Paul M. Sommers
Department of Economics, Middlebury College, Middlebury, VT, USA.
DOI: 10.4236/jss.2026.148047   PDF    HTML   XML   1 Downloads   31 Views   Citations

Abstract

A regression model is used to predict points (countries are awarded 3 points for a win, 1 point for a draw, and 0 points for a loss) in the men’s FIFA World Cup between 1994 and 2022. Key predictor variables include population size of the 248 country-tournament observations and their GDP per capita in their year of participation. Both variables are statistically significant. Predicted point totals are then compared to actual points to compile a list of overachievers and underachievers over the last eight World Cup tournaments.

Keywords

FIFA World Cup, Soccer, World Cup Predicted Points

Share and Cite:

1. Introduction

The Fédération Internationale de Football Association (FIFA) World Cup is arguably the most-watched single sports event in the world. Twenty-four countries entered the World Cup of 1994 hosted in the United States. Thirty-two countries entered the quadrennial men’s tournament in 1998 and every subsequent World Cup until 2026, when three countries—Canada, the United States, and Mexico—hosted forty-eight countries.

In 1994, FIFA began awarding three points for a win. Previously, teams earned two points for a win. The motivation for the rule change was to encourage more offense and discourage tactics that relied on settling for draws. Under the 32-team format used through 2022, a World Cup team could earn a maximum of 21 points (3 points each for 3 wins in the three-game round-robin Group Stage, 3 points each for the 4 wins in the single-elimination Round of 16, Quarterfinals, Semifinals, and Finals). At the conclusion of the tournament, countries might understandably want to know if they performed better or worse than expected.

The purpose of this brief research note is to help soccer fans answer the question of whether their country’s team performed better or worse than expected based not on the country’s pre-tournament seeding, but on the size of the country (by population), the country’s gross domestic product per capita, and what continent the country is located in. Apart from population size, richer countries would be expected to do better than poorer countries because they can afford better sports facilities, better coaching staffs, and talent development programs, not to mention “import … good soccer experience” (Kuper & Szymanski, 2009), that is, lure soccer players to become citizens (or nationals) of the country they will represent. At the conclusion of each tournament, which countries were overachievers? Which ones were underachievers?

2. The Data

For each of the eight FIFA World Cup tournaments between 1994 and 2022, data were collected on each country’s point total, current population and gross domestic product per capita (hereafter, GDP/N) in current U.S. dollars (at the time of that year’s World Cup). Population and GDP/N for each of the 248 country-tournament observations in our sample are from the World Bank Open Data at https://databank.worldbank.org/reports.aspx?source=2&series=SP.POP.TOTL&country=# and https://databank.worldbank.org/reports.aspx?source=2&series=NY.GDP.PCAP.CD&country=, respectively.

Several countries that played in FIFA World Cup tournaments during the study period experienced changes to their political designation or name. For example, FIFA allowed Serbia and Montenegro to compete under the name and flag of Serbia and Montenegro in the 2006 FIFA World Cup, although Montenegro had declared its independence just days before the 2006 World Cup. Serbia played as an independent nation without Montenegro at the 2010 World Cup.

Tournament point totals (3 for a win, 1 point for a draw, and 0 for a loss) increase when teams advance and play additional matches. The point totals reported in this study combine group-stage performance with knockout-stage advancement. Knockout-stage matches that end in a draw are decided by penalty shootouts. Penalty shootout victories are counted as draws, worth one point each. For example, Argentina’s 14 points in the 2022 World Cup are based on two victories in the group stage (6 points), two knockout stage wins (6 points), and two knockout stage shootout victories (one in the quarterfinals and one in the final, for an additional 2 points).

3. Methodology

To assess the effect of population size and GDP per capita on a country’s point total in each of the eight World Cups between 1994 and 2022, an ordinary least squares (OLS) regression of the following form was run:

