Thursday, November 1, 2018

The new FIFA calculation: a brief comparative analysis (II)

See here for my previous post about the new FIFA calculation method. It explains how I came by the set of numbers used here.

There are suspicions that the number of points to be gained per match in relation to the points gaps that will occur in the ranking when using the new method for a prolonged period, will result in too little chances to improve a team's ranking in an relatively short amount of time. The previous post showed that the climb in ranking positions for 'ín form' teams indeed was (much) slower than in the 'old' FIFA-ranking.


In this part I will focus on the points and the points gaps of the June 2018 ranking compiled with the new method. And compare these with the corresponding elo ratings.

The classic elo-rating has the advantage that the goal difference in a match plays a magnifying role in the number of points a team can gain in a match. 
The new FIFA-method doesn't have this advantage. FIFA's method on the other hand gives extra advantage to teams (aka the big shots) reaching the knock-out stages of important tournaments: they can't lose points anymore in these knock-out matches. The vast majority of the 211 teams in the ranking however shall not profit from this rule/advantage.

First a plot with the point-gaps in the June 2018 ranking, in comparison between the new method and elo. The x-axis shows the ranking position at which the gaps occur.


You can see that the points-gaps at the top end of the ranking are equally big for both methods. A large part of the ranking sees relatively small gaps (5-15 points), also for the new FIFA method. Towards the bottom end of the ranking the gaps grow bigger again.

The frequency distribution of the gaps shows that the elo-ranking even has slightly bigger gaps than the new FIFA-method:




The earned and lost points in the eight years between 2010 and 2018 are plotted for both methods in a frequency distribution:



In this graph you can see the fairly symmetrical distribution of the elo-method. The new FIFA-method clearly shows a skewed distribution. Especially the bin +50 to +100 points is overrepresented.
The average points difference over all teams in this method is 12, while for elo this average is, as expected, 0.

The top-25 of the June 2018 ranking calculated with the new FIFA-method:


1 Germany 1906
2 Brazil 1863
3 Spain 1815
4 Argentina 1809
5 Portugal 1762
6 France 1749
7 England 1708
8 Mexico 1699
9 Chile 1698
10 Colombia 1694
10 Netherlands 1694
12 Belgium 1688
13 Italy 1672
14 Uruguay 1641
15 USA 1638
16 Switzerland 1625
17 Croatia 1610
18 Peru 1607
19 Costa Rica 1583
20 Poland 1556
21 Iran 1543
22 Australia 1540
23 Sweden 1532
24 Ghana 1527
25 Japan 1521

The teams that performed well in the knock-out stages of the major tournaments in the 8 years  between June 2010 and June 2018 surface on top. Points-gaps at the very top are relatively big (around 30-40 points each). From spot 7 downwards the gaps become smaller and smaller.

Finally, the effects of the World Cup 2018. The top-25 of the August 2018 ranking with points and in the last column the position in the actual August ranking:

1 Brazil 1892 3
2 France 1873 1
3 Germany 1842 15
4 Belgium 1825 2
5 Spain 1811 9
6 Argentina 1774 11
7 Portugal 1766 7
8 England 1747 6
9 Croatia 1726 4
10 Uruguay 1720 5
11 Mexico 1710 16
12 Chile 1698 12
13 Colombia 1697 14
14 Netherlands 1694 17
15 Italy 1672 21
16 Switzerland 1651 8
17 USA 1639 22
18 Sweden 1607 13
19 Peru 1585 20
20 Denmark 1566 9
21 Iran 1562 32
22 Russia 1558 49
23 Costa Rica 1543 32
24 Poland 1537 18
25 Ghana 1526 45

Doesn't look too farfetched now, does it ? Brazil on top, not far before France. Belgium climbing further up. Germany and Argentina not so heavily punished. Croatia on a more normal spot 9. Denmark and Russia idem dito on spot 20 and 22.

General conclusion: the points distribution and the points gaps distribution of the new method after 8 years resembles the characteristics of the elo-rating. And the elo-ranking is known for it's inertia.
The old ranking could be severely manipulated by not playing friendlies, Wales and Romania have proven that in the recent past. This new ranking method at least does not have that major drawback anymore. 

Maybe the I-factors and/or the scaling factor must/shall be tweaked after each World Cup (like the former Confederation coefficient), who knows. For now we will have to deal with this method as it stands.

About me:

Software engineer, happily unmarried and non-religious. You won't find me on Twitter or other so called social media. Dutchman, joined the blog in March 2018.

1 comment:

  1. Excellent analysis. The way I see it, it is actually the use of faulty old rankings to start the new rankings, which is making things even worser. Combining that with slow movement means that teams like Wales will continue to be rewarded for manipulating the system for a long time. At the same time, teams like Russia, Japan, South Korea, Ghana, Iran would remain in their unfair low rankings for years and can only move up slowly.

    ReplyDelete