Ranking Procedure

In this section, I attempt to explain the general concepts behind how the International, League and Club rankings are calculated.

Eligible Teams and Leagues

For the International and League rankings this is based on three main criteria:

• Is the team recognised by either FIFA or their local confederation?
• If not, are they an internationally recognised territory (i.e. not politically sensitive)?
• Have they played sufficient number of games in their history to enable the ranking calculation to yield a reasonably sensible result?

With these rules in mind, I settled on a list of 224 countries. Those not recognised by FIFA are listed below for clarity.

• Bonaire (full member of CONCACAF)
• French Guiana (full member of CONCACAF)
• Greenland (unaffiliated to any specific confederation, but has been active within last 8 years)
• Guadeloupe (full member of CONCACAF)
• Kiribati (associate member of OFC)
• Martinique (full member of CONCACAF)
• Mayotte (unaffiliated to any specific confederation, but has been active within last 8 years)
• Northern Mariana Islands (associate member of AFC)
• Reunion (associate member of CAF)
• St. Maarten (NL) (full member of CONCACAF)
• St. Martin (FR) (full member of CONCACAF)
• Tuvalu (associate member of OFC)
• Zanzibar (associate member of CAF)

The list of eligible countries will be reviewed annually.

For the Club Rankings, the criteria used are slightly more stringent:

• Is the team an active member of the top-level division (i.e. that which provides entry into international club competitions) from any of the countries within the League Ranking?
• Are the results of that league actively available to enable regular updates to take place? To this end I have selected a list of 160 leagues where I'm comfortable results can be obtained easily. These are listed below.
• Has the club itself played a minimum of 20 eligible matches within the last 2 years (rolling period)? An eligible match is effectively one from either the top-level division (as defined above) or from a recognised international club competition (including qualifying rounds). Friendlies, Promotion/Relegation playoffs (involving teams from lower tiers) and all domestic cup games do not count for the purposes of this ranking.

The 160 eligible leagues can be split into 4 zones in the following fixed ratio - Africa (36), Asia (36), Americas (36), and Europe (52). The number of leagues included (and the ratio split between zones) might change over time subject to my preference.

At the outset of each calendar year, the latest League Rankings are reviewed, with adjustments made to the existing list. The leagues are selected in ranking order, but are only accepted where it is possible to obtain results easily online. For this purpose I use a single web resource, so where leagues are added to/removed from this website, I will update the list accordingly.

Africa

• Algeria
• Angola
• Benin
• Botswana
• Burkina Faso
• Burundi
• Cameroon
• Congo
• DR Congo
• Egypt
• Ethiopia
• Gabon
• Gambia
• Ghana
• Guinea
• Ivory Coast
• Kenya
• Lesotho
• Libya
• Madagascar
• Mali
• Mauritius
• Morocco
• Mozambique
• Nigeria
• Rwanda
• Senegal
• South Africa
• Sudan
• Swaziland
• Tanzania
• Togo
• Tunisia
• Uganda
• Zambia
• Zimbabwe

Americas

• Antigua & Barbuda
• Argentina
• Aruba
• Barbados
• Belize
• Bermuda
• Bolivia
• Brazil
• Cayman Islands
• Chile
• Colombia
• Costa Rica
• Cuba
• Curacao
• Dominican Republic
• Ecuador
• El Salvador
• French Guiana
• Guadeloupe
• Guatemala
• Guyana
• Haiti
• Honduras
• Jamaica
• Martinique
• Mexico
• Nicaragua
• Panama
• Paraguay
• Peru
• St. Kitts & Nevis
• Suriname
• Trinidad & Tobago
• United States
• Uruguay
• Venezuela

Asia

• Australia
• Bahrain
• Bangladesh
• China
• Fiji
• Hong Kong
• India
• Indonesia
• Iran
• Iraq
• Japan
• Jordan
• Kuwait
• Kyrgyzstan
• Lebanon
• Malaysia
• Maldives
• Myanmar
• New Caledonia
• New Zealand
• Oman
• Palestine
• Philippines
• Qatar
• Saudi Arabia
• Singapore
• Solomon Islands
• South Korea
• Syria
• Tahiti
• Tajikistan
• Thailand
• Turkmenistan
• United Arab Emirates
• Uzbekistan
• Vietnam

