# International Rankings and A-Z teams (Alphabet Soup Method)

Joined: 4:06 PM - Aug 04, 2012
I had an idea, and I would like to see if any of you know of something like this already in existence and what you think of the prospect.

So as we all know with international sports the biggest problem with ranking is lack of matches, and lack of data is the worse situation for mathematical analysis. To further complicate this countries often send B teams or worse. If these B or below teams are included in rankings it can make the ratings very unrealistic, if they are not it often leaves big gaps in the ratings. What if we were to add some sort of handicap for these teams, for instance adding to their final score based on a simple equation with different amounts depending on the sport... and if so, how much handicap should be granted?

The next thing to consider is how many classes of teams could we have and still be practical about it. I would think that at least A, B, and C teams are needed for classification. But there are probably other cases where D or below may even be beneficial, but I think at least adding these 3 would be a great improvement.

Since this is predominantly a football forum you all know better than I of examples in that sport where varying classifications for teams in rankings may help. Here are a few outside of football that I can think of off the top of my head, mostly from an Americas perspective.

ICE HOCKEY
Canada, Sweden, Finland, and United States often have B or C teams in competitions for various reasons.

COLD WAR
Due to the ridiculous amateur rules in the past this same concept could be applied to get much more accurate historic rankings in instances where professionals were barred from playing.

BASKETBALL
United States often sends what may be considered G or H teams (such as in the Pan American games). Even other countries have regularly sent lesser teams to competitions such as Argentina and Brazil in the South American Championship. With the new FIBA qualification format European teams are now often forced to do this, unlike the previous 2 basketball examples which are voluntary.

AUSTRALIAN RULES FOOTBALL AND AMERICAN FOOTBALL
USA send what may be considered Q, R, and S teams to Gridiron Football in international play. Australia doesn't even have an official national Aussie Rules team which would give them even a more extreme classification, probably X, Y, or Z, as matches with Australia are mostly just thrown together with whatever Australian is in the vicinity.

BASEBALL
Cuba in recent years has started banning players that abandon the country from playing for the Cuban national club. This greatly degrades the Cuban side, and their rankings are no longer truly representing the kind of baseball talent that Cuba produces. United States often sends B or below teams to these competitions as well.

CLASSIFICATIONS FIRST DRAFT (remember, this is all relative in comparison to country's full available potential in the particular sport. This is also a condensation of above example classifications)
a: primary fully organized team (no handicap)
b: backup team, but fully organized: Such as Canada in most Ice Hockey tournaments, secondary professionals
c: lower quality organized team: Such as USA in American Football or USA in basketball Pan American games or during Cold War... semi-professionals or high quality amateurs
d: disorganized team that is thrown together almost at random: such as Aussie Rules teams representing Australia using complete unknowns or regular people.

In the case of my rankings I could use these 4 classifications to temporarily reduce the country's economic power for the match. example: If Cuba's economic power is 100,000 and they are using a C team their economic power for the match may be 50,000 if using a 50% handicap for C teams. This would then be included in the current equation for adjusted score.

Joined: 12:28 PM - Apr 07, 2007
Maybe entering an elite country's B/amateur team as a separate entity in the rankings would be a way to avoid conflating results with the country's A team, but still get something relevant out of the results of the weaker team.

Joined: 6:46 AM - Jul 04, 2011
Alternatively, use the B/C teams as an 'off-ranking' measuring stick to link teams that haven't played eachother, like how Mark uses island team's results against passing ships.

Joined: 10:54 AM - Apr 05, 2012
I think the question is also trying to solve the issue of something like:
Canada only plays against USA "A" and Turkey only plays against USA "B". Is there a way to 'guesstimate' beforehand the strength of USA "B" relative to "A" (as a fixed percentage, for instance) in order to be able to rank Canada and Turkey relative to each other?

It will be argued that B-teams don't even always perform worse than the proper A-teams. We can also make the case that the Spanish regional teams, which are necessarily a subset of the Spanish player pool (with the possible exception of Catalonia, which at times has been a selection of players based in rather than from the region), can perform better (results-wise) than the Spanish team proper.

