Women’s World Cup predictions and results

Whether or not you’ve been waiting all month for these, here they are.

Match Group
54% Germany 2 1 Canada 46% A
45% Nigeria 0 1 France 55% A
54% Japan 2 1 New Zealand 46% B
47% Mexico 1 1 England 53% B
53% USA 2 0 N Korea 47% C
44% Colombia 0 1 Sweden 56% C
52% Brazil 1 0 Australia 48% D
59% Norway 1 0 Eq Guinea 41% D
60% Germany 1 0 Nigeria 40% A
48% Canada 0 4 France 52% A
54% Japan 4 0 Mexico 46% B
45% New Zealand 1 2 England 55% B
59% USA 3 0 Colombia 41% C
46% N Korea 0 1 Sweden 54% C
51% Brazil 3 0 Norway 49% D
58% Australia 3 2 Eq Guinea 42% D
47% France 2 4 Germany 53% A
55% Canada 0 1 Nigeria 45% A
49% England 2 0 Japan 51% B
48% New Zealand 2 2 Mexico 52% B
49% Sweden 2 1 USA 51% C
54% N Korea 0 0 Colombia 46% C
36% Eq Guinea 0 3 Brazil 64% D
48% Australia 2 1 Norway 52% D
56% Germany 0 1 Japan 44% 1A v 2B
54% Sweden 3 1 Australia 46% 1C v 2D
51% England 1.3 1.4 France 49% 1B v 2A
51% Brazil 2.3 2.5 USA 49% 1D v 2C
47% Japan     Sweden 53% Semi
48% France     USA 52% Semi
            3rd place
            Championship

Based on these probabilities, with 28 matches played I would have expected 54.47% accuracy on the predictions. The model has actually had 69.64% accuracy. My conclusion: Well, the FIFA rankings (on which these probabilities are based) seem to represent the teams fairly well. I manipulated the FIFA rankings a bit. I increased Germany’s points by about 10% (host advantage); decreased Japan, Australia, and New Zealand’s points by about 2% (geographic separation); increased the European teams’ points by about 2% (geographic proximity); and decreased North Korea’s points by about 4% (political and geographic isolation). We can expect some exciting semifinals and although as of now Sweden statistically should come away with the Cup, expect the USA to win it all.

© 2011 John Schneider. All rights reserved.

World Cup probabilities 5.2

So, we are down to the 3rd place match and the final. My model likes Germany to win the 3rd place match and Spain the final. I personally want to see Germany and the Netherlands win. We’ll see.

Round of 16
61% Uruguay 2 1 Korea 39%
55% USA 1 2 Ghana 45%
64% Netherlands 2 1 Slovakia 36%
65% Brazil 3 0 Chile 35%
59% Argentina 3 1 Mexico 41%
50% Germany 4 1 England 50%
53% Paraguay 0.5 0.3 Japan 47%
56% Spain 1 0 Portugal 44%
Quarterfinals
54% Uruguay 1.4 1.2 Ghana 46%
45% Netherlands 2 1 Brazil 55%
52% Argentina 0 4 Germany 48%
32% Paraguay 0 1 Spain 68%
Semifinals
39% Uruguay 2 3 Netherlands 61%
42% Germany 0 1 Spain 58%
3rd Place
44% Uruguay     Germany 56%
Final
47% Netherlands     Spain 53%

The model continues to perform within expectations.

Matches played 62
Cumulative expected model accuracy 58.9%
Actual matches resulting in probable outcome 40
Actual model accuracy 64.5%

Distribution of expected resutls.
Fit of expected results versus actual results.

World Cup probabilities 5.1

Matches for the quarterfinals have been determined. Here is how the knockout round is going so far. The matches for the semifinals, third place match, and final (all in italics) are theoretical and based on the probabilities up to the quarterfinals.