Europe

• Albania
• Andorra
• Armenia
• Austria
• Azerbaijan
• Belarus
• Belgium
• Bosnia & Herzegovina
• Bulgaria
• Croatia
• Cyprus
• Czech Republic
• Denmark
• England
• Estonia
• Faroe Islands
• Finland
• France
• Georgia
• Germany
• Gibraltar
• Greece
• Hungary
• Iceland
• Israel
• Italy
• Kazakhstan
• Latvia
• Lithuania
• Luxembourg
• Macedonia
• Malta
• Moldova
• Montenegro
• Netherlands
• Northern Ireland
• Norway
• Poland
• Portugal
• Republic of Ireland
• Romania
• Russia
• Scotland
• Serbia
• Slovakia
• Slovenia
• Spain
• Sweden
• Switzerland
• Turkey
• Ukraine
• Wales

Eligible Matches

For the International Rankings, the results of all competitive matches played within the last 8 years (rolling period) count. By competitive, I mean qualifiers and tournament matches from each of the following competitions only.

• World Cup
• Confederations Cup
• African Nations Cup
• Asian Cup
• European Championships
• Gold Cup
• Oceanian Nations Cup
• Copa America

Where a country fails to play a minimum of 20 matches, the lookback period (and therefore list of results which count for that team) is further extended covering the same (or equivalent) tournaments until this requirement is met. All earlier results are excluded.

Where a country fails to meet the minimum requirement of 20 competitive matches within its entire history, friendly matches (only those against other teams within the ranking) are then taken into consideration (most recent first) until the criteria is met, although in the case of Greenland (16), Bonaire (17) and Kiribati (11), these territories simply haven't played 20 eligible games (competitive or non competitive) ever.

The countries which haven't met the minimum requirement of 20 competitive matches are listed below for clarity (numbers in parentheses represent competitive matches played to date - correct as at end of September 2018):

• Bonaire (6)
• Greenland (0)
• Kiribati (4)
• Kosovo (12)
• Mayotte (0)
• Northern Mariana Islands (3)
• Reunion (0)
• South Sudan (14)
• Turks & Caicos Islands (19)
• Tuvalu (9)
• Zanzibar (0)

For the League Rankings, the results of all competitive matches involving teams acting as representatives of a given league structure played within the last 8 years (rolling period) count. By competitive, I mean qualifiers and tournament matches from each of the following competitions only.

• FIFA Club World Championship
• CAF Champions League
• CAF Confederation Cup
• CAF Super Cup
• AFC Champions League
• AFC Cup
• AFC Presidents Cup (now defunct)
• UEFA Champions League
• UEFA Europa League
• UEFA Super Cup
• CONCACAF Champions League
• CONCACAF League
• OFC Champions League
• CONMEBOL Copa Libertadores
• CONMEBOL Copa Sudamericana
• CONMEBOL Recopa Sudamericana

Where the representatives of a particular league structure* fail to play a minimum of 20 matches within the 8 year period, the lookback period (and therefore list of results which count for that league) is further extended covering the same (or equivalent) tournaments until this requirement is met. All earlier results are then excluded. Friendly matches are not used to top-up any shortfall for leagues who are still to meet the minimum games played requirement.

*Note that for the purposes of the League Ranking, the results of all Canadian teams (Toronto FC, Montreal Impact, and Vancouver Whitecaps) playing in MLS count towards the United States. No teams from the Canadian Soccer League play in CONCACAF club competition, but if they were ever to do so, then this league would be added to the rankings accordingly. Likewise the results of FC Vaduz (a Liechtensteiner club playing within the Swiss league structure) who technically represent Liechtenstein in UEFA club competitions count towards Switzerland.