$\begin{array}{l}Point{s}_{i,t}=b_0+b_{1}Populatio{n}_{i,t}+b_{2}Populationsquare{d}_{i,t}+b_{3}GDP/N_{i,t}\\ +b_{4}GDP/Nsquare{d}_{i,t}+b_{5}Europe+b_{6}Africa+b_{7}SouthAmerica\end{array}$ (1)

where Points_i,t is country i’s point total (3 for a win, 1 for a draw, and 0 for a loss) in World Cup year t; Population_i,t is country i’s total population (in millions) as of July 1^st in year t; GDP/N_i,t is country i’s GDP per capita (in thousands of current U.S. dollars) in year t; and Europe, Africa, and South America are regional dummy variables identifying the country’s location (i.e., all other countries in Asia, North America, and Oceania are excluded to avoid the so-called “dummy variable trap”). Population and GDP/N are each squared to capture a curved or inverted-U relationship between, say, Points and Population (or Points and GDP/N). Countries with large populations that have participated in the World Cup—China (2002), the United States (all years between 1994 and 2022 with the exception of 2018), and Nigeria (1994, 1998, 2002, 2014, 2018), to name a few—have struggled in the FIFA World Cup since 1994 to advance far in the knockout stages. Similarly, a linear model assumes that Points increase with GDP/N. But, rich countries that have participated in the World Cup between 1994 and 2022—Switzerland (all eight World Cups but 1998 and 2002), Norway (1994 and 1998), and Qatar (2022)—have rarely advanced to the later stages of the tournament. After Population and GDP/N are taken into account, three regional dummies capture the dominant effect countries from these continents have had on the tournament. Between 1994 and 2022, all finalists and all but three semifinalists (with the exceptions of South Korea and Türkiye in 2002 and Morocco in 2022) have come from only Europe and South America.

The resulting regression model in Equation (1) can be used to generate predicted points. The difference between actual points minus predicted points (hereafter, the residual) will be used to identify overachievers (a positive residual) and underachievers (a negative residual). Previous studies have relied on explaining variation in goal difference per game, the dependent variable, and not points (see Kuper & Szymanski, 2009: pp. 289-290) or performance in the preliminaries leading up to a World Cup (Sylla & Magel, 2016). But, in the World Cup group stage, points, not goal differential, determine which teams advance, while the difference between goals scored and allowed serves as the tiebreaker to separate teams that are tied on points.

4. The Results

Table 1 shows the regression results with and without the regional dummies. The regression results with the regional dummies show that Population and Population squared (b₁ = 0.023393, p < 0.01 and b₂ = −0.00002, p < 0.01) are significant at better than the 0.01 level. GDP/N and GDP/N squared (b₃ = 0.09317, p = 0.032 and b₄ = −0.00111, p = 0.038) are significant at better than the 0.05 level. Hawksworth et al. (2014) found that neither population nor GDP per capita explained how well a country in the World Cup performs. For both Population and GDP/N, the number of points is an inverted U-shaped relationship. That is, as population (or GDP per capita) increase, points also increase up to a certain point, after which points

decrease. The peak GDP per capita (where $\frac{\partial Points}{\partial GDP/N}\text{=}0$ and $GDP/N=-b_3/2b_4$ )

occurs at $41,786, close to the GDP per capita of France in 2022 ($40,989). Points are significantly higher for countries in South America (b₇ = 5.43548, p < 0.01) and Europe (b₅ = 4.75824, p < 0.01), but not for countries in Africa (b₆ = 1.53761, p = 0.106). The overall R² for Equation (1) is 0.27.

Table 1. OLS Regression Results, FIFA World Cup Points, 1994-2022.

Dependent variable: World Cup points
Variable	With Regional Dummies	Without Regional Dummies
Population (millions)	0.023393***	0.01005*
	(0.00548)	(0.0056)
Population squared	− 0.00002***	− 0.00001**
	(5.06E−06)	(5.38E−06)
GDP per capita (thousands)	0. 09317**	0.14390***
	(0.04308)	(0.03842)
GDP per capita squared	− 0.00111**	− 0.00169***
	(0.00054)	(0.00053)
Europe	4.75824***
	(0.71557)
Africa	1.53761
	(0.9485)
South America	5.43548***
	(0.87296)
Constant	0.37979	3.57718***
	(0.80561)	(0.54751)
Observations (N)	248	248
R-squared	0.277	0.08
Adjusted R-squared	0.256	0.065

***p < 0.01, **p < 0.05, *p < 0.10, Standard errors in parentheses.

Despite differences in host status, qualification pathways, group composition, and tournament-specific competitive conditions, the estimated slope coefficients and their statistical significance are virtually the same when Equation (1) controls for tournament fixed effects. Although there are repeated observations of several of the same national teams during the study period (Argentina, Brazil, Germany, Mexico, and South Korea, and Spain appeared in all eight tournaments), country-clustered standard errors increase slightly but not enough to affect the significance tests on any of the predictor variables.

