Forum

xml	No replies possible in the archive

Author: Ricardo
Date: 04-09-2006, 09:15

Bert is Fifa ranking a allowed subject on this forum? It's about coefficients, but not about the Uefa?
Not many are interested in this, due to which the previous topic is already archived.
I have tried to calculate the Dutch coefficient, but failed. I only came to 1308/1311 when it should have been 1322.
But well now I came to another strange conclusion.

The last 4 years we got average coefficeints of (according to my (incorrect) calculations until August 1st 2006):
Last year: 693
year before:740
before: 412
before:647

And after August 1 we have played 2 matches:
1. a win in a friendly agaisnt Ireland: 483 points
2. a win(hardly) in a qualifier against Luxemburg: 37.5 points

While the friendly would not help us getting ourt coefficient higher, Luxemburg is really bringing it down: If Holland would have lost, it would hardly have had any consequences on the ranking. We need a win from Belarus(worth 982.5 points) to not lose too much ranking points.

Author: ferdi
Date: 04-09-2006, 10:07
Edited by: ferdi
at: 04-09-2006, 11:47

I think I have detected an error in your calculations. The qualifier vs. Luxemburg gives 375 pts, not 37.5 pts.

Generally it is true that friendlies should be avoided because you cannot get more than 600 pts, which will always lower the coefficient for a top-10 country. Playing countries like Luxemburg, Andorra, San Marino in a qualifier is even worse, because the maximun number of points is 375 for these matches. However as every top country in europe has to play such a minnow twice every two years in the qualifiers, the effect will even out.

Maybe it would be better to have three accounts for each country, one for final WC and continental tournaments, one for qualifiers, and one for friendlies. The coefficient could then be something like 0.5 * (av. final tournament pts.) + 0.35 * (av. qualifier pts.) + 0.15 * (av. friendly pts.)

Author: bert.kassies
Date: 04-09-2006, 10:35

@Ricardo,

1) This subject is very much on-topic and welcome on this forum. As all other ranking methods used in football. Discussing the merits of UEFA coefficients we should be informed on other rankings.

2) The previous topic is archived because it was started in July. All topics started in July are archived at the beginning of September. No matter how many replies. No matter the last date of reply.

Re: FIFA Ranking
Author: Ricardo Date: 04-09-2006, 16:05	The calculation mistake is because there is a minimum measurement for a team of 0.50. I counted Luxemburg(pos. 195) as (200-195)/100 = 0.05 then multiplying this with 2.5(qual) and 3(win) and 100 gives 37.5. using .50 gives indeed 375. phew - not that bad as I feared.

Author: panda
Date: 04-09-2006, 17:27

OK, how accurate does FIFA ranking need to be?

What is it used for?

Is it just for seeding the pots in WC qualifiers, and deciding who are the top seeds in WC finals?

Excuse my ignorance, but I have a feeling Uefa seed EC finals on past performance in EC. What about seeding the pots for EC qualifiers, is that also on FIFA ranking?

We agreed that in the past, there have been some strange anomalies (USA = 5th), but can we say that this ranking system had any bad consequences (I think we thought the WC groups were not unfair)? For a few countries? Specific examples?

Because for sure one of the main subjects of debate for the Uefa club rankings is: is it fair? Is it fair to everybody?

I mean, obviously an EC group with Italy France and Ukraine seems like a group of death, but maybe it is just bad luck (for Scotland etc)?

Author: Michele
Date: 04-09-2006, 17:41

For the past many years, European qualifying groups have been based on results from the previous two qualifying tournaments. The method is simply to take the total number of points and divide it by the total number of games played. So to make significant moves in this ranking, you need to play two good qualifying cycles in a row.

The reason why Italy, France, Ukraine (3 of the 8 best teams in the world) and Scotland (best team in the world in their own opinion

) is that Italy played less than optimal in the 2004 and 2006 qualifiers and are only 2nd seeded, Ukraine weren't very good for 2004 and 2006 couldn't give them a better seeding than 3rd and Scotland really have been no good in the period in question.

