Latest updates

-

Thursday, April 1, 2021

FIFA ranking: April 2021 final preview

Here is the final preview of the April 2021 ranking. FIFA will publish this ranking next Thursday April 8th. 

After the disqualification of Chad and the subsequent forfeit of their last two group matches in AFCON qualifying, I expect, conform the inclusion of other awarded matches, their awarded matches to be included in this ranking.



Two teams record a new low:

Togo (133 - 129 in April 2018)

Aruba (205 - 202 in February 2008)


Best movers:

11 - Guinea-Bissau
10 - Namibia
9 - Armenia


Worst movers:

-9 - Mozambique
-8 - Libya
-7 - Estonia


Best movers in the top 50:

5 - Serbia
4 - Nigeria, Scotland, Egypt


Worst movers in the top 50:

-6 - Romania
-5 - Republic of Ireland
-3 - Northern Ireland

Stay safe and healthy everybody !


pos Apr.

team

pts Apr.

+/- ranking

+/- points

1

Belgium

1783,38

0

3

2

France

1757,30

0

2

3

Brazil

1742,65

0

0

4

England

1686,78

0

17

5

Portugal

1666,17

0

4

6

Spain

1648,13

0

3

7

Italy

1642,06

3

17

7

Argentina

1641,95

0

0

9

Uruguay

1639,08

-1

0

10

Denmark

1631,55

2

18

11

Mexico

1629,56

-2

-2

12

Germany

1609,12

1

-1

13

Croatia

1605,75

-2

-11

14

Switzerland

1604,58

2

12

15

Colombia

1600,66

0

0

16

Netherlands

1598,04

-2

-11

17

Wales

1570,36

1

8

17

Chile

1569,52

0

3

19

Sweden

1568,94

1

11

20

USA

1555,42

2

7

21

Poland

1549,87

-2

-9

22

Senegal

1542,45

-2

-16

23

Austria

1523,42

0

-8

24

Ukraine

1515,85

0

-5

25

Serbia

1512,90

5

21

25

Tunisia

1512,88

1

10

27

Peru

1511,88

-2

0

28

Japan

1509,34

-1

7

29

Turkey

1505,05

3

18

30

Venezuela

1500,71

-2

0

31

Iran

1499,56

-2

4

32

Nigeria

1487,16

4

13

32

Algeria

1486,69

-1

-1

34

Morocco

1479,32

-1

-2

35

Paraguay

1476,02

0

0

36

Slovakia

1475,24

-2

-3

37

Hungary

1468,75

3

9

38

Russia

1462,65

1

2

39

Korea Republic

1460,25

-1

-5

40

Czech Republic

1458,81

2

3

41

Australia

1457,49

0

0

42

Norway

1452,34

2

2

43

Romania

1449,23

-6

-17

44

Scotland

1441,43

4

5

45

Egypt

1432,66

4

1

45

Jamaica

1432,64

2

-4

47

Republic of Ireland

1427,25

-5

-29

48

Northern Ireland

1425,74

-3

-14

49

Costa Rica

1423,40

1

-4

50

Ghana

1419,05

2

-5

50

Greece

1418,70

3

6

52

Iceland

1415,99

-6

-22

53

Ecuador

1413,01

4

4

54

Finland

1412,45

1

1

55

Cameroon

1404,87

-5

-22

55

Bosnia-Herzegovina

1404,87

1

-5

57

Mali

1400,94

-3

-11

58

Cote d'Ivoire

1397,77

3

14

59

Qatar

1394,43

-1

3

60

Burkina Faso

1392,88

-2

2

61

Congo DR

1376,65

-1

-9

62

North Macedonia

1374,73

3

13

63

Slovenia

1369,88

-1

-9

64

Montenegro

1368,49

-1

-2

65

Saudi Arabia

1364,39

2

11

66

Albania

1362,09

0

2

67

Honduras

1361,21

-3

-6

68

Iraq

1352,88

1

5

69

El Salvador

1341,24

1

-3

69

Canada

1340,56

4

9

71

Bulgaria

1339,03

-3

-11

72

Guinea

1331,79

0

-2

73

United Arab Emirates

1330,16

1

4

73

Cape Verde Islands

1330,16

7

24

75

South Africa

1325,64

-4

-15

76

Curacao

1323,06

0

10

76

China PR

1322,96

-1

0

78

Panama

1321,98

0

10

79

Syria

1309,79

-3

-3

80

Oman

1301,51

1

-1

81

Bolivia

1300,11

-2

-7

82

Benin

1293,10

0

-9

83

Haiti

1290,62

1

6

84

Uzbekistan

1287,22

1

4

84

Uganda

1286,76

-1

-14

86

Israel

1283,36

1

4

87

Zambia

