Tag Archives: Soccer Power Index

Forecasting 2018 FIFA World Cup in Russia

This a short post to share my forecast for the coming football 2018 FIFA World Cup to be played during the following five weeks in Russia.

As I introduced in a similar post four years ago for the 2014 World Cup (here), I have a work colleague who not only is a tremendous aircraft salesman but also has a great sense of humor and manages in his free time late in the night to set up a contest for office staff to try to guess winners, matches’ scores, top scorers, etc., of major international soccer competitions. The 2018 FIFA World Cup in Russia, which will start this week, could not be missed. Nacho managed to set up the contest in time.

I have approached the game of forecasting this World Cup with the same method as previous times, as I have not watched a single match of national teams’ football since the previous World Cup and I have no clue of who is who and how they come to the competition. I have relied on ESPN rankings, and used its offensive and defensive coefficients to build with a simple algorithm all the scores of the competition:

• taking into account the coefficients of both sides
• when the difference between them was narrow, I put a draw, if resulting coefficients were high 2-2, if low, 0-0.
• the same for victories, if the difference was high 3-0, if small and low coefficients, 1-0.
• I also checked the numbers that different scores were repeated in the group phase in the previous two World Cups, as the most repeated ones are 2-1, 1-0, 0-0, in order to assign them in similar proportion.
• I also checked the amounts of goals scored in the group phase of previous 2 World Cups (100 and 136 goals), to adjust the overall number of goals I would distribute.
• Checked the goals the previous top scorers managed in World Cups to a assign a similar number.

What did I forecast?

• A World Cup won by Brazil against Spain in the final in the penalty shootout.
• An oddity: the algorithm provided that Spain would face 3 shootouts in this World Cup, we will see.
• The forecast also provided that England would beat Colombia in a shootout, that may be even odder, given the historical bad luck of England at shootouts.
• As top scorer I put Neymar (Brazil) with 6 goals.

2018 FIFA World Cup Russia forecast.

Filed under Sports

Forecasting France Euro 2016

I have a work colleague who not only is a tremendous negotiator and aircraft seller but also has a great sense of humor and manages in his free time late in the night to set up a contest for office staff to try to guess winners, matches’ scores, top scorers, etc., of major international soccer competitions. The France Euro 2016 which starts this afternoon could not be missed. Nacho managed to set up the contest in time.

In this post I am going to explain how I went about forecasting the results of the UEFA Euro 2016.

“when in doubt, build a model”, Nate Silver.

The readers of this blog may already know how much I do like to build models to produce forecasts, guesstimates, etc. In relation to forecasting this UEFA Euro 2016 there is some background that has shaped my mind in relation to the subject in the recent years, let me give you some hints:

Having shared this background, you may understand that I tried to remove all the beauty of guessing and my football knowledge out of the forecasting process (1).

• ESPN Soccer Power Index (SPI) ranking, introduced by the economist Nate Silver. I used its offensive and defensive scores plus weight for each of the scores based on a tip indicating that in competitive matches the defensive factor tends to be slightly more important (see “A Guide to ESPN’s SPI rankings”) (2).
• The frequency of different scores in the group phases of the Euro 2012 and the World Cup 2010, the in the round of 16, quarter finals and semi-finals.

• A few simple rules about how to allocate results given the difference between SPI ratings of the two nations playing each match. (3)
• The total number of goals during group phases the latest Euro and World Cup. In order to cross check that the total numbers of goals that my forecast yielded was in check with previous competitions.

It may sound very complex. It is not. It requires a bit of reading (which most of it I did years ago), retrieving the latest ratings, giving it a bit of thought to set up the model and then, not even looking at the names of the teams, you go about allocating the scores based on raw figures. Let’s see how my forecast fares this time! (4)

Les grandes personnes aiment les chiffres” (5), the Little Prince.

(1) In fact I have not watched a single national team football match from any country since the World Cup in Brazil in 2014.

(2) See here the blog post I published yesterday in which I made a more thorough review of the ESPN SPI index.

(3) I set up rules like “if the difference of the combination of indices of the two nations is below this threshold, I take it as a draw, if it is between x and y as victory by 1 goal, if higher…”, etc.

(4) This way of forecasting allowed me to finish 4th out of 47 in 2010, 15th out of 87 in 2014. As it removes biases it allows to be better than the average, though it prevents you of guessing outliers, gut feelings, etc.