Round of 16
61% Uruguay 2 1 Korea 39%
55% USA 1 2 Ghana 45%
64% Netherlands 2 1 Slovakia 36%
65% Brazil 3 0 Chile 35%
59% Argentina 3 1 Mexico 41%
50% Germany 4 1 England 50%
53% Paraguay 0.5 0.3 Japan 47%
56% Spain 1 0 Portugal 44%
Quarterfinals
54% Uruguay     Ghana 46%
45% Netherlands     Brazil 55%
52% Argentina     Germany 48%
32% Paraguay     Spain 68%
Semifinals
38% Uruguay     Brazil 62%
43% Argentina     Spain 57%
3rd Place
45% Uruguay     Argentina 55%
Final
50% Brazil     Spain 50%

As to the model accuracy to this point — it is still doing well.

Matches played 56
Cumulative expected model accuracy 59.0%
Actual matches resulting in probable outcome 36
Actual model accuracy 64.3%

Distribution of expected resutls
Fit of expected results vs actual

World Cup probabilities 5.0

Let’s start with the advancements to the Round of 16. The probabilities in the model were calculated using FIFA rankings at the beginning of the tournament. I did not adjust the probabilities after each match like I do for the match predictions. It’s possible to do so, but more work than I wanted to do this year. Still, the model faired well, though performing slightly better than expected.

Teams advanced 16
Cumulative expected model accuracy 57.73%
Actual highest probability teams advancing 11
Actual model accuracy 68.75%

Distribution of expected advancements
Expected results vs actuals

 

The model continues to do well with the match to match probabilities.

Matches played 54
Cumulative expected model accuracy 59.22%
Actual matches resulting in probable outcome 34
Actual model accuracy 62.96%

Distribution of expected results
Fit of expected results vs actuals
So, for the knockout round so far and the probabilities for upcoming matches which have been determined:

61% Uruguay 2 1 Korea 39%
55% USA 1 2 Ghana 45%
64% Netherlands 2 1 Slovakia 36%
65% Brazil 3 0 Chile 35%
59% Argentina 3 1 Mexico 41%
50% Germany 4 1 England 50%
53% Paraguay     Japan 47%
56% Spain     Portugal 44%
54% Uruguay     Ghana 46%
45% Netherlands     Brazil 55%
52% Argentina     Germany 48%

World Cup probabilities 4.0

Eight teams have advanced. How well did my model predict those teams? Within expectations. Here is the analysis.

Teams advanced 8
Cumulative expected model accuracy 54.72%
Actual highest probability teams advancing 5
Actual model accuracy 62.50%

Distribution of teams advancingFit of expected advance probability vs actual

 

As for the match to match probabilities: 

Matches played 40
Cumulative expected model accuracy 58.82%
Actual matches resulting in probable outcome 25.5
Actual model accuracy 63.75%

Distribution of results
Fit of expected result vs actual result

 

And here is the results table with the probabilites for upcoming fixtures, including Round of 16 matchups that have already been determined.

