Evaluation of Rio 2016 Olympic medal tally forecasts: Who has won?

Fabian ten Kate, Gerard Kuper and Elmer Sterken

August 2016

 

At the Rio 2016 Summer Games 974 medals have been awarded to athletes from 87 countries. With 306 medal events (with rugby sevens and golf making their debut) and 53 events awarding two bronze medals the total number of medals is 659. However in some events athletes have tied:

         In the womenís 100m freestyle swimming event two gold medals and no silver medals are awarded.

         In the 100m butterfly swimming event for men three silver medals, and no bronze medals, are awarded.

         In the men's 200m single kayak event two bronze medals are awarded.

         In the women's 100m backstroke swimming event two bronze medals are awarded.

Eventually, 307 gold medals, 307 silver medals and 360 bronze medals were awarded.

 

Kosovo (gold, women judo, 52kg) Fiji (gold, rugby sevens) and Jordan (gold, men taekwondo, 68kg) won their first ever Olympic medal

Medal winning is concentrated: the Top 8 from the medal tally won 47% of all medals which is slightly below the 50% in 2004, 2008, and 2012. The Top 8 took 54% ofthe gold medals which is lower compared to 58% in Londen 2012.

 

Actual results and forecasts

Table 1 Our predictions and the actual results (Top 15) ranked according to the actual results

Ten Kate et al.

Rio 2016

 

Rank

G

S

B

T

Rank

G

S

B

T

USA

1

35

33

34

102

1

46

37

38

121

GBR

4

17

13

13

43

2

27

23

17

67

CHN

2

35

21

18

74

3

26

18

26

70

RUS

3

24

26

33

83

4

19

18

19

56

GER

5

12

11

18

41

5

17

10

15

42

JPN

6

12

10

12

34

6

12

8

21

41

FRA

8

11

12

14

37

7

10

18

14

42

KOR

9

10

11

10

31

8

9

3

9

21

ITA

10

9

9

10

28

9

8

12

8

26

AUS

7

11

17

13

41

10

8

11

10

29

NED

12

8

9

7

24

11

8

7

4

19

HUN

14

7

5

4

16

12

8

3

4

15

BRA

11

8

9

10

27

13

7

6

6

19

ESP

23

3

8

5

16

14

7

4

7

18

KEN

19

4

4

4

12

15

6

6

1

13

 


 

From Table 1 we can conclude that

        China earned fewer gold medals than we anticipated, and is back at their pre-2008 performance;

        Also Russia performed worse than we expected obviously because of the ban of Russian athletes. Before the start of the Games we expected that Russia would lose seven gold medals, and this is what has happened;

        Great Britain performed better than we expected earning over 20 more medals than we predicted. This results equals their performance of the home Olympics in 2012;

        Australia did not meet our expectations, and Brazil did not fully capitalized their home advantage with rank 15 and six gold medals;

        we overestimated the total number of medals for the Netherlands by five medals, but we correctly estimated the number of gold medals.

 

Comparing forecasts

 

We compare ten forecasts published on http://www.topendsports.com/events/summer/medal-tally/predicted-2016.htm. Unfortunately, we only have the Top 10 of each forecast limiting our analysis.

 

There are two types of medal tally forecasts: expertís opinions (Table 2) and model-based forecasts (Table 3).

 

Table 2 Predicted Top 10 countries with total gold medals won based on regularly updated results

 

 

Gracenote (3 Aug 2016)

Cazeneuve-SI (29 July 2016)

Barra (Dec 2015)

Olympic medals predictions (4 Aug 2016)

AOC Benchmark (Dec 2015)

Only top 8

 

Rank

Gold

Rank

Gold

Rank

Gold

Rank

Gold

Rank

Gold

AUS

5

16

3

20

7

14

8

11

7

13

BRA

10

8

CHN

2

29

2

45

2

32

2

30

1

39

FRA

8

13

5

15

8

14

6

17

6

15

GBR

3

18

4

16

5

16

4

18

5

15

GER

4

16

7

14

6

16

5

17

8

13

HUN

ITA

JPN

7

14

8

13

4

18

7

14

4

18

KOR

9

10

9

10

9

12

10

7

NED

NZL

10

7

10

10

9

10

RUS

6

15

6

14

3

23

3

19

3

25

USA

1

38

1

45

1

35

1

50

2

35

 

Most expertís forecast are updated just before the start of the Games and account for Russian athletes who are banned from the Games because of use of doping in the past. The model of Gonzales (Tuck School) is the only model one publishing updated forecasts because of banned Russian track and field athletes.

