Talk:Strongest football nations by Elo Ratings

From Wikipedia, the free encyclopedia

Contents 1 1970s 2 Averaging, original research 3 Calculating these averages 4 millie and mackenzi caradice reporting about there day!

[edit] 1970s

Why are bot the Strongest football nations by Elo Ratings and the Top 10 in this article caltulated since 1970?? There seams to be no reason to ignore previous years. I'm not sure how far back the available relevant information goes, but I would think it goes way back since there's a Top 20 per decade since 1950. There are coutries that are very much relegated because of this ommisions, most notably Uruguay.

Here are the 1872-2008 statistics.

--Mariano(t/c) 13:17, 12 April 2007 (UTC)

There is an explanation on the "Strongest nations" site, but it may have to be elaborated and certainly can use a discussion. I think that a cut-off is unavoidable, though the precise date is somewhat arbitrary.

Elo ratings can be calculated from the very first international game in 1872 (an exciting 0-0 draw between Scotland and England), but the first international games outside the British isles didn't take place until the 20th century (ignoring two US-Canada games in 1885-6) and the Elo ratings over that time mostly reflect the strength of the English team compared to the Scottish team. Brazil's first game was in 1914, Spain's in 1920, Mexico's in 1923, Colombia's in 1938. As mentioned in the "Strongest nation" intro, many nations didn't partipate until the 1950s, e.g. Ivory Coast and Cameroon's first games were in 1960.

Also, in earlier years nations would sometimes play very irregularly. A country like Canada may have played it's first game in 1885, but didn't play at all between 1927 and 1957. Furthermore, countries initially tended to play "local" teams only (Australia played New Zealand in 1922, but didn't play a European or South American team until 1969), which I would expect to cause the Elo score to reflect a local rather than world-wide relative strength.

You'll probably agree that it does not make sense to compare the average Elo rating of England's team since 1872 with that of Cameroon's team since 1960. We also have to consider that ratings only "tend to converge on a team's true strength relative to its competitors after about 30 matches". If that's true, the first meaningful scores available for Algeria and Iran is for 1966, for Iraq, Ivory Coast and Senegal 1969, for Congo and Guinea 1970, for Cameroon, Kuwait and Zambia 1971, and for Saudi Arabia 1975.

So, I picked 1970 because it's a round number and because since that time intercontinental play has been common and almost all non-minnow football nations have had a meaningful Elo score (short-lived and young countries necessarily dropping out).

Perhaps we can add some earlier decades, though a top 20 before the 1920s would be rather meaningless and most European nations didn't play between 1940 and 1945.

Afasmit 17:45, 12 April 2007 (UTC)

You are talking about a bunch of nations not having regular international matches until the 1970s, but non appear not even in the first 20. Having played less matches is not necesarilly a disadvantage, specially if they already have crossed the at least 30 matches line. --Mariano(t/c) 20:13, 12 April 2007 (UTC)

It's not a matter of advantage or disadvantage for the teams that have played longer or shorter (it can go both ways I suppose), but comparing averages over different periods is just not right (as in apples and oranges).

The nations I mentioned are indeed not in the top 20, but most are in the top 50. It's possible to make a fair, smaller (top 20 or so) list of averages going further back in time, but I feel that the longer list in combination with a bunch of top 20s per decade has a wider appeal and is more informative. --Afasmit 01:35, 13 April 2007 (UTC)

But the system of the Elo rating itself avoid favouring a team with more years in bussines. That's the hole point! Otherwise, you couldn't use it to compare a chess master with 50 years of activity with one that has just started. --Mariano(t/c) 13:04, 13 April 2007 (UTC)

As I understand it, the Elo score reflects the relative strength of a “player” in a pool based on the most recent results. Equally strong players over the last 30 games or so would have the same score, even if one started out with, say, 200 points more than the other. So, a 50 year old veteran and a relative novice chess player are not compared by their career accomplishments but by “what they’ve done lately”.