Re: FIFA Ranking
Author: panda Date: 04-09-2006, 17:44	Thanks Michele - so the interest in correctness of FIFA ranking (as I think was stated on a previous thread about this topic) is above all - Why have a system of rankings if it is not accurate? The practical effects are secondary.....?

Re: FIFA Ranking
Author: Michele Date: 04-09-2006, 18:03 Edited by: Michele at: 04-09-2006, 18:05	Yeah, pretty much... On Forum 2, there is a thread debating the seedings for WC2010 as they progress Forum 2 Unfortunately, only 2 different systems are used in order to calculate the seedings, so the thread could easily use some more attention.

Author: MartinW
Date: 05-09-2006, 04:00

I have also tried to calculate the FIFA Ranking for England, my calculation for August 2006 ranking is 1,425 but FIFA amount is 1,434. Same as Ricardo's problem with Netherlands ranking, it is a small difference but difficult to understand how it happens.

I think that we are making the calculation properly, but the main problem lies in the cut-off dates used. For example, see this schedule on the FIFA website about the date of publication for the monthly rankings. It says the Aug rankings were released on Wed 16 Aug, but you can see at the top it says "Deadline: Thursday prior to the release date".

It is not entirely clear, but I think this means that the ranking includes all matches and data up to Thu 10 Aug. (I don't know why they need 6 days to make the calculation, maybe they don't have such good technical knowledge as Bert!) So I think that the Aug 2006 rankings do not include the results of the Euro2008 quals and friendlies played on Wed 16 Aug.

So the next question is when making the calculation of the points from the match played on Wed 16 Aug, which month's ranking do you use for working out the strength coefficient of your opponent? I guess that all matches played after the cut-off date of Thu 10 Aug should probably use the new Aug ranking number for the opposition strength. (This may be wrong but it is my best guess.)

Next is how to apply this is previous years. I have not been able to find the schedule of dates for release of monthly rankings in the last 4 years so I have had to make a sensible guess assuming that the dates are very similar to the 2006 year release dates. Some of the 2006 dates make sense, such as release dates on 13 Sep, 18 Oct, 22 Nov are obviously meant to include the results from the main international match dates up to 6 Sep, 11 Oct and 15 Nov. But for example England had a match played on 12 Feb 2003 and I don't know if this should use the Jan 2003 ranking or Feb 2003 ranking for the opponent's strength.

Another problem is where to make the cut-off when looking at the result in the last 1 year, 2 years, 3 years and 4 years. Do you follow the the date exactly e.g. 8 Sep 2005 to 7 Sep 2006, 8 Sep 2004 to 7 Sep 2005, 8 Sep 2003 to 7 Sep 2004 and 8 Sep 2002 to 7 Sep 2003? But because the international match date can change by a few days each year this can cause problems. Such as the Sep international match dates in 2003 were 6 Sep 2003 and 10 Sep 2003. Surely common sense suggests these matches are being "replaced" by the current matches and should be moved back 1 year in importance, even though the actual date may be a few days different from the specific cut-off date.

In summary I think we need to understand how FIFA apply the calculation to the above date cut-off problems before we can make a completely accurate re-calculation. I'm sure this explains why I have a difference of 9 points but there are too many factors to try them out and guess which can correct the difference.

Author: ferdi
Date: 05-09-2006, 12:00

panda wrote:

{i>but can we say that this ranking system had any bad consequences (I think we thought the WC groups were not unfair)? For a few countries? Specific examples?{/i>

At the time the draw was made for the groups in the 2006 world cup final tournament, the coefficient of Portugal was suffering from the lack of qualification matches for the Euro 2004. It is hard to say how many points Portugal lost exactly, since we do not know the results of the hypothetical qualification matches. So we do not know whether Portugal could have been seeded instead of Italy. However, Portugal were lucky to get Mexico as seeded team in their group, so they did not really suffer.

At the same time the coefficient of Germany was maximally reduced because of the lack of qualifiers for WC 2006, but since they were seeded anyway as hosts, they also didn't suffer.

The next possible team to suffer will be Switzerland. They will drop dramatically in the next 15 months in ranking, because they play no qualification matches for Euro 2008. This will still have an effect on their ranking when the draw for the WC final tournament 2010 will be made.