Joined: 4:06 PM - Aug 04, 2012
Kaizeler wrote: I think the question is also trying to solve the issue of something like:
Canada only plays against USA "A" and Turkey only plays against USA "B". Is there a way to 'guesstimate' beforehand the strength of USA "B" relative to "A" (as a fixed percentage, for instance) in order to be able to rank Canada and Turkey relative to each other?

It will be argued that B-teams don't even always perform worse than the proper A-teams. We can also make the case that the Spanish regional teams, which are necessarily a subset of the Spanish player pool (with the possible exception of Catalonia, which at times has been a selection of players based in rather than from the region), can perform better (results-wise) than the Spanish team proper.
Exactly, that is the purpose I had in mind for the alphabet method.

For the second paragraph I don't think I'd agree for the most part. Yes, there are instances a B team can perform better, but this is just random fluctuation that will solve itself with more matches, or in rare cases, poor administration that will eventually be worked out. I wouldn't say that Basque or Catalonia certainly performed better than Spain at any point in time, because they really haven't played enough matches to evaluate that. In addition, didn't they usually play against B squads? A Catalonia + Basque team will always play better in the long run than a separate Catalonia or Basque side (in terms of brute strength) assuming enough matches can be played to make a proper mathematical deduction.

Also, the difference between a region like Basque or Catalonia in comparison to Spain can be quite minimal. I did some experiments with NBA statistics and was grouping players by where they grew up into virtual national sides. I would add up each sides ability and compare. For instance, there was a USA team, a Lithuania team, a Florida team, Split Dalmatia, California, Texas, Wisconsin, Illinois, and so on. I even made some city sized national teams like Oakland, New Orleans, and Milwaukee. Looking at only USA teams, obviously USA had the highest score. California, Texas, Florida and several other large states were very close to each other. And these regions were not that far behind the USA. So if you faced California and USA national basketball team against each other California could win quite often... but if you played enough matches USA would still come out ahead. It is mathematically impossible for California to be a part of the USA and still be better than the USA (assuming you cloned California players and they made up part of the USA team), because in order for that to happen California would have to be better than itself. However, if California were to gain independence and certain socioeconomic conditions of basketball changed, then California could potentially become better than the rest of the United States combined as a separate entity. Things were very similar on the next level down... cities were often only slightly behind the state that they were in. Milwaukee, for instance, was neck and neck with Wisconsin, only down by a few numbers. And off note, there were many regions within the USA that ranked very poorly.

Joined: 10:54 AM - Apr 05, 2012
abramjones wrote:A Catalonia + Basque team will always play better in the long run than a separate Catalonia or Basque side (in terms of brute strength)
Theoretically (and objectively), of course. Ditto for those teams vs Spain in general as you mention. The practical problem though is often:
abramjones wrote:assuming enough matches can be played to make a proper mathematical deduction.
Although when it comes to regions you could probably easily factor them in in your model anyway, as economic and demographic data tends to generally available.

As for the "B" selections, or cases where e.g. Canada chooses to disregard all NHL players, or something, I'd say you will need an understanding of that the (self-imposed) restriction is on a case-by-case basis to even begin to be able to estimate the impact on the talent pool.

I've quickly generated a simulation of what I mean.
Suppose I've established a new country and I'm generating new players, one by one. The skill level of these players follows a normal distribution, with an average of 1000 skill units and a standard deviation of 200 units. You may be able to calculate these indicators regarding PER to get a perspective for the NBA, for instance.

As soon as I have 11 players I can form a (football) team. But the talent pool is so limited, that literally everyone makes the team. The 12th player generated will most likely be able to make the team (there's only a 1-in-12 chance that he's worse than all the others). The same will apply for the next few players, and the skill level threshold to join the team will rise quickly, but after a short while it will tend to stabilize. At some point you will already have 11 very good players, and it will take a lot of time to generate a player able to break into the "A" team.

The evolution of this threshold is illustrated by the chart below (only the general trend is what's at stake here, and obviously dependent on the choice of the initial parameters):

Simon Kuper (in Soccernomics, I think) makes the case that a team's strength can be defined by the strength of its worse player. These two points are perhaps an interesting view for your consideration; doubling a country's population will not necessarily double their team's skill output. Law of diminishing returns and all that... I could have calculated "team strength" as an average of the top 11 players or some other metric, but for simplicity's sake let's go along with this definition.