1280,21

3

15

88

Gabon

1278,10

-2

-2

89

Belarus

1276,79

-1

8

90

Armenia

1273,28

9

40

91

Georgia

1259,51

-2

-7

92

Vietnam

1258,06

1

0

93

Lebanon

1256,08

-1

-7

93

Congo

1255,88

-3

-9

95

Jordan

1251,58

0

9

96

Luxembourg

1248,29

2

13

97

Cyprus

1240,78

3

17

98

Kyrgyzstan

1240,08

-2

0

99

Madagascar

1238,24

-5

-19

100

Bahrain

1233,23

-3

-5

101

Mauritania

1224,76

0

18

102

Kenya

1206,81

2

20

103

Trinidad and Tobago

1201,33

0

1

104

Palestine

1196,80

-2

-9

105

India

1184,36

-1

-3

106

Thailand

1178,07

4

0

107

Zimbabwe

1175,50

5

-1

108

Guinea-Bissau

1171,19

11

25

109

Korea DPR

1169,96

6

0

110

Azerbaijan

1168,50

-2

-11

111

Namibia

1168,44

10

25

112

Niger

1165,96

1

-6

113

Faroe Islands

1165,45

-6

-18

114

Sierra Leone

1162,62

2

-2

115

Estonia

1162,45

-7

-18

115

Malawi

1162,36

8

22

115

Mozambique

1161,55

-9

-23

118

Central African Republic

1157,64

-4

-13

119

Libya

1156,25

-8

-21

120

Kosovo

1151,73

-3

-3

121

Tajikistan

1151,10

-1

7

122

New Zealand

1149,43

-4

0

122

Sudan

1149,13

5

25

124

Kazakhstan

1147,35

-2

5

125

Philippines

1135,94

-1

0

125

Angola

1135,88

0

2

127

Guatemala

1129,30

3

17

127

Antigua and Barbuda

1129,02

-1

2

129

Rwanda

1123,69

4

21

130

Turkmenistan

1106,51

2

0

131

Comoros

1106,44

-1

-6

132

Equatorial Guinea

1105,08

2

8

133

Togo

1104,18

-5

-11

134

Lithuania

1102,84

-5

-11

135

St. Kitts and Nevis

1091,12

5

17

136

Suriname

1089,43

5

16

137

Tanzania

1088,05

-2

1

138

Latvia

1081,66

-2

0

139

Myanmar

1081,26

-2

0

140

Ethiopia

1079,35

6

13

141

Chinese Taipei

1078,48

-3

0

142

Burundi

1075,64

-4

-2

143

Solomon Islands

1072,78

-2

0

144

Hong Kong

1072,00

-1

0

145

Yemen

1070,54

0

0

146

Lesotho

1069,91

-3

-2

147

Nicaragua

1060,50

2

6

148

Kuwait

1056,30

0

-1

149

Afghanistan

1052,24

1

0

149

Botswana

1051,74

-3

-14

151

Liberia

1047,45

1

0

152

Gambia

1043,45

5

9

153

Malaysia

1040,02

0

0

154

Eswatini

1039,24

-1

-1

155

Maldives

1038,31

0

0

156

Dominican Republic

1036,22

3

18

157

New Caledonia

1035,12

-1

0

157

Andorra

1034,90

-6

-13

159

Singapore

1020,27

-1

0

160

Grenada

1017,58

0

3

161

Tahiti

1014,27

0

0

162

Barbados

1010,97

0

2

163

Fiji

996,27

1

0

163

Vanuatu

995,62

1

0

165

Guyana

990,63

2

3

165

Papua New Guinea

990,55

1

0

167

St. Vincent and the Grenadines

989,32

1

3

168

Bermuda

987,78

1

5

169

South Sudan

982,61

-6

-15

170

Belize

977,95

0

4

171

Nepal

968,37

0

0

172

Mauritius

964,94

0

0

173

Indonesia

964,07

0

0

173

Cambodia

963,79

0

0

175

Malta

955,89

1

5

176

St. Lucia

953,45

-1

0

177

Moldova

948,16

0

-2

178

Puerto Rico

938,86

1

-3

179

Chad

935,01

-1

-12

180

Montserrat

933,64

3

13

181

Cuba

923,03

-1

-13

182

Macau

922,10

0

0

183

Djibouti

918,91

1

0

184

Bangladesh

916,89

2

0

185

Laos

912,07

3

0

186

Liechtenstein

911,05

-5

-13

186

Bhutan

910,96

3

0

188

Dominica

904,26

-4

-15

188

Brunei Darussalam

903,90

3

0

190

Sao Tomé e Príncipe

903,25

-3

-11

191

American Samoa

900,27

1

0

192

Mongolia

896,97

-2

-9

193

Samoa

894,26

1

0

194

Cayman Islands

884,44

-1

-13

195

Gibraltar

880,16

0

-8

196

Timor-Leste

