Barcelona vs AC Milan Passing Distribution (2nd Leg)

Some changes can be noticed clearly when the passing distribution in the 2 legs are compared. Here is the link to the first leg passing distribution.

1. With David Villa playing rather than Fabregas, Messi could drop deeper and work more cohesively with Xavi and Iniesta.
2. Moving Pedro to the left seems to be a fruitful decision too. First, Alves can work directly with Messi once again and at the same time Jordi Alba is “forced” to be more discipline defensively (i.e. doesn’t need to overlap very often as Pedro is already there). This impact can also be seen in AC Milan passing distribution in the two legs. See how Abate-Boateng thick link in the first leg is no longer present in the 2nd leg (replaced by Constant-Shaarawy link).

Euro 2012 Multidimensional Scaling Analysis

Based on OPTA’s data (via footytube), below is the multidimensional scaling plot based on per game data of passes, passes completed, touches, crosses, dribbles, corners taken, and offsides.

I was using Classic Multidimensional scaling, from wiki:

Multidimensional scaling (MDS) is a set of related statistical techniques often used in information visualization for exploring similarities or dissimilarities in data.

Also known as Principal Coordinates Analysis, Torgerson Scaling or Torgerson–Gower scaling. Takes an input matrix giving dissimilarities between pairs of items and outputs a coordinate matrix whose configuration minimizes a loss function called strain.

Based on the figure below, it is indicated that Germany and Netherland should have similarities, the same can be said with Ukraine, Czech, and Sweden. However, as far as my memory goes in watching the Euro 2012 games, I failed to notice how the teams are similar. In addition, what should be the label for the x and y dimensions?

Any comment?

Note: The input matrix in this case is as shown below and can be downloaded here.

I then used R to do the multidimensional scaling analysis.

Munchen vs Madrid Passing Distribution

Didn’t watch the match, but have been reading how Toni Kroos played an important role in the match. Based on the graph, he IS quiet central for Munchen in the match.

Some notes:

1. Why were there so many passes by Alonso to both central defenders Ramos and Pepe?
2. Centrao-Arbeola and Alaba-Lahm have similar passing distribution towards their central defenders (backward pass), but see how they differ in terms of forward passes.

Chelsea vs Barcelona Passing Distribution

Which side is Barcelona?

Chelsea vs Barcelona Passing Distribution

The number of passes by all Chelsea players: 194.
The number of passes by all Barcelona players: 779 (Xavi alone completed 127 successful passes and received 133 passes!).

Ball Possession: Chelsea 28% – 72% Barcelona.

But, the most important fact is still Chelsea 1, Barcelona 0.

Preview: Netherlands vs Spain

Netherlands:

Spain:

Few notes:

• This match may be more tightly contested than Spain vs Germany. This is because Netherlands, unlike Germany, does not have an over reliance towards a single player (like Schweinsteiger in Germany). So, it might be tougher for Spain to dominate the possession like it did against Germany.
• On another note, Netherlands past view goals have been due to Sneijder’s geniuses. So, if Spain can take care of Sneijder, Netherlands possibility to score should be minimized.

Germany Pass Network (aggregated from all matches)

Notes:

• To prevent the network becoming too cluttered, an edge is only shown when its weight is greater than or equal to 15.
• Thanks to JJ Merelo for the data.

Xavi pass-network (ESP 0:1 SUI)

Throughout the five games Spain has played in the 2010 FIFA World Cup, Xavi has been one of its most important player. He has passed the ball more than anyone else in his team, and also more than anyone else in the World Cup (up to quarter final stage). He has passed the ball 464 times (92 times/game) with a passing accuracy of 80%.

If we look at the authority centrality in Spain team, Xavi has always been among the top 2 most central player in his team. This is the definition of authority-centrality from ORA:

A node is authority-central to the extent that its in-links are from nodes that have many out-links. Individuals or organizations that act as authorities are receiving information from a wide range of others each of whom sends information to a large number of others. Technically, an agent is authority-central if its in-links are from agents that have are sending links to many others. The scientific name of this measure is authority centrality and it is calculated on agent by agent matrices.

In other word, Xavi is authority central because he received passes from other players who passes the balls a lot (i.e. played the distributor role in the team).

Let’s visit Xavi passing game during Spain’s first game against Swiss (which it lost).

*The size of the jersey indicates the authority-centrality of each player.

The thickness of the line indicates the frequency of passes between Xavi and the other player. Very interestingly, Xavi only passed the ball successfully to David Villa 3 times throughout the whole game. This fact is even more surprising since Torres only came into play in minute 61 replacing Busquets (note: Xavi did not have any successful pass towards Torres in this game). As a comparison, in Spain subsequent games where Torres played the game from the beginning, Xavi had 5 completed passes towards Villa (vs Honduras), 5 (vs Chile), 18 (vs Portugal), and 8 (vs Paraguay). And successful passes towards Torres are: 3 (vs Honduras), 1 (vs Chile), 5 (vs Portugal), 2 (vs Paraguay).

It can also be observed the preference of Xavi to pass towards Iniesta rather than to David Silva. And also he passed the ball more often to Sergio Ramos rather than to David Silva. David Silva is definitely a competent player, but it doesn’t seem he meshed well during the match. And Xavi (one of Spain most influential player) seemed not very comfortable playing along with him (indicated from the number of passes among them). Guess what, this match against Swiss was the first and also the last game (upto quarter final) for David Silva. He did not even enter other matches as a substitute.