Author: MartinW
Date: 14-09-2006, 03:23

Have you seen the newly released Sep 2006 version of the FIFA rankings? You can see the full details here.

For my own calculations of England's ranking co-efficient the difference with FIFA's is as follows:

Aug 06:  My rank = 1,425   FIFA rank = 1,434   Diff = 9
Sep 06:  My rank = 1,468   FIFA rank = 1,477   Diff = 9

So it seems there is some consistency as both my calculations and FIFA's calculations give England an increase of 43 points during Sep 06. I think this means that whatever the error between our two methods, it cannot be included in the results of matches from Sep 05, Sep 04, Sep 03 or Sep 02 as the points from these matches have been shifted back a year and had weighting changes.

The error must be in a match from another month which counts equally towards Aug 06 and Sep 06 rankings. It won't be possible to work out where the error is until we get to a month where the change in the FIFA ranking is different from the change in my ranking calculation.

It's also noticeable that the points changes and ranking changes are a lot larger from month to month using the new method. Now it is very essential to use the correct month's ranking when working out the opposition's ranking co-efficient as there are many countries moving more than 10 places in the rankings in one month.

Re: FIFA Ranking
Author: Ricardo Date: 14-09-2006, 15:55	Well, This does not work for me, rather unfortunately: My Aug: 1309 Fifa Aug: 1322 My Sep: 1294 Fifa Sep: 1327 I must be donig something very wrong here.

Author: MartinW
Date: 15-09-2006, 03:40

Ricardo - have you seen my post above (5 Sep) about the issues that I think are affecting the calculation?

I believe we need to be very careful about the cut-off date between each month. We must know up to what date is included in each month's calculation so that the opposition ranking strength is taking from the previous month's ranking list. I think it is very easy to have some differences because of this, especially from more than one year ago when we do not know what dates the monthly calculation was released.

Author: Osiris
Date: 16-09-2006, 06:36
Edited by: Osiris
at: 16-09-2006, 06:45

Ricardo,
making a calculation jf points won in a match with Luxembourg you made a mistake. 200-195=5 but FIFA decided to give coefficient 50 for all teams at the places 150-200. So the win over Luxembourg means 50x3x2.5=375 not 37,5
Rgds.

P.S. Sorry, I didn't read you post carefully, you already pointed out to the minimum coefficient 50.

MartinW,

agree with you all differences in calculations caused by the date of match in comparison to the date of calculation of current month rating.
In my opinion it would be better if in ALL September matches the rank of opponents were taken from August FIFA Rank

Author: Martijn
Date: 23-09-2006, 16:06

I have tried to calculate the Dutch coefficient as well.

When I included the match against Belarus played on the 7th of september 2002, it summed up to 1327...

I think the only parameter that causes (y)our differences is the strength of the opponent. I tried different methods, when I simply used last month's ranking for the entire next month, I got the best result.
I do think this is just pure luck and that my ranking will be wrong next month but we'll see.

Perhaps I should mail FIFA to see if they still have the deadlines from previous years.

Re: FIFA Ranking
Author: Ricardo Date: 25-09-2006, 12:51 Edited by: Ricardo at: 25-09-2006, 13:00	Martijn, What are your seperation-dates of the years? 1 September of each year? So The 4 included years would be 1/9/02-1/9/03-1/9/04-1/9/05-1/9/06? Or are they between the first and second match in September(5th?) By 1/9 of every year aI get: 1/9/02 - 1/9/03: 122.4367 1/9/03 - 1/9/04: 123.8113 1/9/04 - 1/9/05: 370.2727 1/9/05 - 1/9/06: 677.5518 Total: 1294.073

Author: Tobcoach
Date: 23-10-2006, 08:30

I've recently entered al games of the past 4 years (almost 3600!) into an Excel-sheet, but somewhere along the line things go wrong.

To my knowledge, I have all the separation dates of the months correctly, but still France has more points than it should have, Argentina loses almost 100 points and so on. The only few countries that come up with the exact amount of points are the ones in the lower regions of the ranking list. The errors make France the number 1 instead of Brazil, and the Dutch are down to 9th instead of 7th. Further down the list the difference between my positions and the real postitions are even greater.