The countries which haven't met the minimum requirement of 20 competitive matches are listed below for clarity (numbers in parentheses represent competitive matches played to date - correct as at end of September 2018):

• Afghanistan (2)
• Anguilla (0)
• Bahamas (3)
• Bonaire (16)
• British Virgin Islands (0)
• Canada (0) - as stated above all results of MLS-affiliated clubs count towards the US
• Cuba (15)
• Djibouti (4)
• Dominica (16)
• Greenland (0)
• Grenada (2)
• Guam (4)
• Kiribati (0)
• Kosovo (14)
• Laos (18)
• Liechtenstein (0) - as stated above all results of Liechtenstein clubs count towards Switzerland
• Mayotte (0)
• Montserrat (5)
• Northern Mariana Islands (0)
• Sao Tome & Principe (6)
• South Sudan (18)
• St. Kitts & Nevis (11)
• St. Maarten (NL) (3)
• St. Martin (FR) (2)
• Timor Leste (0)
• Turks & Caicos Islands (0)
• Tuvalu (0)
• US Virgin Islands (19)

For the Club Rankings, the results of all 160 selected domestic leagues (including playoff phases where all competing teams are current members of the top tier), as well as results from the accepted group of international club competitions (same list as for the League Rankings) played within the last 2 years (rolling period) count.

As stated earlier in this piece, league phases involving teams from lower tiers, and all domestic cup matches (including super cups and equivalents) do not count.

All teams involved in any eligible match (including those counting towards the League Rankings) are included within the ranking calculation, however only those teams who have played in a minimum of 20 matches will receive an overall points score, and hence ranking.

Calculation procedure

In my opinion, a strong ranking system is one that is both robust and uses a simple underlying methodology with minimal variables. To this end, the criteria used are limited to the following:

• Relative strengths of the two teams involved in any given match
• The location of that match
• The match result (adjusted for extra time and penalty kicks)
• The relative weighting of that match versus others played by the same team
• An adjustment (known as a Ranking Factor) used to apply penalties to particular teams/leagues for lack of activity or where teams/leagues from a particular zone are over-ranked versus expectation.

Base Match Points

Within all three rankings, each of the teams/leagues is assigned a coefficient between 0 and 10, where 10 is strongest, 0 is weakest, and 5 should be considered average. These are closely aligned to the points scores of each team prior to the application of the Ranking Factor (see below), and are used to define the relative strengths of the two teams involved in any given match. When considered in conjunction with the match location, an expected outcome for each match can be derived.

Teams/leagues then score points (known as the Base Match Points, BMP) for each individual match (including adjustments for extra time and penalty kicks where applicable) depending on whether they meet, exceed or fall short of this expectation. In practise, this means that there might be a situation where a team wins a game, but actually loses points because, for example, they might have been expected to win a match by 3 goals, but the margin of victory was only 1 goal. The same would be true for the losing team in that scenario, except in that case, that team would gain points as they have outperformed expectation.

The point scores for all eligible matches are then weighted according to how recently they were played versus all other matches for that particular team/league. This is best explained using a worked example. Suppose a team has 50 counting matches. The most recent game carries a Match Weighting (MW) of 50/50, the second most recent game carries a MW of 49/50, the third most recent carries a MW of 48/50, and so on, right down to the oldest (counting) game which will carry a MW of just 1/50. 

The Weighted Match Points (WMP) can then be computed as follows:

WMP = Sum of (BMP x MW)

                Sum of MW

Ranking Factor

The Ranking Factor (RF) is constructed as the product of two separate measures - a team/league adjustment, and a zonal adjustment. 

The team/league adjustment penalises teams/leagues who play fewer than the minimum required number of games within the rolling time period for that particular ranking (8 years for International and League, 2 years for Club). Ultimately a factor between 0.8 and 1 is derived. The rationale behind this is simple. Teams/leagues could effectively remain idle for a long period of time and hold onto their existing points and positions, which in turn could lead to a misrepresentation of their true position. In my rankings therefore, teams need to remain suitably active to score highest.