Note: In the blog post from yesterday I mentioned that the latest complete ranking from the ESPN SPI index that I could retrieve dated from October 2015. That is the one I have used, therefore, Germany results as winner. Of the latest ranking, covering the Top 25 nations, only 13 countries of the 24 competing at the Euro 2016 are included. I could have set up an hybrid ranking taking the latest rankings and ratings for the top 13 from June and using the October figures for the lower 11 teams. I decided to go on with a single set of data. If I had done so, the maing changes would have come from the semifinals onwards. France would have appeared as winner instead of Germany. We’ll see if that was a good decision.

France Euro 2016: “group of death”?

Tomorrow will start the UEFA Euro 2016. Fans all over Europe start getting excited by it. This year’s competition is played in France, with some matches taking place in Toulouse, one of them Spain – Czech Republic, which some friends and I will be able to watch live!

This post is intended to be a quick one to discuss, as I did for the 2014 World Cup in Brazil, which groups are the most difficult ones, the so-called “group of death“. Media all over Europe states that it is group E, with Belgium, Ireland, Italy and Sweden the one which is the toughest. To discover which is effectively such group I’ll focus on a couple of rankings: FIFA’s and ESPN’s Soccer Power Index, as I did in 2014.

In its website, FIFA explains the procedure which it uses to compute the ranking, which is based on the following formula:

M x I x T x C = P

M: winning, drawing or losing a match

I: importance of the match

T: strength of opposing team

C: confederation strength weights

P: points for a game

According to that formula, the latest ranking (June 2nd), filtered for European teams, has the following teams at its top:

With the information of both the ranking and the points I went to check which of the groups of the Euro 2016 were the strongest, both taking a look at the overall group and looking from the perspective of the “favourite” team (the one with the highest ranking), which was the one facing a toughest group (total points of the other 3 teams composing the group). See the results below:

As you can see the most difficult groups in terms of total points are:

• C (Germany, Northern Ireland, Poland, Ukraine) with 3,897.
• F (Austria, Hungary, Island, Portugal) with 3,895.
• E (Belgium, Ireland, Italy, Sweden) with 3,869.

Looking at the average ranking, the most difficult groups are:

• F (Austria, Hungary, Island, Portugal) with 18.
• C (Germany, Northern Ireland, Poland, Ukraine) with 18,75.
• D (Croatia, Spain, Czech Republic, Turkey) with 20,25.

And excluding the points of the favorite team in each group, which is the favorite facing the toughest group?

• Portugal in group F, facing 2,714.
• Germany in group C, facing 2,587.
• Spain in group D, facing 2,576.

Then, combining the 3 approaches, to me, it becomes clear that the toughest group is F, with Austria, Hungary, Island and Portugal, by the total amount of points (2nd), ranking of the teams (1st) and in relation to what Portugal will face (1st).

The second most difficult group would be C, with Germany, Northern Ireland, Poland and Ukraine, by the total amount of points (1st), ranking of the teams (2nd) and in relation to what Germany will face (2nd).

You can see that, using FIFA ranking, and despite of conventional “wisdom” (press), group E would be nothing but the 3rd or 4th most difficult group, i.e. an average group out of 6.

ESPN Soccer Power Index (SPI) ranking was introduced by the economist Nate Silver of worldly fame, who many readers will know from his forecasts on elections in the USA (check his blog FiveThirtyEight).

In a post from 2009, when the SPI was introduced, just before the 2010 World Cup, he explained how the index was computed (“A Guide to ESPN’s SPI rankings”). As he explained, the process had 4 main steps:

• Calculate competitiveness coefficients for all games in database.
• Derive match-based ratings for all international and club teams.
• Derive player-based ratings for all games in which detailed data is available.
• Combine team and player data into a composite rating based on current rosters; use to predict future results.

The main difference in relation to FIFA ranking algorithm is that it takes player-based ratings for those players who play in clubs in the Big Four leagues (England, Spain, Italy, Germany) and the UEFA Champions’ League. The player-based rating is merged into the national team coefficient. The player-based rating weighs heavily in national teams with many players playing in the main leagues (e.g. England or Spain national teams) and less heavily in other nations which roster is composed of many players not playing in clubs of the 4 main leagues (e.g. Russia).

Other details of the ESPN’s approach are similar to those used by FIFA: e.g. giving weights to results depending on the opponent, measuring the competitiveness of the match, the different confederations, etc.

ESPN provides a daily update of its ranking, however it includes only the top 25 world-wide teams, out of which 15 are European and only 13 represented in the UEFA Euro 2016, that is about half of those 24 competing.