879,43

1

0

196

Somalia

879,13

1

0

198

Guam

872,83

1

0

199

Pakistan

866,81

1

0

200

Seychelles

865,69

2

0

201

Bahamas

861,84

-5

-18

201

Tonga

861,81

2

0

203

Eritrea

855,56

2

0

204

Sri Lanka

853,07

2

0

205

Aruba

850,10

-5

-17

206

Turks and Caicos Islands

843,65

-3

-18

207

US Virgin Islands

829,24

0

-15

208

British Virgin Islands

826,27

0

-16

209

Anguilla

805,32

0

-16

209

San Marino

804,71

1

-5


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.

35 comments:

  1. Hello Ed, I have a question: why are the games Switzerland vs Finland and Sweden vs Estonia given an importance of 5? Are they considered to be outside the window?

    Thank you for all your work, I really appreciate it!

    ReplyDelete
    Replies
    1. Hi Rodrigo, you're welcome and that's a really good question.

      Officially a FIFA International match window does not include the Wednesday, but as this March window was extended by one day (for the UEFA confederation) to accommodate 3 match days of WC qualification, one could indeed argue if also friendlies played on that last day should be included in the window.

      To be honest, I don't know for certain if these friendlies should be calculated with an I-factor of 5 or 10. The extended windows in November and October last year had no friendlies on the Wednesday and we saw in the last September window (officially 'opened' for UEFA only) that an AFC-friendly (scheduled inside the window) was calculated with an I of 10. That's the 'nearest' example of how to handle here, but these are certainly different circumstances.

      We will see next week what FIFA decides here. I will of course adapt and reapply my formulas when FIFA has shown that they calculate this specific case with an I-factor of 10. It is a first anyway.

      Delete
    2. Thank you very much for your answer.

      Yes, I saw that this window finished a day later for European teams and therefore I assumed that those games should have an I-factor of 10, because if we stick to what FIFA says, they're technically inside the window

      We can only wait.

      Delete
    3. Switzerland is above croatia according new fifa ranking. So, it must be factor 10 was used

      Delete
  2. Hello Ed, I guess the San Marino FA are going to be happy as they aren't the worst nation by themselves anymore

    ReplyDelete
    Replies
    1. I doubt if the San Marinese FA concerns itself with the position of their A-team in the FIFA ranking at all :)
      But indeed, the CONCACAF minnows are finally 'catching up' with San Marino.

      If I could wish for one special match it would be between Anguilla and UEFA's weakest NT to settle matters once and for all...

      Delete
    2. We've got the mouthwatering prospect of a San Marino-Andorra clash in these World Cup qualifiers.

      Probably the most intriguing question in that group with England likely to win it and Poland likely to finish 2nd (Hungarians may disagree with me).

      Andorra probably clear favourites to finish 5th, but it'd be amazing if San Marino could finally win a game.

      Delete
  3. Well well, FIFA is on a roll :)

    First, they published the ranking on Wednesday instead of the normal Thursday. Second and more important, they finally decided to give their points totals now rounded on two decimals. This makes it much easier to decide which matches should be included. Kudos to FIFA.