I had to make a few assumptions in my calculations:

1) the continent-coefficient before the WC 2006. After the WC it is easy: Europe 1,00 , S-America 0,98 and the others 0,85. Before the WC I've calculated all remains the same, except for S-America: 0,99 instead of 0,98.
2) Although I think I have the separation dates correctly, it's still an educated guess (extracted from the times the ranking has been published, most of the times the deadline is Thursday prior to publication, except when a continental final is played, in that case the deadline is set on the day of the final).
3) What to do with games between the deadline and publication of the new list? I've assumed that those games have to be calculated with the new ranking.
4) How to round the numbers? If European or S-American teams play against eachother, rounding the numers is not a problem. But what if an African team plays a European team? The contintal factor is 0,925 which means that the eventual points do not automatically round up to whole numbers. Same thing for the eventual year-totals. For example: if a country has 56,6 point for the first year, 122,7 for the second year, 211,5 for the 3rd year and 518,5 for the last year, what is the overal total of points? Is that al 4 added and then rounded up to the nearest whole numers, or do you have to round up first, and then add up. With the first case, you come to a total of 909,3 = 909 points, with the second method it's 911 points.

So, I have a total list, but it's far from perfect (yet). I don't know if I will ever be able to match the FIFA-list totally.

If anyone could help me with my assumptions (in a way that they aren't assumptions anymore), please be my guest.

Author: Thomaz
Date: 23-10-2006, 13:36

I ve had the same problem.
However, I think that there is an error in the explanation FIFA gave to calculate ranking points.
I think the the confederation strength is not the average of the two teams taking part, but is just strength of the opponent ie:
England v Argentina
England: in their rankings point calculation, conderation strength will be 0.98.
Argentina: in their rankings point calculation, conderation strength will be 1.00.
NOT 0.99 for both. Another example:
England v Australia
England: in their rankings point calculation, conderation strength will be 0.85.
Argentina: in their rankings point calculation, conderation strength will be 1.00.
NOT 0.925 for both.
By doing this, your calculated rankings for a country should be within 1 or 2 points (sometimes exact). Then the error will be due to cut off dates and exact ranking of the teams at a given time.
Do you all agree with this?

Author: Tobcoach
Date: 23-10-2006, 14:13

The FIFA is pretty explicit though about the confederation strength. The PDF-file explaining the new ranking system contains a full page which explicitly goes through all the possible results, inluding the way how the confederation strenght is calculated.

Could this be the way FIFA tries to discourage people trying to duplicatie the ranking system themselves?
Or is it just a minor oversight by FIFA?

Anyway, it's a useful tip Thomaz and I will try this approach on my own Excel-sheet. I'll let you know what the outcome is.

Re: FIFA Ranking
Author: Ricardo Date: 23-10-2006, 19:14	Thanks for the tip. I will try to do that too (for the Dutch coefficient) as soon as I got my USBkey back (soonest Thursday)

Author: MartinW
Date: 24-10-2006, 03:58

The suggestion to use the actual regional co-efficient of the opponent instead of the average of the two teams does not help in my calculations. I have the England ranking on my spreadsheet and using the average regional co-efficient my number is 3 less than Fifa for Oct rank (9 less than Fifa for Sep rank).

Using the opponent regional co-efficient means the points for each game is reduced and so if I use this my number is 22 less than Fifa for Oct rank (28 less than Fifa for Sep rank).

My assumptions are as follows: The Oct ranking was published last week Wed 18 Oct. It says the cut-off is the previous Thu 12 Oct and so this ranking obviously includes all the matches played up to then. If, for example, there would be an international match played on Sat 14 Oct then I would assume the opponents Oct rank is used and the match counts for the first time towards the Nov published ranking.

Anyway even though I have a small difference of 3 for Oct ranking I don't think it is due to rounding error because I had a difference of 9 in the previous month which is too large for rounding error.

Author: Tobcoach
Date: 24-10-2006, 09:39

With all respect Thomaz, your tip isn't working for me. In my database, the European and S-American teams lose even more points than they already did.

I've discovered something else. The FIFA is not consistant on their information about the matches taken into account for the ranking. Some examples:

"Of the 125 international matches played last month, 96 were confederation championship qualifiers: UEFA (44 matches), CONCACAF (19), AFC (11) and CAF (22). Twenty-nine friendlies were also contested."