The zonal adjustment is used to ensure that certain regions (i.e. the confederations) are not over-ranked versus the others. This is the most common problem I have encountered when devising new or tweaking existing football ranking systems. Given the significant volume of matches played between teams/leagues within the same zone, it is not usually too difficult to rank teams/leagues from the same zone against each other, however the opposite is true when trying to rank teams from one zone versus another, simply because it is relatively rare for inter-zonal matches to take place.

The way in which my zonal adjustments operate is relatively simple to understand. In all cases, the relative performance of each of the 6 confederations within major competitions (see below) is converted into a benchmark coefficient score for each zone. The coefficient of the highest ranked team/league from each confederation is then compared to their zonal benchmark, and should it exceed this mark, the point scores of all teams/leagues from the entire confederation are adjusted downwards using the simple ratio of the two measures. There is no equivalent upward adjustment.

For the International Ranking, the major competitions used are the last 2 editions of the World Cup (75%) and the last 2 editions of the Confederations Cup (25%), whereas for the League and Club Rankings, the major competitions used are the last 8 editions of the Club World Championship (100%).

Therefore the final points score for any given team/league is simply WMP x RF.

Iterations

The sharp-witted amongst you will have already identified a key aspect of the rankings. That is that the calculations form an infinite loop. The coefficients are driven off the points scores of each team/league, and these in turn form a key part of the calculation of updated points scores.

For example, a team's point score increases. As a result, their coefficient also increases. This creates a series of slightly adjusted expected results, and as such a different overall points score. That new points score, then drives another coefficient, which in turn creates another different points score, and so on...

To get around the problem, I perform a fixed number of iterations each time the rankings are refreshed (fixing the coefficients in place for each individual iteration, so as to avoid circular references arising), however for various reasons (most notably to reduce historical bias) before performing the first iteration, I set the coefficients of all teams/leagues to be exactly 5 (i.e. the mid-point between 0 and 10). This means that the rankings are effectively run from scratch each time, which eliminates (amongst other items) the problem of new teams being added with no prior record. This is one of the most fundamental differences between my ranking and that of the ELO Ratings, which whilst relying on a fixed start position for each team, all future movements are dependent only on the results which have occurred previously. In my rankings, all eligible results for a given team are therefore directly linked with all others, whether these took place earlier or later.

One slight variation on the above pertains to the Club Ranking, where I utilise the latest League Rankings (which are always calculated immediately prior to the Club Rankings) as the starting coefficient for all teams from that particular league. This covers against the situation where a particular league has not entered teams into international club competitions within the last 2 years and as such the results of all teams from that particular league structure are effectively matches against one another, thus locking in an average coefficient of 5 (for the teams of that particular league). This would otherwise lead to a situation where the average performer within that league structure sits higher/lower than the League Rankings might suggest, so could ultimately result in misrepresented ranking positions.

Quirks

There is one key quirk which is worth explaining. It relates to the impact that the results of other teams (and their movements) might have on you. Such is the way that BMP is computed, it relies on knowing the relative difference between the coefficients of the two teams involved in a given match.

Let us suppose that you are Team A and you have recently defeated Team B by 2 goals. If the relative difference between the coefficients of both teams was x, you will score a specific number of points for that 2 goal victory, but then consider what happens if that coefficient difference was to narrow due to other unconnected results. Suddenly that 2 goal victory seems better than before, and indeed the points scored will reflect that (better for you, worse for them). The same is true where the coefficient difference was to widen, except this time is would be worse for you, better for them.

Extend this concept out further and you'll find that entire zones can move up and down en masse from month to month (largely due to the comparatively huge volume of intra-zonal matches versus inter-zonal matches). This can happen even where no further inter-zonal matches have taken place, as any shift of weighting (caused by older matches dropping out, new same-zonal matches being added) can affect the balance. The zonal adjustments can help to regulate this to a point, but cannot eliminate it entirely. One to keep an eye on...

Scott.