Equation (1) with the regional dummies can be used to estimate a country’s predicted number of points. For example, Brazil 2002 (with a 2002 population, in millions, of 178.503 and a 2002 GDP/N, in thousands of U.S. dollars, of 2.8559) would have a predicted number of points equal to 9.6363. Brazil 2002’s residual (i.e., actual points minus predicted points) is 11.3637, the largest residual and hence the sample’s biggest overachiever.

Table 2 lists all 248 country-tournament observations in the World Cup between 1994 and 2022, from biggest overachiever (largest positive residual) to biggest underachiever (largest negative residual).

Table 2. Over- and Under-achievers in the FIFA World Cup, 1994-2022.

Rank	Year	Country	Points	Predicted Points	Residual
1	2002	Brazil1	21	9.6363	11.3637
2	1998	France	19	8.1042	10.8958
3	2018	Belgium	18	7.3133	10.6867
4	2010	Netherlands	18	7.3671	10.6329
5	2018	France	19	8.569	10.431
6	2014	Germany	19	8.7959	10.2041
7	2010	Spain	18	7.9946	10.0054
8	2014	Netherlands	17	7.3219	9.6781
9	1998	Croatia	15	5.7395	9.2605
10	2006	Italy	17	8.3095	8.6905
11	2002	South Korea	11	2.5144	8.4856
12	2014	Argentina	16	7.7592	8.2408
13	2022	Morocco	11	3.0732	7.9268
14	1994	Brazil	17	9.3492	7.6508
15	2014	Costa Rica	9	1.3748	7.6252
16	2018	Croatia	14	6.4051	7.5949
17	2002	Germany	16	8.5876	7.4124
18	2022	France	16	8.5898	7.4102
19	2006	Germany	16	8.8557	7.1443
20	2006	France	15	8.4634	6.5366
21	2006	Portugal	13	6.7918	6.2083
22	1994	Italy	14	7.7928	6.2072
23	2010	Germany	15	8.8689	6.1311
24	2022	Argentina	14	7.9215	6.0785
25	2002	Türkiye	13	6.9404	6.0596
26	2002	Senegal	8	2.222	5.778
27	2010	Ghana	8	2.6169	5.3831
28	1994	Sweden	12	7.0124	4.9876
29	1998	Netherlands	12	7.2332	4.7668
30	2018	Uruguay	12	7.2749	4.7251
31	2014	Belgium	12	7.3015	4.6985
32	1994	Bulgaria	10	5.4399	4.5601
33	1994	Saudi Arabia	6	1.5374	4.4626
34	2014	Colombia	12	7.5511	4.4489
35	2010	Argentina	12	7.587	4.413
36	1994	Romania	10	5.7812	4.2188
37	2010	Uruguay	11	6.8924	4.1076
38	2022	Netherlands	11	7.1577	3.8423
39	2006	Argentina	11	7.2205	3.7795
40	2002	Mexico	7	3.2301	3.7699
41	2018	Russia	8	4.2797	3.7203
42	2022	Australia	6	2.3126	3.6875
43	2002	Spain	11	7.3422	3.6578
44	2022	Senegal	6	2.4682	3.5318
45	2006	Ghana	6	2.5297	3.4703
46	2022	Croatia	10	6.5683	3.4317
47	1998	Brazil	13	9.6697	3.3303
48	2014	Mexico	7	3.8232	3.1768
49	2002	Costa Rica	4	0.8354	3.1646
50	1998	Italy	11	7.9333	3.0667
51	1998	Argentina	10	7.3312	2.6688
52	2006	England	11	8.4276	2.5724
53	1998	Jamaica	3	0.7454	2.2546
54	2022	Switzerland	6	3.7894	2.2106
55	2018	Mexico	6	3.8223	2.1777
56	1994	Mexico	5	2.8499	2.1501
57	2022	Japan	7	4.9096	2.0904
58	2002	Japan	7	4.916	2.084
59	1998	Mexico	5	2.9457	2.0543
60	2010	Japan	7	4.9932	2.0068
61	2022	Portugal	9	6.9982	2.0018
62	2006	Brazil	12	10.0264	1.9736
63	1994	Netherlands	9	7.1139	1.8861
64	2018	Sweden	9	7.1539	1.8461
65	1994	Nigeria	6	4.2875	1.7125
66	2002	Cameroon	4	2.3534	1.6466
67	2018	Senegal	4	2.4178	1.5822
68	2022	Tunisia	4	2.5282	1.4718
69	2018	England	10	8.5469	1.4531
70	2022	England	10	8.5484	1.4516