This is a piece of text taken from the publication of the ranking of October 2006. When I search the FIFA database, I come up with 28 friendlies instead of the 29 mentioned by the FIFA.

And let's look at the year 2003. FIFA claims 870 matches were played: 72 WorldCup qualifiers, 16 Confederation Cup games, 19 Continental final games, 350 Continental qualifiers and 413 friendlies.

Searching the FIFA database, I count 882 matches. 3 more Continental qualifiers and 9 more friendlies.

Also the FIFA-database includes matches with nations that aren't recognized by the FIFA, i.e. Martinique, Guadeloupe, Timor Leste, etc. Matches against those nations are not recognized by the FIFA as international "A" matches and therefor are not taken intom account for the ranking.

Although I suspect the irregularities only affect some 'small' countries, it becomes hard to precisely copy the ranking list. Because: which matches did the FIFA include, and (more importantly) which did the FIFA not include?

The FIFA match-database is the database I used for my calculations. Now, it turns out there are matches in that database which are not included into the ranking procedure, and maybe matches that are included don't appea in the FIFA-database. Do I perhaps need to look at another database (such as the RSSSF, or world-results.net)?

Author: Tobcoach
Date: 24-10-2006, 21:44

Now it's getting frustrating...

I've tried to come up with the points for Argentina, and I've tried it in all possible ways. Every time I fall about 80 points short of the real points.

This are the 56 matches of Argentina in the FIFA-datbase:

20-11-02 Friendly Japan - Argentina 0 - 2
31-01-03 Friendly Honduras - Argentina 1 - 3
04-02-03 Friendly Mexico - Argentina 0 - 1
08-02-03 Friendly USA - Argentina 0 - 1
12-02-03 Friendly Netherlands - Argentina 1 - 0
30-04-03 Friendly Libya - Argentina 1 - 3
08-06-03 Friendly Japan - Argentina 1 - 4
11-06-03 Friendly Korea Republic - Argentina 0 - 1
16-07-03 Friendly Argentina - Uruguay 2 - 2
20-08-03 Friendly Argentina - Uruguay 3 - 2
06-09-03 FIFA WC Prel. Comp. Argentina - Chile 2 - 2
09-09-03 FIFA WC Prel. Comp. Venezuela - Argentina 0 - 3
15-11-03 FIFA WC Prel. Comp. Argentina - Bolivia 3 - 0
19-11-03 FIFA WC Prel. Comp. Colombia - Argentina 1 - 1
30-03-04 FIFA WC Prel. Comp. Argentina - Ecuador 1 - 0
28-04-04 Friendly Morocco - Argentina 0 - 1
02-06-04 FIFA WC Prel. Comp. Brazil - Argentina 3 - 1
06-06-04 FIFA WC Prel. Comp. Argentina - Paraguay 0 - 0
27-06-04 Friendly Argentina - Colombia 0 - 2
30-06-04 Friendly Argentina - Peru 2 - 1
07-07-04 Copa America Argentina - Ecuador 6 - 1
10-07-04 Copa America Argentina - Mexico 0 - 1
13-07-04 Copa America Argentina - Uruguay 4 - 2
17-07-04 Copa America Peru - Argentina 0 - 1
20-07-04 Copa America Argentina - Colombia 3 - 0
25-07-04 Copa America Argentina - Brazil 2 - 2 (PSO 2-4)
18-08-04 Friendly Japan - Argentina 1 - 2
04-09-04 FIFA WC Prel. Comp. Peru - Argentina 1 - 3
09-10-04 FIFA WC Prel. Comp. Argentina - Uruguay 4 - 2
13-10-04 FIFA WC Prel. Comp. Chile - Argentina 0 - 0
17-11-04 FIFA WC Prel. Comp. Argentina - Venezuela 3 - 2
09-02-05 Friendly Germany - Argentina 2 - 2
09-03-05 Friendly Argentina - Mexico 1 - 1
26-03-05 FIFA WC Prel. Comp. Bolivia - Argentina 1 - 2
30-03-05 FIFA WC Prel. Comp. Argentina - Colombia 1 - 0
04-06-05 FIFA WC Prel. Comp. Ecuador - Argentina 2 - 0
08-06-05 FIFA WC Prel. Comp. Argentina - Brazil 3 - 1
15-06-05 FIFA Confederations Cup Argentina - Tunisia 2 - 1
18-06-05 FIFA Confederations Cup Australia - Argentina 2 - 4
21-06-05 FIFA Confederations Cup Argentina - Germany 2 - 2
26-06-05 FIFA Confederations Cup Mexico - Argentina 1 - 1 (PSO 5-6)
29-06-05 FIFA Confederations Cup Brazil - Argentina 4 - 1
17-08-05 Friendly Hungary - Argentina 1 - 2
03-09-05 FIFA WC Prel. Comp. Paraguay - Argentina 1 - 0
09-10-05 FIFA WC Prel. Comp. Argentina - Peru 2 - 0
12-10-05 FIFA WC Prel. Comp. Uruguay - Argentina 1 - 0
12-11-05 Friendly Argentina - England 2 - 3
16-11-05 Friendly Qatar - Argentina 0 - 3
01-03-06 Friendly Croatia - Argentina 3 - 2
30-05-06 Friendly Argentina - Angola 2 - 0
10-06-06 FIFA World Cup Argentina - C�te d'Ivoire 2 - 1
16-06-06 FIFA World Cup Argentina - Serbia 6 - 0
21-06-06 FIFA World Cup Netherlands - Argentina 0 - 0
24-06-06 FIFA World Cup Argentina - Mexico 2 - 1
30-06-06 FIFA World Cup Germany - Argentina 1 - 1 (PSO 4-2)
03-09-06 Friendly Brazil - Argentina 3 - 0
11-10-06 Friendly Spain - Argentina 2 - 1