    A consequence is that the ranking positions are now based on these real points (instead of the rounded integer points), so teams will from now on have their own position and it will hardly occur anymore that two or more teams share a position.

    Indeed, the friendlies on the last Wednesday of the match window are calculated with an I-factor of 10, so I will adapt my formulas.

    My assumption that the awarded Chad matches would be included, was correct.

    FIFA excluded two matches that I included:
    - 2021-03-13 Kenya - South Sudan; friendly 1:0 in Nairobi
    - 2021-03-26 Uzbekistan - Ghana; friendly 2:1 in Tashkent
    It was a Ghana B-team so that's not surprising.

    FIFA included two matches that I excluded:
    - 2021-03-25 Bahrain - Syria; friendly 3:1 in Manama
    - 2021-03-27 Nepal - Bangladesh; friendly 0:0 in Kathmandu
    Now this last match was part of a friendly tournament in Nepal and there was a final played in that tournament two days later between the same two teams. Strangely, this final was not included by FIFA !

    I will further research the small differences (maximum 0.16 point) that I find for 19 teams in comparison between my calculations and FIFA's unrounded points. I will report on these differences later.

    ReplyDelete
    Replies
    1. Hi Ed:
      I had included the EQG-LBR from 2019.07.21 and the BOT-LBR from 2019.09.30. When I exclude them, it fixes a lot of issues in CAF.

      Delete
    2. Hi Martin,

      Thanks! Those two matches are indeed also my prime suspects.

      Delete
    3. How can you find what matches are included into calculation by fifa? Any official source?

      Delete
    4. Hey mick, that's the whole problem: there is no complete, reliable, official, public source which matches are included.
      I dare to say that the sheet which you can find under Important info in the top menu, contains the nearest thing. It's not official, but all other qualifications are true :)
      Beware, the April version is not yet uploaded as there are still some small differences to resolve, but I plan to do that this weekend.

      Delete
    5. Hi Ed:

      Recall the mess FIFA did with the GUY-BRB match of 2019.09.06? It was awarded to GUY on 2019.12.06 because of BRB's ineligible players with the score of 3-0. I commented here:
      http://www.football-rankings.info/2019/07/fifa-ranking-july-2019-final-preview.html?showComment=1564079425110#c2719149305776975154

      Well, it seems we may have to adjust this game again. Explanation follows.

      Delete
    6. Sorry, I wrote the post but forgot to load it. Here it is what we should do about the BRB-GUY game:

      1) One way to update was to assume the match was never played on 2019.09.06 (inclusion=0), and instead was played on 2019.12.06, with GUY entering the match with 1006.4436 points for BRB and 942.43539. With that, I get that the exchanged points would have been 8.416549. BUT (assuming all matches up to today are correct) I end up with 1010.93507 points for BRB and 990.65872 for GUY.
      2) What I proposed in my post from 2019.07.25 was to use the November 2018 ranking (BRB: 1005.4952, GUY: 943.39295), and make an adjustment of points. With that, the exchanged points would have been 8.3895, BUT (assuming all matches up to today are correct) I end up with 1010.97657 points for BRB and 990.62666 for GUY.
      3) So it seems that FIFA did something else. We need that the points exchaged were 8.40. I think that FIFA averaged both situations. More precisely, they averaged the ex-ante ranking in both situations, and used 1005.9694 for BRB and 942.91417. With that, the exchanged points would have been 8.4030356 and, assuming all matches up to today are correct, I end up with 1010.94647 points for BRB and 990.64809 for GUY. With that, I also fix Jamaica's points.

      In any case, we would need that the points exchanged between BRB and GUY were around 8.40. I have found at least another way to get it, but it implies to assume that FIFA made a mistake.

      Delete
    7. Assuming my previous post is right, then in CONCACAF I only have to fix PAN and NCA. I believe the culprit is the PAN-NCA played on 2020.02.25.

      One way to fix this is to assume that the match was NOT included on its original date, but at a later date. Concretely, it would have been included AFTER the PAN-GUA match of 2020.03.04. It could be FIFA's mistake, or that they originally thought the match was not up to FIFA's standards, but then changed their minds.

      Anyway, with that assumption (as if the PAN-GUA was played first, and later the PAN-NCA), then the points exchanged between PAN and NCA would be 1.136472 (and not 1.130404), and with that I would fix all problems in CONCACAF.