Current and past chess (FIDE) Elo scores are often tried to be compared (see Comparing top chess players throughout history), but (from that article) “Elo himself was of the opinion that it was futile to attempt to use ratings to compare players from different eras; in his view, they could only possibly measure the strength of a player as compared to his or her contemporaries”. -- Afasmit 19:24, 16 April 2007 (UTC)

I think it makes sense to include back to 1970's, as much as I'd rather see a wider range. Maybe it could be possible to include a table lower on the page with 50's til today added? If it's too much work to update it regularly, maybe that one could be updated less often, yearly or something. 90.228.255.194 08:59, 13 July 2007 (UTC)

[edit] Averaging, original research

Isn't this whole page original research anyway? We could make loads of just as valid 'stats' pages based on performing various arbitrary calculations on the ELO table. What's more, this page is time-consuming to maintain, and i'm not even sure it makes that much sense to average ELO ratings. --Iae 11:01, 14 April 2007 (UTC)

Of the offenses listed on the original research policy page, this page could perhaps represent "an analysis or synthesis of established facts, ideas, opinions, or arguments in a way that builds a particular case favored by the editor, without attributing that analysis or synthesis to a reputable source". Not a Brazilian (citizen or fan), I don't have a case to make here. That's not very convincing (I could be lying), though Brazil probably would come out on top no matter how you calculate it. More importantly, the "analysis" is a simple and simply verifiable average over time of linked-to data at a reputable source. Other lists I've built (e.g. List of highest mountains or List of most expensive paintings) involved lots of research and constituted new syntheses as well, but have not been considered original research by anyone (yet).

The time periods chosen are arbitrary, indeed, and there's probably the weakness of this page. Decades seem a reasonable choice though and for the main list the longest period up to present with accurate data seems to make sense as well. The cut-off date for this main table is defended above and in the intro, but can easily be adjusted given convincing arguments. There is even more “arbitration” in the choice of parameters of the Elo score calculation, of course.

The motivation for the original research policy was "to prevent people with personal theories attempting to use Wikipedia to draw attention to their ideas". Perhaps it is a fault to draw attention to the idea that average Elo scores are a good way to compare competitors' strengths. Fortunately this is not an original idea. For example, Elo himself wrote a book in which he rated chess players by their highest five-year average score (kind of funny, as he admitted that comparing players from different eras was futile (see above), a problem which we don't have here). I don’t have the book, but I would be surprised if he doesn’t write in it that the average over a five year span is a better measure of a chess player’s strength than, say, the peak score reached, as he opined that individual ratings “…might well be compared with the measurement of the position of a cork bobbing up and down on the surface of agitated water with a yard stick tied to a rope and which is swaying in the wind." Similarly, Jeff Sonas at his chessmetrics website [1] calculates averages of a variation of the Elo score for different time spans (1-20 years) to compare contemporary players. His stuff may have inspired this list, though I wasn't conscious about this. Perhaps I should add these references to the page. --Afasmit 19:24, 16 April 2007 (UTC)

OK cool. I was confused because ELO ratings are kind of like averages anyway (with a higher weight being assigned to those values closer in time), but I wasn't aware of the precedent. You've convinced me though and as long as the method used to calculate the data is always kept transparent (which it is now) then this seems useful. --Iae 22:00, 16 April 2007 (UTC)

[edit] Calculating these averages

How are these averages calculated? Because the Elo ratings are not updated at regular intervals, it's not at all accurate to simply average every rating that a team has had during a decade. BassoProfundo 16:29, 30 June 2007 (UTC)

The average is calculated as (rating1 x timespan1 + rating2 x timespan2 + ...)/ (total time span) Afasmit 19:51, 30 June 2007 (UTC)

[edit] millie and mackenzi caradice reporting about there day!

whe have had to look at your website to find some infomation about brazil because mackenzi(maca)needs to no because we love you! —Preceding unsigned comment added by 90.194.162.111 (talk) 18:49, 29 May 2008 (UTC)