Anyone fancy a gamble? Be my guest.

Author: Tobcoach
Date: 31-10-2006, 08:57
Edited by: Tobcoach
at: 31-10-2006, 09:01

It's thinkable that after the (mostly friendly) games of November, and in particular November 15th, Italy will be the new number one.

Even with all the problems as noticed a few posts earlier, Italy is a strong contender for the number one spot.

In my calculations, Italy will gain top notch if they don't lose to Turkey and Brazil doesn't win against Switzerland.
I'm pretty sure my ranking contains a few errors, so to be on the safe side my conclusion is: If Italy wins and Brazil loses, the Azzurri will be the new number one. By the way, it will not be the first time Italy would be best ranked. In the early days of the FIFA-ranking, Italy already was the number one for one month (November '93).

Furthermore I suspect the Czech will start a slow, but steady descent from the top positions, being that their strong performances of 2003 and 2004 will have less effect over the next year. I'm pretty much convinced that they will lose a spot to Portugal (that country will progress to the rankings over the next year, providing the results are good). If things go very wrong for the Czechs, they can even lose spots to Spain and Nigeria.

And Spain will most likely lose a spot to Nigeria, even though the Africans will not play any considerable match until March 2007 (source: FIFA.com, allthough on November 15, Cameroon-Nigeria pops up on both countries' subsite, the match isn't included in the fixtures-database of the FIFA though).

Why does Spain lose a spot, even if they win against Romania? Because a friendly game does not generate enough points for Spain to obtain their current year-average. With a win they will get 531 points, the average of Spain will drop somewhere around 10 points. Combined with the "points win" of Nigeria an a slight "loss" of points for Spain due to the re-calculation of points from the last four years, gives Nigeria a slight advantadge against Spain.

The interesting game between Holland and England will not cause any big movements. If Holland wins they will get close to England, and the English will probably lose a spot to Germany. If England wins, the Germans will get close, and Holland loses a spot to Portugal. Germany will benefit from a draw, overtaking England, and Portugal will get very close to Holland.

So, the main change could be the top spot, and the descent of Czech Republic, as well as a possible re-entry into the top 10 of Nigeria (last time: september '95, since then they went down as far as 82nd in November '99).

Biggest losers overall in November could well be Trinidad&Tobago, who will have to fear for their top-100 ranking.

So much for seeing into the future...