In order to review which one would be the group of death using the ESPN SPI I took the latest available complete ranking I could find, dating from October 2015, which is half a year away, but reflected the situation at about the end of the qualifying phase for the Euro 2016. See the ranking below:

As I did with the FIFA ranking above, with the information of both the ranking and the ESPN SPI ratings I went to check which ones of the groups of the Euro 2016 were the strongest, both taking a look at the overall group and looking from the perspective of the “favourite” team (the one with the highest ranking), which was the one facing a toughest group (total ratings of the other 3 teams composing the group). See the results below:

As you can see the most difficult groups in terms of total ratings are:

• D (Croatia, Spain, Czech Republic, Turkey) with 309.
• B (Slovakia, Wales, England, Russia) with 307.
• C and E with 303.

Looking at the average ranking, the most difficult groups are:

• B (Slovakia, Wales, England, Russia) with 24.
• D (Croatia, Spain, Czech Republic, Turkey) with 24.5.
• E (Belgium, Ireland, Italy, Sweden) with 28.

And excluding the points of the favorite team in each group, which is the favorite facing the toughest group?

• England in group B, facing 224.
• Spain in group D, facing 223.
• Belgium in group E, facing 219.

Then, combining the 3 approaches, to me, it becomes clear that the toughest group is B, with Slovakia, Wales, England and Russia, by the total amount of points (2nd), ranking of the teams (1st) and in relation to what England will face (1st).

The second most difficult group would be D, with Croatia, Spain, Czech Republic and Turkey, by the total amount of points (1st), ranking of the teams (2nd) and in relation to what Spain will face (2nd).

You can see that, using ESPN SPI ranking (from October), and despite of conventional “wisdom” (press), group E would be nothing but the 3rd most difficult group.

Some readers may be tempted to think that I arrived at this result because I used a ranking from half a year ago, that if we were to take the latest ratings (if fully available) group E would emerge as the toughest one… not so. See the preliminary table using the information available for those 12 teams:

There you can see that with the latest ratings the most competitive group would be either D or C, very much like with FIFA rating (from June as well).

It is interesting to note how by using FIFA or ESPN SPI the weight given to the group F (Portugal) is completely different.

Finally, in both ratings the big absence in the tournament is the Netherlands, arguably about the 10-14th team in the world, the 6th in Europe. A pity for the competition.

1 Comment

Filed under Sports

Forecasting 2014 FIFA World Cup Brazil

I have a work colleague who not only is a tremendous negotiator and contracts’ drafter but also has a great sense of humor and manages in his free time late in the night to set up a contest for office staff to try to guess winners, matches’ scores, top scorers, etc., of major international soccer competitions. The 2014 FIFA World Cup in Brazil, which will start tomorrow, could not be missed. Nacho managed to set up the contest in time.

To set up the background as to how I have approached the game of forecasting this World Cup:

• I had written a review of the book “Soccernomics“, which among other things advocates the use of data in order to make decisions in relation to football transfer market, forecasting, etc. This book relies somewhat heavily in “Moneyball” another book which I read some months ago with a similar scope but with baseball as the theme sport.
• When the draw of the World Cup took place last December, I wrote a couple of blog posts discussing what was the so-called “group of death” basing the analysis on FIFA and ESPN rankings.
• During the last year, I read a couple of books which approach how we make decisions and how to remove different kind of biases from the thought processes of making them: “Thinking Fast and Slow” (by the 2002 winner of the Nobel Prize in Economics Daniel Kahneman) and “Seeking Wisdom“.
• Finally, last year I followed the open course “A Beginner’s Guide to Irrational Behavior” by Dan Ariely (though I missed the last exam due to my honeymoon and could not get credit for it).

Having shared this background, you may understand that I tried to remove all the beauty of guessing and my football “knowledge” to the forecasting process. I rather made use of  ESPN Soccer Power Index (SPI) ranking, introduced by the economist Nate Silver. I used its offensive and defensive scores plus the tip indicating that in competitive matches the defensive factor tends to be slightly more important (see “A Guide to ESPN’s SPI rankings”).

Once I plugged in the numbers from the index and used the referred tip on the defensive side, I built a simple model to guess each of the World Cup matches. Once you take this approach you will find that the model gives you plenty of results such as Nigeria 1.32 – 1.53 Bosnia… What to do with it? When the result was very tight I resolved it as a draw, otherwise a victory for the team with the highest score.

In very few instances I forecast that a team would score 3 or more goals in a match. I bore in mind that in the 2010 World Cup 80% of the matches ended up with scores of 1-0 (26% of the matches), 2-1 (15%), 0-0, 1-1 or 2-0 (each 13%).  That a team scores more than 3 goals in a match will certainly happen in some games, but I did not bother to guess in which ones, the odds are against.

The prize pot of the game organized by this colleague is not particularly big (few hundreds euros). The main point of the game is enjoying the chit-chat with work colleagues. My second main point is putting this rational approach to work and see how it fares.

Finally, what did I forecast?