      My guess is that something similar (matches considered at a later date) would have happened with other games in CAF (which is the only confederation in which I have discrepancies in points).

      Delete
    8. If a match was updated afterwards, my assumption is all matches played after that match are supposed to be updated as well because points after the match (Es from the excel sheets) are changed. If you change only points of teams involved in the single game by adding/substracting points, then it leads you to inconsistent data. Barbados played next game let's assume against puerto rico. Calculation of puerto rico points after the game depends on barbados points which were changed in the meanwhile. Therefore puerto rico points must be changed too etc. Published rankings cannot be changed, but all points obtained in the chain, needed for calculation for next rankings must be rebalanced.

      Delete
    9. Hi Martin and mick,

      here's my take on things in an intermediate report:
      - Martin, the correction for the GUY-BRB match could well be conform your proposal. At the time we could only compare to the integer values in the December 2018 ranking. The small correction you propose doesn't lead to other integer points for GUY and BRB or other teams, so that's okay with me, as it indeed resolves the differences for GUY, BRB and JAM.
      - your fix for the two remaining CONCACAF differences for NCA and PAN seems too farfetched for me. The matches you mention count for the same April 2020 ranking and it doesn't make any sense to me to assume a different match date (other than the actual match date) for a match when the calculation of the subsequent ranking, where both matches are included, still has to be performed. I don't believe in such artificial explanations, nor in 'FIFA's mistake' as an explanation.
      - the remaining CAF differences (ZAM, BOT and ETH 0.06, NIG 0.05, SDN and GHA 0.01) after exclusion of matches BOT-LBR on 30 Sep 2019 and EQG-LBR on 21 Jul 2019) are difficult to resolve. FIFA reports constraints for each month's ranking in terms of the number of matches included in each ranking. The EQG-LBR exclusion already violates the FIFA-constraint for the July 2019 ranking of 146 included matches, so another one has to be included instead.
      I have researched several other potential inclusion-exclusion reversals but none showed any improvement regarding the found differences, stronger: all showed severe deteriorations of the differences, introducing both new teams with differences and bigger existing differences.
      - indeed mick, all subsequent matches where for one or both of the teams the starting points are changed because of an inclusion/exclusion earlier, need to be recalculated also. Excel is quite good and fast in such cascading calculations.

      More research is needed to resolve the remaining CAF differences and the PAN/NCA difference properly...

      Delete
    10. Btw Martin, I really appreciate your efforts in finding and reporting about fixes. Good to know there are more people than just me interested in producing the correct information.
      And this whole frustrating shebang is necessary, because FIFA neglects their core business: administrating WORLD football properly.

      Delete
    11. So, you use excel for those calculations? I aske because i can find no formulas or macros in attached calculation sheets.

      Delete
    12. The uploaded sheet is no more than a result sheet with values. I don't want to share the sheets where all the magic happens.

      Delete
    13. Ed, when can we expect an update of that results sheet (including matches played in the last window)? Een mooie asjeblief met suiker erop could be yours.

      Delete
    14. Well, the fixes for the remaining small differences are not obvious, so I'll have to decide what to do with them. I think I will introduce the necessary small corrections, counting from this month. It will probably have to wait until next weekend though, I'm rather busy at the moment.

      Delete
  4. This comment has been removed by the author.

    ReplyDelete
  5. Hi Ed, here what could be solve the problem;

    1. Regarding on some CAF teams differences, I found the culprit which are 3 friendlies played on late October 2020 / early November 2020 : ETH-ZAM (both 22/10 and 25/10) and ETH-SDN (6/11). The points obtained on these matches were eventually has been counted towards November 2020 ranking (ETH = 1064.332168 points, ZAM = 1266.274405, SDN = 1123.864875). However, it seemingly FIFA has abolished the points obtained on these two matches: ETH-ZAM (22/10) (with exchange 1.53335 points) and ETH-SDN (6/11) (with exchange 0.24870 points) , which eventually changing ETH, ZAM and SDN points in December 2020 ranking (ETH = 1065.616984, ZAM = 1264.740889, SDN = 1124.113574). These points were eventually affected the other matches, generally in CHAN 2020 and CAN qualifiers, which played on January until March 2021, and finally fixing the points on some CAF teams. These scenario is quite similar when FIFA abolished the points in match CIV-ZAM (played on 19/6/2019), which changing their points during February 2020 ranking released.