There could be other reasons why David Silva has not played since the match against Swiss. But this could be an analogy in an organization setting. Talent alone may not guarantee a place in an organization, one should make sure he/she can mesh together with other team member (especially with the influential people in the organization.)

Germany pass-networks (against Australia) – 2010 FIFA World Cup

Germany won its first game in 2010 FIFA World Cup against Australia 4:0.

Inspired by what FAS.research did in visualizing the pass-network of soccer matches (there are a number of matches from the 2010 FIFA World Cup as well), here is the Germany pass-network against Australia.

Here are the number of passes completed by each German players:

• Philipp LAHM – 71 pass(es)
• Arne FRIEDRICH – 68 pass(es)
• Per MERTESACKER – 64 pass(es)
• Bastian SCHWEINSTEIGER – 54 pass(es)
• Holger BADSTUBER – 47 pass(es)
• Sami KHEDIRA – 41 pass(es)
• Thomas MUELLER – 34 pass(es)
• Manuel NEUER – 24 pass(es)
• Mesut OEZIL – 24 pass(es)
• Lukas PODOLSKI – 15 pass(es)
• Miroslav KLOSE – 7 pass(es)
• CACAU – 5 pass(es)
• Mario GOMEZ – 1 pass(es)
• Marko MARIN – 1 pass(es)

The defenders contributed 250 out of 456 passes (54.8%)!

And the following are the number of passes received by each player:

• Arne FRIEDRICH – receive 63 pass(es)
• Per MERTESACKER – receive 60 pass(es)
• Bastian SCHWEINSTEIGER – receive 59 pass(es)
• Philipp LAHM – receive 58 pass(es)
• Thomas MUELLER – receive 42 pass(es)
• Sami KHEDIRA – receive 40 pass(es)
• Lukas PODOLSKI – receive 34 pass(es)
• Holger BADSTUBER – receive 33 pass(es)
• Mesut OEZIL – receive 31 pass(es)
• Miroslav KLOSE – receive 14 pass(es)
• Manuel NEUER – receive 12 pass(es)
• CACAU – receive 10 pass(es)
• Mario GOMEZ – receive 6 pass(es)
• Marko MARIN – receive 4 pass(es)

Passes to defenders 45.9%, midfielders (Schweinsteiger, Khedira, Oezil and Gomez) 29.2%, forwards (Klose, Cacau, Podolski, Marin and Mueller) 22.3%.

Source of inspiration: FAS Research

A simple network analysis on hedge fund holdings

I’ve been following the website Financial Network Analysis lately. Wanting to get my hands dirty, I tried to generate simple network map of hedge fund holdings.

Based on the hedge fund portfolio tracker compiled by marketfolly, I constructed a network map of some stocks which are held by at least two hedge funds.

Here are the list of hedge funds tracked:

• Appaloosa Management (David Tepper)
• Baupost Group (Seth Klarman)
• Berkshire Hatheway (Warren Buffett)
• Blue Ridge Capital (John Griffin)
• Bridger Management (Roberto Mignone)
• Conatus Capital (David Stemerman)
• Fairholme Capital Management (Bruce Berkowitz)
• Greenlight Capital (David Einhorn)
• Harbinger Capital Partners (Phil Falcone)
• Lone Pine Capital (Stephen Mandel)
• Maverick Capital (Lee Ainslie)
• Pabrai Investment Fund (Mohnish Pabrai)
• Paulson & Co (John Paulson)
• Pershing Square Capital Management (Bill Ackman)
• RBS Partners (Eddie Lampert)
• Shumway Capital Partners (Chris Shumway)
• Soros Fund Management (George Soros)
• Third Point LLC (Dan Loeb)
• Tiger Global (Chase Coleman)
• Tiger Management (Julian Robertson)
• Viking Global (Andreas Halvorsen)

The red nodes are the fund managers for each hedge fund. The bigger the size of the node of a stock (blue nodes) means that the stock are held by more hedge funds. Here is the network map (click on the image for larger size):

As can be seen Apple, Wells Fargo, and Visa are held by the most hedge funds (7 to be exact).

Directorship Interlocks among US 51 Largest Companies (by Market Caps)

I tried to map the directorship interlocks among the companies.

The data sources are from:

• Finviz (I used the screener to find the largest companies – as per 4 February 2010)
• Reuters (the list of directors for each company)

Software used:

• Ucinet (For the creation of affiliation matrices derived from the DL file)
• ORA (For the visualiation)

The results are here:

• Directorship Interlocks Map – the larger the node size the higher the betweenness of the company
• List of Directors & Company Abbreviations List (http://www.box.net/shared/d1nant9r9a)

The top 10 companies with the highest betweenness centrality are as follows:

• IBM – INTL BUSINESS MAC
• BA – BOEING CO
• GE – GEN ELECTRIC CO
• WFC – WELLS FARGO & CO
• UTX – UNITED TECH
• JPM – JP MORGAN CHASE C
• DIS – WALT DISNEY-DISNE
• MCD – MCDONALDS CP
• BRK – BERKSHIRE

I will come up with more statistical measures of the network in the subsequent post.