A World Cup won by Brazil against Argentina in the final. With Spain beating Germany for the third place (in the penalties). For my English readers: England defeated by Colombia in the 1/8 of final. For the ones from USA, it doesn’t make the cut from the group phase. We will see along this month how well do I fare.

2014 FIFA World Cup Brazil forecast.

1 Comment

Filed under Sports

Brazil 2014 FIFA World Cup: “group of death”? (using ESPN ranking)

In a previous blog post I used FIFA world rankings to see which was the “group of death” of the following Brazil 2014 World Cup finals.

I received some comments questioning FIFA ranking based on the position of some specific countries: Switzerland, Portugal, Argentina, Colombia, Chile… I am sure that when one looks at how each country is playing he will believe that this or that country plays much better than the other placed higher in the ranking. But, the goodness of the ranking is that it removes perceptions from the process and simply establishes a set of rules by which all teams are going to be measured. It then goes on computing teams’ results along the year and the positions in the ranking are established, for good and bad.

In one of the comments I received I got the suggestion to rather use ESPN Soccer Power Index (SPI) ranking. I was even more attracted to that hint as the ESPN SPI index was introduced by the economist Nate Silver of worldly fame, who many readers will know from his forecasts on recent elections in the USA (check his blog FiveThirtyEight).

In a post from 2009, when the SPI was introduced, just before the 2010 World Cup, he explained how the index was computed (“A Guide to ESPN’s SPI rankings”). As he explained, the process had 4 main steps:

• Calculate competitiveness coefficients for all games in database
• Derive match-based ratings for all international and club teams
• Derive player-based ratings for all games in which detailed data is available
• Combine team and player data into a composite rating based on current rosters; use to predict future results.

ESPN SPI ranking at the end of Nov 2013.

The main difference in relation to FIFA ranking algorithm is that it takes player-based ratings for those players who play in clubs in the Big Four leagues (England, Spain, Italy, Germany) and the UEFA Champions’ League. The player-based rating is merged into the national team coefficient. The player-based rating weighs heavily in national teams with many players playing in the main leagues (e.g. England or Spain national teams) and less heavily in other nations which roster is composed of many players not playing in clubs of the 4 main leagues (e.g. Russia).

Other details of the ESPN’s approach are similar to those used by FIFA: e.g. giving weights to results depending on the opponent, measuring the competitiveness of the match, the different confederations, etc.

You can see the top ranked countries at the picture above.

Without entering on whether this or that country is far better placed in one or the other ranking based on perceptions, one simple yardstick to measure them is to see how many of their 32 top countries are not among the 32 countries qualified for the World Cup:

• FIFA ranking: 7 teams among the top 32 are not in the World Cup: Ukraine (18), Denmark (25), Sweden (27), Czech Republic (28), Slovenia (29), Serbia (30) and Romania (32). All coming from Europe, and not qualified for the World Cup due to the limited amount of places for UEFA countries (they all placed 2nd or 3rd in their groups).
• ESPN SPI ranking: 6 teams among the top 32 are not in the World Cup: Paraguay (19), Serbia (20), Ukraine (21), Peru (27), Sweden (29) and Czech Republic (30). 4 countries from Europe and 2 from South America, out for the same reason. Here however, Paraguay is still placed 19th despite of being the last country of the CONMEBOL qualifying.

With the information from the ESPN SPI ranking I produced the same table:

Brazil 2014 groups heat map based on ESPN SPI ranking.

And then, the same analysis as in my previous post follows.

The most difficult groups in terms of total ratings are:

1. B (Spain, Netherlands, Chile, Australia) with 327.
2. D (Uruguay, Costa Rica, England, Italy) with 323.
3. G (Germany, Portugal, Ghana, USA) with 322.

Looking at the average ranking, the most difficult groups are:

1. D (Uruguay, Costa Rica, England, Italy) with 14.
2. G (Germany, Portugal, Ghana, USA) with 15,25.
3. B (Spain, Netherlands, Chile, Australia) with 17,5.

And excluding the rating of the favorite team (pot 1) in each group, which is the favorite facing the toughest group?

1. Uruguay in group D, facing 239.
2. Spain in group B, facing 238.
3. Germany in group G, facing 234.

Then, combining the 3 approaches, the toughest group is between B (in terms of combined ratings) or D (in terms of average rating and from the favourite point of view).

Using the ESPN ranking group G would definitely would not be the toughest one, but the 3rd toughest.

I would understand ESPN journalists calling group B or D the toughest one. What strikes me is why FIFA website content editors call group B the “group of death” if by their ranking that group would be the group G!

It will be interesting to see how one ranking fares against the other at the time of predicting the actual development of the Brazil 2014 World Cup.