Similarly, we can repeat this exercise considering a scenario where the top, say, 200 players are excluded to proxy the strength of a B team. Obviously such a team may only be put together after we have 211 players in total, and it will follow the same pattern of starting off relatively weak, growing rapidly and then tending to stabilize:

And then you can put both of these analyses together to estimate, given a current population size, how good a "B" team would be relatively to an "A" team following these restrictions (in the very very long run these curves will tend to overlap, because the 200 players you exclude will stop making a difference). Although a number of the points discussed above will be quite subjective to determine.

Joined: 12:28 PM - Apr 07, 2007
Some brief observations for a player-skill-based analysis:

1. Players are not interchangable... a team isn't equivalent to e.g. the 11 best players if they are all goalkeepers, or if none of them are.  Effectively you need 11 curves, one for each position (or 4 for GK/DF/MF/FW).  Luckily, the sum of normals is normal... if they're independent.
2. Players retire (or get injured).
3. Skill levels also aren't fixed - players improve by playing with/against better players

These are just minor adjustments of course.

Joined: 12:28 PM - Apr 07, 2007
in the very very long run these curves will tend to overlap, because the 200 players you exclude will stop making a difference
They will tend to overlap but the relative difference will still be the important thing.  When there are many skilful players, small absolute differences in skill are still large relative differences at the top end, because most good teams are made up of elite players who are all already outliers in their country's skill distribution.

E.g. Portugal without the top 200 players will still be a lot better than average teams and minnows, but will have declined sharply relative to other elite nations.  So while you can say that "Portugal B" and "Portugal A" tend towards equal strength in the large number limit, the point is that it is still the small numbers (the best players) which make all the difference.

In a population of 10 million, the best 11 players are not dramatically better than the next best 11 players in absolute terms (compared to Joe Public FC) but the relevant comparison is to the base level for elite professional players, not the average player.

Joined: 10:54 AM - Apr 05, 2012
Brief observation 4: I'm also assuming the national coach has perfect knowledge of all of the players' skill levels. ;)

Joined: 12:28 PM - Apr 07, 2007
Kaizeler wrote:I think the question is also trying to solve the issue of something like:
Canada only plays against USA "A" and Turkey only plays against USA "B". Is there a way to 'guesstimate' beforehand the strength of USA "B" relative to "A" (as a fixed percentage, for instance) in order to be able to rank Canada and Turkey relative to each other?
Going back to this question, the relative strength of USA "B-team" and USA "A-team" is not an international issue.  It may be that across many countries, one can find a general consistent relationship between Country "B-team" and Country "A-team", such as Kaizeler is working on, but perhaps it is also possible to make the considerations bespoke to each country by considering domestic aspects only.

Observing that club rankings (especially domestic ones) should be more accurate in the sense of more regular fixtures between the same subset of teams, maybe it is possible to try and map the national "A-team" and "B-team" onto the domestic league in order to gauge their strengths.   Note that it needn't be the domestic league of the USA.  If the players of A and B teams are largely based abroad, then perhaps using a European club ranking is a better comparison.   One could take the 11th lowest ranked club of a player in the USA A-team, and the same of the B-team, and use the difference as the relative strength A vs B.

Or some sort of averages of the rankings of the clubs of the players in the A-team and B-team.

Joined: 12:28 PM - Apr 07, 2007
One could consider the results of U21/U23 international teams, and a ranking of them, to assess the impact of the removal of a country's elite players.  There are various results over the years of a country's A team against its own U21 team, which could be incorporated.

Joined: 6:46 AM - Jul 04, 2011
nfm24 wrote:One could take the 11th lowest ranked club of a player in the USA A-team, and the same of the B-team, and use the difference as the relative strength A vs B.

Or some sort of averages of the rankings of the clubs of the players in the A-team and B-team
I'm not too sure about that part, overall maybe it works out but in theory there's too many variables, like what if that 11th-ranked club is being carried by an exceptional player, or the B national team has a bit-part player from a higher-ranked club?

And how do you handle players that are out on loan, do they get ranked by their Chelsea or their Vitesse?

Joined: 12:28 PM - Apr 07, 2007
Probably an average over 11 players is better to smooth out this sort of issue.