cabella did not press at all, the PSG Centre Backs had loads of time to pick passes. All of these let Cavani, Di Maria and ibrahimovic do their thing.

After 71 minutes, Michel brought on Ocampos and Djadjedje for Barrada and De Ceglie. The system chanMarseille Line-Up 2ged from a 4-3-3. this meant it was man for man in the midfield. This allowed Marseille to have more possession of the ball and where able to challenge PSG all over the pitch. Djadjedje also put pressure on the left side of PSG’s system. When he pushed forward Pastore who came on for Ibrahimovic had to track back more. When he did not it enabled them to exploit the spaces inbetween the PSG players particularly down the right wing. Consequently, leading to a number of quality opportunities, which they did not take. PSG still had chances, which came as a result of Cabella again not tracking back or marking one of PSG’s midfield 3. This meant Diarra, Silva or Sarr (who came on for Silva) had to cover 3 players and were drawn out of position. This then effected the defenders who were also pulled out of position and so Pastore, Lavezzi and Cavani created problems near the end of the game.

To Conclude…

Although the final result was 2-1 to PSG, they should have won by a higher margin. PSG were not really troubled throughout the contest and during the match had high quality matches to put the game out of sight. If Marseille had started with a 4-3-3, I believe Marseille could have won the game. They would have been defensively better off as well as pose more of a threat in an attack. During the last 20 minutes in which Marseille had changed their system, they caused PSG’s problems and looked the better side. I’d question if the top clubs in france all played the right way against PSG, then there stranglehold on Ligue 1 would be broken. This was a missed opportunity for Marseille.

The ideas for diagrams came from @TomPayneftbl and


Borussia Dortmund???


So what has happened to Dortmund?

Although they are bottom of the bundesliga, their statistics are not all that bad.

From some summary statistics from, Dortmund apparently have a rating deserving 15th place. They have 17.1 shots per game. This is the 3rd highest in the league. They hold 54.4% possession, which is the second highest in the league. Their pass success percentage was 77.3%, which is 5th highest in the league. I am not sure how is rating the teams, but on these summary statistics they should be higher than 15th place.

People have been stating that their defensive performances have been just as bad as their attacking performances. They are conceding 9.6 shots per game. This is 3rd best in the league. They’ve got 20.7 tackles per game, which is 6th best in the league. Although this is not essential. The Interceptions per game dortmund has 15.8 per game. This is 3rd lowest in the league. And this is not essential for good performance. Dortmund have 12.2 fouls per game, which is 2nd lowest in the league As well they cause 3.8 offsides per game, which is the best in the league.

With their attacking statistics, as stated earlier they have 17.1 shots per game. They have 5.2 shots on target per game, which is 5th best in the league. There is 11.8 dribbles per game, which is 7th highest in the league. Dortmund players have been fouled 14.3 times per game, this is joint 10th in the league.

If we look at the areas Dortmund are shooting from its a different picture. With the 6 yard box dortmund has the worst percentage of shots in the league (3%). Then in the 18 yard box they have 51% of their shots which is joints 10th in the league. And 46% of their shots are outside the box which is 5th in the league. This clearly needs to improve. Dortmund have scored 14 goals this season. 8 have come from open play, 2 from counter attacks, 3 from set pieces and 1 own goal. If we compare this to Bayern, they have scored 32 goals. 23 from open play, 2 from counter attacks, 4 from set pieces, 2 penalties and 1 own goal. The big difference is their ability to score goals from open play.

Dortmund need to get further into the penalty area and score more from open play. My suggestion, even though it only is my opinion, would be to move the ball quicker. Have a faster passing tempo and try to penetrate the penalty area and the 6 yard box. I am sure soon enough we will see the Dortmund of old.

Performance Profiling

First post in a while and I apologise for that. So I have completed 3 months of my Masters course at the University of Chichester. One topic I have learnt about is performance profiling. This method takes me back to my childhood playing football manager games. The process was used on games as a way to analyse the attributes of the different players. This idea, I personally think is brilliant, it gives analysts and gamers a way to compare the performances of a team against the opposition performances within their league. They can be used to analyse certain areas of performance or the general performance of a team. So, for instance, the analyst might be looking to analyse their teams attacking performance. So they’d use shots, shots on target, crosses, possession, attacking third entries and penalty area entries figures.

Performance Profile Example

Performance Profile Example

From the above example, the different colours represent your teams performance at 25%, 50% and 75%. Green represents your teams 25% level, Blue represents your teams 50% level and Red represents your teams 75% level. So in this example we can see that this teams percentage of successful penalty area entries has a 25% level below 40. The 50% level is just above 60 and the 75% level is at about 80. Apart from these the other attacking areas for each level are at 100 or above 80. So in comparison to the other teams in their league, they are producing statistics better than any other team in their division. Where as their percentage of successful penalty area entries is in need of improvement. Their 25% level is one of the worst in their league. But their 50% and 75% levels are about average for every other team in their league.