71	2014	France	10	8.5495	1.4505
72	1998	Yugoslavia	7	5.564	1.436
73	2010	Ivory Coast	4	2.5758	1.4242
74	2022	Costa Rica	3	1.5824	1.4176
75	1998	Nigeria	6	4.6113	1.3887
76	1994	Germany	10	8.6254	1.3746
77	2014	Switzerland	6	4.628	1.372
78	1998	Germany	10	8.644	1.356
79	2018	Iran	4	2.6721	1.3279
80	2022	Cameroon	4	2.6958	1.3042
81	1998	Morocco	4	2.7013	1.2987
82	2010	Australia	4	2.709	1.291
83	2006	Australia	4	2.7676	1.2324
84	1998	Romania	7	5.8236	1.1764
85	2006	Spain	9	7.8863	1.1137
86	2006	Switzerland	8	6.8925	1.1075
87	2002	China	0	−1.0375	1.0375
88	1998	Iran	3	1.9652	1.0348
89	2006	South Korea	4	3.0046	0.9954
90	1994	Spain	8	7.0819	0.9181
91	2010	South Korea	4	3.0887	0.9113
92	2002	South Africa	4	3.2389	0.7611
93	2014	Algeria	4	3.3315	0.6685
94	2010	New Zealand	3	2.3561	0.6439
95	2006	Ukraine	7	6.4154	0.5846
96	2022	South Korea	4	3.43	0.57
97	2006	Mexico	4	3.4534	0.5466
98	2006	Ivory Coast	3	2.5029	0.4971
99	2018	Tunisia	3	2.5134	0.4866
100	2010	Mexico	4	3.5909	0.4091
101	2014	Ivory Coast	3	2.6607	0.3393
102	2014	Brazil	11	10.703	0.297
103	2022	Iran	3	2.7352	0.2648
104	2010	Honduras	1	0.7466	0.2534
105	2010	South Africa	4	3.7613	0.2387
106	2018	Saudi Arabia	3	2.8432	0.1568
107	2022	Ghana	3	2.874	0.126
108	2022	Qatar	0	−0.0654	0.0654
109	2022	Mexico	4	3.9886	0.0114
110	2018	Switzerland	5	5.0031	−0.0031
111	2022	Saudi Arabia	3	3.0473	−0.0473
112	2002	Denmark	7	7.1288	−0.1288
113	1998	Denmark	7	7.1289	−0.1289
114	2002	England	8	8.2552	−0.2552
115	1994	South Korea	2	2.2575	−0.2575
116	1998	Cameroon	2	2.3182	−0.3182
117	2014	Chile	7	7.3455	−0.3455
118	2010	Paraguay	6	6.3664	−0.3664
119	2006	Ecuador	6	6.4315	−0.4315
120	2018	South Korea	3	3.4362	−0.4362
121	2006	Netherlands	7	7.4496	−0.4496
122	2002	United States	7	7.4506	−0.4506
123	2018	Costa Rica	1	1.4936	−0.4936
124	2018	Colombia	7	7.4993	−0.4993
125	1998	Saudi Arabia	1	1.5332	−0.5332
126	2010	Brazil	10	10.5437	−0.5437
127	2002	Russia	3	3.5888	−0.5888
128	2018	Brazil	10	10.5913	−0.5913
129	2006	Angola	2	2.6414	−0.6414
130	2022	Brazil	10	10.6546	−0.6546
131	1994	Russia	3	3.6688	−0.6688
132	2014	Honduras	0	0.788	−0.788
133	2018	Denmark	6	6.794	−0.794
134	2006	Costa Rica	0	0.9395	−0.9395
135	1994	Belgium	6	6.9749	−0.9749
136	2018	Japan	4	4.9828	−0.9828
137	2010	The Democratic People's Republic of Korea	0	1.0054	−1.0054
138	2002	Republic of Ireland	6	7.0845	−1.0845
139	1998	Paraguay	5	6.1015	−1.1015
140	1998	South Korea	1	2.1391	−1.1391
141	2006	Saudi Arabia	1	2.1655	−1.1655
142	1998	England	7	8.1874	−1.1874
143	2010	Chile	6	7.2107	−1.2107
144	2014	Uruguay	6	7.2265	−1.2265
145	1994	Argentina	6	7.2333	−1.2333
146	1998	South Africa	2	3.2539	−1.2539
147	1994	Cameroon	1	2.2762	−1.2762
148	2006	Iran	1	2.3159	−1.3159
149	1998	Tunisia	1	2.3459	−1.3459
150	2002	Tunisia	1	2.359	−1.359
151	2006	Tunisia	1	2.4546	−1.4546
152	2022	United States	5	6.5351	−1.5351
153	2018	Australia	1	2.6263	−1.6263
154	2014	Iran	1	2.6515	−1.6515
155	2002	Saudi Arabia	0	1.654	−1.654
156	2018	Panama	0	1.691	−1.691
157	2014	Ghana	1	2.7357	−1.7357
158	2010	Switzerland	4	5.808	−1.808
159	2014	Greece	5	6.8773	−1.8773