    2. Regarding PAN-NCA differences, I believe the only way to fix their points is by assuming match PAN-NCA (played on 25/2/2020) played after PAN-GUA (played 4/3/2020). It seems awkward, but let shall be agree this error should be concerned for the further findings. Credit to Martin.

    3. Regarding GUY-BRB differences, as we thought, the points obtained during the match both played on 6/9/2018 has been abolished once the match was forfeited. Hence, GUY obtained 942.4217004 points and BRB obtained 1006.46645347 points. These points eventually being included as starting point for ‘reformed’ match which resulted GUY 3-0 BRB, with exchange 8.4170 points, finally fixing the both teams’ points.

    Hopefully this will be useful to solve your fixing and not regret to try. Thanks for Mick’s and Martin’s effort. Really appreciate it.

    ReplyDelete
    Replies
    1. just let me know if this doesn't work out. I am waiting for your response

      Delete
  6. Hi mono,

    sorry but I hadn't had the time until now to thoroughly research your suggestions. I think you hit the sweet spot regarding the CAF-teams. The correction of the matches ETH-ZAM and ETH-SDN in the December 2020 ranking solves indeed all remaining differences in CAF. Well done!

    Regarding the CONCACAF teams: after the plausible GUY-BRB adjustment Martin suggested, the only remaining differences of 0.01 point are PAN and NCA. The suggestion of Martin to assume that GUA-PAN was played before NCA-PAN (instead of what actually happened) indeed solves those tiny differences also, so that must be somehow what happened. I just try to get my head around what has happened there exactly. Because the first published ranking after those matches (April 2020) with the integer point-value of 1305 for PAN showed already that both PAN-matches must have been included in that ranking. If the NCA-PAN match were only first included in the next to April (i.e. June) ranking or later the PAN points in April would have been 1306. This I can still not understand. But I can't find any other possible cause also, so I settle for this 'explanation'.

    Anyway, I want to thank Martin and mono for their efforts and (reasonable) suggestions. I'm now producing the new match points calculation sheet from scratch because the earliest changes already occured in December 2019 (the BRB issues). When ready (somewhere tomorrow) it will be uploaded to it's usual location.

    ReplyDelete
    Replies
    1. You're welcome, Ed. That's fine for me although. Let's talk about the future. Actually, I'm really not hyped for the next ranking release. I believe there is a few of friendly matches scheduled before that release, which outside the International Match Calendar. But I also believe there will be a lot of changes for August's release. As you know, various tournaments scheduled before that release. We got EURO 2020, Copa America, CONCACAF Gold Cup and FWC Qualifiers, which is a lot of work to do.

      Delete
  7. Hi Ed
    I'm still having trouble coming up with the reversing exchange points on the matches on 12-6-18 between Barbados-Guyana and Barbados-El Salvador. Is there a way to show the step by step process to recalculate.
    Thanks

    ReplyDelete
  8. Well, that's not so easy.

    First the facts:
    2018-09-06 the CNL qualifier GUY-BRB wsa played and ended 2-2. Matchpoints exchanged from BRB to GUY 0,971.
    2018-10-13 the CNL qualifier SLV-BRB was played and ended 3-0. Matchpoints exchanged from BRB to SLV 3,307.
    In the December 2018 ranking there was an adjustment of the points for BRB and GUY (and probably SLV too) because both CNL qualifiers were awarded 3-0 to resp. GUY and SLV as BRB fielded one or more ineligible players in these matches. Based on the rounded points in that ranking BRB should have 998 (instead of my calculated and rounded 1005), GUY should have 951 (instead of my calculated and rounded 943) and SLV's points of 1327 were 'correct'.

    The following calculation has the assumption that both of the matches, so GUY-BRB and SLV-BRB are awarded in December 2018 3-0 against Barbados and recalculated with some starting points for each of the concerned teams and that there are correction terms introduced in the December 2018 ranking because of that. Furthermore I assume that the correction term for each team was determined as the difference between what they already received for the original match and the resulting points of the recalculated match.