I believe this method of analysing performance is very useful and the reason I have written this post about it is that I believe others should use it. Anybody could use it to do analyse on any team. Any coach could use it to analyse their own teams whether it is a sunday league team, a junior team or a semi-professional team. I believe that these profiles can point towards coaching topics for training for later in the season. This is during the season after your teams have played at least 8 games. This figure is used to get enough opposition performances to compare your own performances too.

Expected Goals v TSR(Total Shots Ratio)


Expected Goals is a calculation that is attached to each attempted shot and measures it’s chance of resulting in a goal. The method of calculation differs between models (and the iterations of those models), but they broadly take into account the same types of things. Such as, but not limited to, a shot’s distance from goal, what phase of play it came from, and even what body part the shot came off of. Theoretically, anything that effects a shot’s probability of resulting in a goal belongs in this model(Opta, 2014). This method creates a percentage possibility of scoring a goal from a certain area in the oppositions half, generally around and in the oppositions penalty area.

Another blog using expected goals finds that the top 4 from this season would achieve the same positions next season (Differentgame Blog, 2014). That prediction was using expected goals from last season. In the same method it predicts Sunderland, Aston Villa and West Ham will be relegated. In another prediction using the expected goals from the last three seasons Everton replace Arsenal in fourth position. Arsenal are predicted 7th in this method. In this method the teams predicted to be relegated are Sunderland, QPR and Burnley. The first method could be effected by the signing the teams bring in within the summer transfer window. The latter method produces a more valid prediction.

Expeted Goals Bin -

Expeted Goals Bin –

TSR is the total shot ratio. TSRx = (Total shots for)x/(Total shots for + total shots against)x. (James’ blog, 2012) From James’ brilliant work there is a critique of a number of methods to improve predictions within football. As well as these critiques he experiments and attempts to improve on previous methods. The total shot ratio method is an interesting method. The equation is easy to understand but I do question the need for the x figure. From England’s draw against Ecuador yesterday, the statistics create a value of 0.48 if the figure x is excluded. With the x figure, the value is 0.2304 if x is 2. Clearly, this is one game against James’ large data set.

In a blog piece by Chris Smith, the equation is TSR = shots for / (shots for + shots against) (Chris’ Blog, 2014). This does not have a squared value. In the piece, he uses the equation on league two teams. He highlights teams that do not fit the prediction value. We now know that Bristol Rovers were relegated from league two even though they had a TSR value of 0.50 on 10/02/14. In the blog chris’ league two table 10 of 12 teams in the top 12 have a TSR value of equal to or above 0.50. He then realigned the table to what the teams positions would be if they finished on the TSR value. Wycombe Wanderers would finished 9th and Bristol Rovers would have finished 15th. This, therefore, provides food for thought.

Total-Shot-Ratio –

To conclude with a comparison, it does not look like TSR is as reliable as Expected Goals. As research is further done in these areas, it would be compelling to find out more in these areas. It would also be interesting to find out whether the predictions come true from the expected goals prediction. We will have to wait until the end of next season.


Chris’ Blog (2014) Total Shots Ratio – TSR [Blog Entry] 10 February 2014. London: Chris’ blog. Available from [Accessed 5 June 2014]

Differentgame Blog (2014) A Shooting Model – An Exp(G)lanation and Application [Blog Entry] 19 May 2014 UK: Differentgame Blog. Available from [Accessed 5 June 2014]

James’ Blog (2012) Introducing TSR2.4 [Blog Entry] 2 December. James’ Blog. Available from [Accessed 4 June 2014].

Opta (2014) On the topic of Expected Goals and the repeatability of finishing skill. [Online] England: Opta. Available from [Accessed 4 June 2014].

Expected Goals: NHL Ice Hockey Example

After watching a presentation by Omar Chaudhuri from Prozone on expected goals at the Science and Football Conference 2014, I thought it would be interesting to research it. MacDonald wrote a paper on NHL Ice Hockey expected goals model (2012). MacDonald looks at previous performance indicators like Shot differential, Fenwick rating and Corsi rating. Shot Differential looks at a team’s average shots for versus the average shots allowed on a per game basis (Sporting Charts, 2014). Fenwick rating is the shots plus the missed shots (MacDonald, 2012). Corsi rating is the shots plus the missed shots plus the blocked shots (MacDonald, 2012). These performance analysis models in Ice Hockey were then compared in an ordinary least squares (OLS) model, ridge regression model and the Corsi rating with hits. The OLS model is a method for estimating the unknown parameters in a linear regression model. The ridge regression model is a model like least squares but shrinks the estimated coefficients towards zero (Tibshirani, 2013). The simple linear regression model is a model with a single regressor x that has a relationship with a response y that is a straight line (Montgomery et al., 2012).

MacDonald found that the OLS model and ridge regression model have higher correlations than goals, shots, Fenwick rating, Corsi rating and Corsi rating with hits (2012). These results are from the 2010-11 season. The OLS model and ridge regression model also have a lower mean squared error (MSE). The Corsi rating with hits is slightly lower compared to the goals, shots, Fenwick rating and Corsi rating. The Corsi rating with hits is not as low as the OLS and the ridge regression model.