160	2014	Nigeria	4	5.8798	−1.8798
161	2010	Portugal	5	6.9175	−1.9175
162	2018	Portugal	5	6.9521	−1.9521
163	2018	Spain	6	7.9976	−1.9976
164	2002	Belgium	5	7.0105	−2.0105
165	2018	Morocco	1	3.0437	−2.0437
166	2002	Paraguay	4	6.0632	−2.0632
167	2006	Togo	0	2.1023	−2.1023
168	1998	Norway	5	7.1291	−2.1291
169	2002	Sweden	5	7.136	−2.136
170	2010	Algeria	1	3.1696	−2.1696
171	2006	Sweden	5	7.2705	−2.2705
172	2014	South Korea	1	3.3263	−2.3263
173	2014	Australia	0	2.372	−2.372
174	1994	Republic of Ireland	4	6.4215	−2.4215
175	2014	Russia	2	4.4439	−2.4439
176	2010	Cameroon	0	2.4982	−2.4982
177	2010	Slovakia	4	6.5198	−2.5198
178	2014	Cameroon	0	2.5748	−2.5748
179	1994	Morocco	0	2.6372	−2.6372
180	2010	United States	5	7.6737	−2.6737
181	2014	Bosnia and Herzegovina	3	5.676	−2.676
182	2014	Ecuador	4	6.7366	−2.7366
183	2010	Slovenia	4	6.7528	−2.7528
184	2002	Croatia	3	5.7747	−2.7747
185	2022	Ecuador	4	6.7859	−2.7859
186	2010	Serbia	3	5.8242	−2.8242
187	2014	Portugal	4	6.8948	−2.8948
188	2002	Argentina	4	6.9092	−2.9092
189	1994	United States	4	6.9312	−2.9312
190	2018	Serbia	3	5.9411	−2.9411
191	2022	Canada	0	2.9673	−2.9673
192	1994	Norway	4	7.0072	−3.0072
193	2022	Spain	5	8.0122	−3.0122
194	2018	Nigeria	3	6.0922	−3.0922
195	2006	Paraguay	3	6.1641	−3.1641
196	1998	Spain	4	7.2174	−3.2174
197	1994	Switzerland	4	7.2431	−3.2431
198	2022	Belgium	4	7.268	−3.268
199	2002	Ecuador	3	6.3064	−3.3064
200	2014	Croatia	3	6.3334	−3.3334
201	2022	Poland	4	7.3357	−3.3357
202	2022	Uruguay	4	7.3598	−3.3598
203	2002	Portugal	3	6.3985	−3.3985
204	2002	Poland	3	6.4606	−3.4606
205	2010	England	5	8.4729	−3.4729
206	2006	Czech Republic	3	6.5377	−3.5377
207	1998	Chile	3	6.637	−3.637
208	2014	United States	4	7.6388	−3.6388
209	2006	Poland	3	6.7556	−3.7556
210	2018	Argentina	4	7.7627	−3.7627
211	2002	Nigeria	1	4.7635	−3.7635
212	1994	Colombia	3	6.824	−3.824
213	1998	Colombia	3	6.9071	−3.9071
214	2010	Denmark	3	6.917	−3.917
C215	2002	Italy	4	7.9409	−3.9409
216	2006	Japan	1	4.9707	−3.9707
217	2014	Japan	1	4.9853	−3.9853
218	1998	Belgium	3	7.0198	−4.0198
219	2010	Greece	3	7.0872	−4.0872
220	2018	Peru	3	7.1379	−4.1379
221	2006	Croatia	2	6.1627	−4.1627
222	2018	Poland	3	7.1843	−4.1843
223	2002	Uruguay	2	6.2592	−4.2592
224	2018	Egypt	0	4.3998	−4.3998
225	2010	Nigeria	1	5.4823	−4.4823
226	1998	Bulgaria	1	5.4958	−4.4958
227	2022	Germany	4	8.8129	−4.8129
228	2018	Iceland	1	5.8157	−4.8157
229	1998	Japan	0	4.8873	−4.8873
230	2014	Spain	3	7.9675	−4.9675
231	1998	Austria	2	7.0333	−5.0333
232	1994	Bolivia	1	6.0663	−5.0663
233	2022	Serbia	1	6.1151	−5.1151
234	2014	Italy	3	8.389	−5.389
235	2022	Denmark	1	6.4688	−5.4688
236	2006	Serbia and Montenegro	0	5.7065	−5.7065
237	2018	Germany	3	8.8355	−5.8355
238	1998	Scotland	1	6.882	−5.882
239	2006	Trinidad and Tobago	1	6.941	−5.941
240	2022	Wales	1	7.0718	−6.0718
241	2002	Slovenia	0	6.1181	−6.1181
242	1994	Greece	0	6.2675	−6.2675
243	2010	Italy	2	8.3762	−6.3762
244	2006	United States	1	7.5745	−6.5745
245	2002	France	1	8.1104	−7.1104
246	1998	United States	0	7.2296	−7.2296
247	2014	England	1	8.4766	−7.4766
248	2010	France	1	8.5234	−7.5234