 

Table 3 Predicted Top 10 countries with total gold medals won based on econometric models (NA is not in Top 10)

 

 

Goldman Sachs (July 2016)

WSJ (28 July 2016)

Gonzales (Tuck School) (July 2016)

ATASS Sports (5 Aug 2016)

Ten Kate et al. (July 2016)

 

Rank

Gold

Rank

Gold

Rank

Gold

Rank

Gold

Rank

Gold

AUS

3

17

10

7

4

17

7

11

BRA

CHN

2

36

2

35

2

38

2

32

2

35

FRA

7

11

7

12

7

11

5

14

8

11

GBR

3

23

4

16

3

30

3

18

4

17

GER

6

12

6

15

5

13

6

13

5

12

HUN

9

8

ITA

8

8

8

8

10

9

JPN

10

7

8

11

9

8

8

12

6

12

KOR

5

13

9

10

6

12

9

9

9

10

NED

10

9

NZL

10

8

RUS

4

14

5

16

4

18

7

13

3

24

USA

1

45

1

42

1

48

1

42

1

35

 

All but one forecast predict the USA to lead the medal table followed by China. In the AOC Benchmark China wins most gold medals followed by the USA. The number of expected gold medals for the USA ranges between 35 and 50. For China the range of gold medals is 29 to 45. In all forecasts France, Germany, Great Britain and Russia are expected to end in the Top 8. The host country is included in only in the forecasts of Gracenote.

 

There are various ways to compare the forecasts. Obviously, one way is to calculate rank correlation coefficients between the various ranking of countries and the actual ranking. To compare the gold medal forecasts we calculate the average deviation of the forecasts and the realisation (Root Mean Squared Error or RMSE) for gold. The results are in Table 4.

 

This table shows that:

-          Gonzales (Tuck School) outperforms all other forecasts both for accuracy of the rank as well as the gold medals;

-          Surprisingly, econometric models perform better than the experts;

-          In particular, the forecasts by Cazeneuve-SIand the Australian Olympic Committee (AOC Benchmark) underperform;

-          Gonzales (Tuck School), the website of Olympic medals predictions and Ten Kate et al. all accurately forecast the rank.

 


 

Table 4 Comparing Top 10 forecasts

RMSE a)

Kendall

 

 

 

 

Gold

tau-a b)

Gracenote (3 Aug 2016)1)

5,000

0,733

Cazeneuve-SI (29 July 2016)1)

8,349

0,556

Barra (Dec 2015)1)

6,542

0,733

Olympic medals predictions (4 Aug 2016)1)

4,637

0,778

AOC Benchmark (Dec 2015)1, 5)

8,456

0,500

Goldman Sachs (July 2016)2)

4,572

0,689

WSJ (28 July 2016)3)

5,648

0,644

Gonzales (Tuck School) (July 2016)2, 6)

4,483

0,778

ATASS Sports (5 Aug 2016)4)

5,459

0,556

Ten Kate et al. (July 2016)2)

6,033

0,778

Notes:

a) Measures the average deviation between two series

b) Measures the strength of the association between ranks

1) Predictions based on recent results, and regularly updated

2) Predictions based on econometric models

3) Hybrid: part subjective reporting, part statistics

4) Method unknown

5) Only lists Top 8

6) Adjusted for banning Russian athletes

 

The home country and TeamNL

We expected Brazil to win eight gold medals. Brazil won seven gold medals which is just one gold less then we predicted. In terms of total medals Brazil underperformed, and did not take full advantage of their home advantage.

 

All forecasts overestimated the total number of medals (19) won by TeamNL. The estimates range from 22 (Cazeneuve-SI) to 30 (Olympic medals predictions). TeamNL underperformed in terms of total medals won (19 realized versus 24 expected), but performed as we expected in terms of gold medals (8) won.