The paper also looks into expected goals with adjusted plus-minus. Basically it involves expected goals per 60 minutes. The paper names Sidney Crosby as the highest adjusted plus-minus during even strength situations based on expected goals offensive component. Most of the NHL’s top offensive players are among the highest in this offensive component for example Sidney Crosby, Alex Ovechkin and Daniel Sedin.

In discussion MacDonald discusses that the teams more hits against than hits are the teams with higher goals. This can be seen in a situational sense but surely does not relate to goals scored of face-offs or goals scored when goalies are pulled. It is pointed out in the paper that the team performing the hits generally does not have possession of the puck. It also states that this is an indicator of the amount of possession a team has. Another reason is that once the player hits another player, the game becomes a 4 v 4 situation temporarily.

The article concludes that the ‘use of expected goals in a ridge regression to estimate adjusted plus-minus, coupled with the results based on goals, shots, Fenwick rating and Corsi rating, can be useful to NHL teams, analysts, and fans as they evaluate the performance of teams and players’ (MacDonald, 2012). This theory is one suggested for all people to gain insight into their sport. MacDonald’s article also presents an example of expected goals in another sport such as football. It also prevents an alternative model to predict events and player performance.



Macdonald, B. (2012). An expected goals model for evaluating NHL teams and players. In Proceedings of the 2012 MIT Sloan Sports Analytics Conference, http://www. sloansportsconference. com.

Montgomery, D.C., Peck, E.A. and Geoffery Vinning, G. (2012) Introduction to Liner Regression Analysis. 5th Edition. USA:Wiley.

Sporting Charts (2014) Team Shot Differential Per Game: 2013-14 NHL Season. [Online] USA: Sporting Charts. Available from [Accessed 04 June 2014]

Tibshirani, B (2013). Modern regression 1: Ridge Regression. [Online] USA: Carnegie Mellon University. Available from [Accessed 04 June 2014]

How Sevilla Won the Europa League 2014

Place of Penalties by Benfica and Sevilla in the Europe League Final 2014.

Place of Penalties by Benfica and Sevilla in the Europa League Final 2014.

Sevilla beat Benfica in the Europa League Final on the 14th May. The match was won on penalties. Sevilla scored all four penalties whereas Benfica only scored 2 of their 4 penalties. The original match was only a 0-0 draw during the original 90 minutes and extra time. Within the picture, you can see that the crosses stand for penalties scored and the minus mark signifies a penalty missed.

Sevilla’s penalties once struck towards the net averaged 0.36 seconds. Whereas, Benfica’s penalties averaged 0.38 seconds. This does not show much difference. A penalty could be struck as hard as possible and saved by the opposition goalkeeper. The difference is found in the run up of each penalty. Benfica averaged 2.983 seconds and Sevilla averaged 2.117 seconds. This is a difference of almost a second.



1. 2.60 seconds = Scored

2. 3.87 seconds = Missed

3. 3.47 seconds = Missed

4. 2.00 seconds = Scored



1. 2.07 seconds = Scored

2. 2.37 seconds = Scored

3. 1.87 seconds = Scored

4. 2.17 seconds = Scored


To me this shows that to score a penalty in a penalty shoot-out. The player needs to decide were they are going to place the ball and be decisive. This is how Sevilla won the Europa League 2014.

High Speed Ball Possession and Playing Position

Cristiano Ronaldo running with the ball.

Cristiano Ronaldo running with the ball.

As I am coming to the end of my university degree, it is the time for jobs. During my final year of university life, I undertook a dissertation. This dissertation looked into the work profiles of youth elite athletes and their playing positions. Through the search for jobs, I came across some academic work by Chris Carling, a performance analyst at Lille Football Club. The paper looks at the analysis of physical activity profiles when running with the ball in a professional soccer team (Carling, 2009).

If I was to compare my dissertation to his work, not only is it a lot better but it is an upgrade. Many authors have attempted to find differences and advantages through analysing elite player work profiles. Although my dissertation looked into youth elite football players, it looks into the distance covered by the players over 90 minutes of a football match. I found that forwards ran more than defenders over 90 minutes. In Carling’s paper it finds that elite players spend most of the time with the ball running at high speeds. The elite players spent 1.7% of the 90 minute match with possession of the ball. It would be interesting to analyse what the youth elite players in my study would have done with the ball in their possession. Carling states that “the ability to move at high speed with the ball seems to be an important facet of contemporary elite soccer and players across all positions should be able to carry out such actions” (2009). He then goes on to mention that people have tried to transfer this into training drills. He also states that this maybe is a way to improve performance if the elite players are able to perform at high speeds.

In the paper’s discussion he cites two other papers Carling et al (2008) and Stolen et al (2005) which both confirm a need for individualised training programs as the distances covered at different speeds vary according to play position. He then states that findings imply that fitness-training with and without the ball should be based on the specific requirements of each playing position. This is in agreement with my study, but would this be the same at youth level. Would players need to train at different speeds with and without the ball. Surely, this is just fartlek training with a ball.

Inconclusion, youth players looking to improve their football should learn to perform at high speeds with the ball at their feet. Training sessions should also look to train players with possession of the ball at high speeds but also without. But dependent on what position the child wants to player, they should train to the specifications of that certain position.