¹Italicized names are winners of the men’s FIFA World Cup.

The three biggest overachievers are Brazil 2002, France 1998, and Belgium 2018. Brazil did exceptionally well in the 2002 World Cup, winning all seven of their matches (without needing extra time or penalty kicks) and that year’s championship. France hosted the tournament in 1998 and defeated Brazil 3-0 in the final to claim their first-ever world title. Belgium did incredibly well in the 2018 World Cup. They finished in third place, which was the country’s best-ever result in the tournament’s history.

The three biggest underachievers are the United States 1998, England 2014, and France 2010. The United States in the 1998 World Cup lost all three group stage games and finished in last place in their group. England in the 2014 FIFA World Cup hosted by Brazil finished at the bottom of their group, losing games to Italy and Uruguay and recording a draw with Costa Rica. England finished with just one point, their worst-ever performance in the tournament’s history. The biggest underachiever was France in 2010 which was eliminated in the first round after finishing last in their group.

5. Concluding Remarks

A regression model using a country’s population and GDP per capita at the time of its participation in the FIFA World Cup is estimated to predict points earned by countries between 1994 and 2022. A comparison between actual points earned and predicted points enables one to easily identify overachievers (actual points earned exceed predicted points) as well as underachievers (predicted points exceed actual points).

Future researchers might estimate a regression model to predict points in the Women’s FIFA World Cup which also operates on a four-year cycle. Although 32 countries will participate in the FIFA Women’s FIFA World Cup in 2027 hosted by Brazil, an extension of the work presented here for the men’s FIFA World Cup to include the results of the 2026 World Cup must recognize that the 48 countries in 2026 could earn more points. That is, the 2026 tournament added a new Round of 32 knockout phase. Spain and Argentina will have played a total of eight (not seven) games each by the conclusion of the 2026 World Cup final.

Conflicts of Interest

The author declares no conflicts of interest regarding the publication of this paper.

References

[1]	Hawksworth, J., Zimmern, W., & Broadfield, D. (2014). The PwC World Cup Index: What Can the Dismal Science Tell Us about the Beautiful Game? https://www.pwc.com/im/en/assets/document/pwc_world_cup_index_-_june_2014.pdf
[2]	Kuper, S. & Szymanski, S. (2009). Soccernomics. https://www.scribd.com/document/980987303/Soccernomics-PDF
[3]	Sylla, M. & Magel, R. (2016). Predicting the Winner of Games in World Cup Soccer Matches. Journal of Advance Research in Mathematics and Statistics, 3, 10-21. [Google Scholar] [CrossRef]