    Now in the March 2021 ranking we finally have two decimal resulting points for all teams and can I determine that, with reverse engineering and recalculating all affected matches from December 2018 until now, the correction term for the match GUY-BRB must have been 7,448 and for SLV-BRB must have been 0,001. Focussing on the GUY-BRB match, this correction term -and the original points exchange of 0,971 points- means that in the recalculated awarded match 8,419 points must have been exchanged. The elo win expectancy for GUY in the recalculated match can then be determined as 15*(1-We) = 8,419 or We = 0,4387. The difference between the starting points of BRB and GUY in the recalculated match must then be 600 * log(1/We - 1) = 64,181. But that's as far as I can go. Exactly which starting points were used by FIFA for each team and how they came by them I don't know.
    Martin suggested to take the average of the starting points in December 2018 without the awarded match included and the starting points in December 2018 with the awarded match included, but that leads to a win expectancy of 0,4398. I've tried other variations, but none gave the exact desired win expectancy.

    So I decided to just introduce the correct correction term for the concerned teams in December 2018: BRB -7,449, GUY 7,448 and SLV 0,001. At least with these terms I can correctly reproduce the complete March 2021 ranking rounded on two decimals.

    ReplyDelete
  9. Hi Ed!
    Let me ask you how did you get that exchanged points 8.419?
    I made several updates in excel sheet where you provided results. But didn't rewrite your cells. I rather put formulas which exactly calculate expectancy, points exchanged and points after match. I made several permutations, i.e. removed match from 06.Sept.2018 and inserted new one at 06.Dec.2018. Your updated values (998,046 for BRB and 950,841 for GUY) did not match. Then I tried simply to update score from 06.09 from 2:2 to 3:0 for GUY. It also did not lead me to correct update of points exchanged. How exactly was recalculation made? And why is El Salvador only team which points are updated as well, since BRB played against VIR and VIN in the meanwhile too?

    ReplyDelete
  10. Hi mick, the premisse is that the match GUY-BRB was played on 6 Sep 2018 and points were then calculated according to the initial 2:2 draw. The resulting points for GUY and BRB were subsequently used in their next matches in October (GUY-TCA and SLV-BRB) and November 2018 (VIR-BRB). Only in December 2018 the GUY-BRB match and the SLV-BRB were awarded 3:0 to GUY resp. SLV and there were at that moment correction terms introduced to the points totals of GUY, BRB and SLV because of these awards.
    These correction terms were subsequently applied in the starting points for the first matches of GUY, BRB and SLV after the December 2018 ranking was published. The points calculations of the October and November 2018 matches of GUY, BRB and SLV were not adjusted.

    As I wrote, I could only determine the 'exact' value of these correction terms because we now have two decimal values for each team available in the March 2021 ranking and I could reverse engineer from March 2021 backwards to December 2018 the 'exact' starting points for each team which played in any affected match between December 2018 and now.
    I still can't think of a logical way that UEFA followed to determine the value of the respective correction terms. All logical explanations I could think of and explored didn't result in the exact correction terms, so a few small differences in the March 2021 ranking remained for each explanation I could come up with.

    ReplyDelete
  11. Did I say UEFA ? I mean FIFA, of course :)

    ReplyDelete
  12. I am just curious how did you come up with 8,419516 exchange. I tried all possible combinations to adjust score 3:0 for GUY and never got such a value:

    1) 1003,871 936,000 (values before Sept match) -> 8,471
    2) 1005,495 (BRB value before Dec) 936,000(GUY value before Sept) -> 8,494
    3) 1005,495 943,393 (values before Dec, I immagined that new game was played at Dec) -> 8,390
    4) 1003,871 (BRB before Sept match) 943,393 (GUY before Dec) -> 8,366



    ReplyDelete
  13. I see I still use the value 8,419516 in the sheet. That's not correct, it should be 8,419 without the rest of the decimals.
    I've adjusted that in the sheet.

    Like I wrote: my experiments in trying to achieve this value were also in vain. I decided to let it rest and just apply this correction factor so that I can at least correctly reproduce the March 2021 values with two decimals.
    It's a pity, but we are talking now about differences of less than 0,01 point. I can live with that :)

    ReplyDelete
  14. Of course, however, thanks for being so active 🙂

    ReplyDelete