Match
35% South Africa 1 1 Mexico 65%
46% Uruguay 0 0 France 54%
54% Argentina 1 0 Nigeria 46%
40% Korea 2 0 Greece 60%
53% England 1 1 USA 47%
49% Algeria 0 1 Slovenia 51%
55% Germany 4 0 Australia 45%
53% Serbia 0 1 Ghana 47%
62% Netherlands 2 0 Denmark 38%
42% Japan 1 0 Cameroon 58%
59% Italy 1 1 Paraguay 41%
35% New Zealand 1 1 Slovakia 65%
41% Ivory Coast 0 0 Portugal 59%
85% Brazil 2 1 N Korea 15%
45% Honduras 0 1 Chile 55%
65% Spain 0 1 Switzerland 35%
36% South Africa 0 3 Uruguay 64%
54% France 0 2 Mexico 46%
50% Greece 2 1 Nigeria 50%
63% Argentina 4 1 Korea 37%
48% Slovenia 2 2 USA 52%
58% England 0 0 Algeria 42%
55% Germany 0 1 Serbia 45%
51% Ghana 1 1 Australia 49%
64% Netherlands 1 0 Japan 36%
55% Cameroon 1 2 Denmark 45%
48% Slovakia 0 2 Paraguay 52%
74% Italy 1 1 New Zealand 26%
65% Brazil 3 1 Ivory Coast 35%
82% Portugal 7 0 N Korea 18%
50% Chile 1 0 Switzerland 50%
68% Spain 2 0 Honduras 32%
50% Mexico 0 1 Uruguay 50%
67% France 1 2 South Africa 33%
57% Nigeria 2 2 Korea 43%
45% Greece 0 2 Argentina 55%
46% Slovenia 0 1 England 54%
55% USA 1 0 Algeria 45%
46% Ghana 0 1 Germany 54%
47% Australia 2 1 Serbia 53%
53% Denmark     Japan 47%
39% Cameroon     Netherlands 61%
39% Slovakia     Italy 61%
67% Paraguay     New Zealand 33%
43% Portugal     Brazil 57%
24% N Korea     Ivory Coast 76%
39% Chile     Spain 61%
56% Switzerland     Honduras 44%
61% Uruguay     Korea 39%
55% USA     Ghana 45%
59% Argentina     Mexico 41%
50% Germany     England 50%

World Cup probabilities 3.1

Now that all teams have played two matches, how is the revised model holding up? Pretty well.

Matches played 32
Cummulative expected model accuracy 59.60%
Actual matches resulting in probable outcome 20
Actual model accuracy 62.50%

Distribution of expected results
Expected result vs actual

Here is the full table of group play fixtures, with probabilities and scores.

  Match  
35% South Africa 1 1 Mexico 65%
46% Uruguay 0 0 France 54%
54% Argentina 1 0 Nigeria 46%
40% Korea 2 0 Greece 60%
53% England 1 1 USA 47%
49% Algeria 0 1 Slovenia 51%
55% Germany 4 0 Australia 45%
53% Serbia 0 1 Ghana 47%
62% Netherlands 2 0 Denmark 38%
42% Japan 1 0 Cameroon 58%
59% Italy 1 1 Paraguay 41%
35% New Zealand 1 1 Slovakia 65%
41% Ivory Coast 0 0 Portugal 59%
85% Brazil 2 1 N Korea 15%
45% Honduras 0 1 Chile 55%
65% Spain 0 1 Switzerland 35%
36% South Africa 0 3 Uruguay 64%
54% France 0 2 Mexico 46%
50% Greece 2 1 Nigeria 50%
63% Argentina 4 1 Korea 37%
48% Slovenia 2 2 USA 52%
58% England 0 0 Algeria 42%
55% Germany 0 1 Serbia 45%
51% Ghana 1 1 Australia 49%
64% Netherlands 1 0 Japan 36%
55% Cameroon 1 2 Denmark 45%
48% Slovakia 0 2 Paraguay 52%
74% Italy 1 1 New Zealand 26%
65% Brazil 3 1 Ivory Coast 35%
82% Portugal 7 0 N Korea 18%
50% Chile 1 0 Switzerland 50%
68% Spain 2 0 Honduras 32%
50% Mexico     Uruguay 50%
67% France     South Africa 33%
57% Nigeria     Korea 43%
45% Greece     Argentina 55%
46% Slovenia     England 54%
55% USA     Algeria 45%
46% Ghana     Germany 54%
47% Australia     Serbia 53%
53% Denmark     Japan 47%
39% Cameroon     Netherlands 61%
39% Slovakia     Italy 61%
67% Paraguay     New Zealand 33%
43% Portugal     Brazil 57%
24% N Korea     Ivory Coast 76%
39% Chile     Spain 61%
56% Switzerland     Honduras 44%

World Cup probabilities 3.0

After the first twelve matches completed, I revised my methodology for probabilites. The old method was: 

  1. Start with the FIFA ranking points
  2. Add my modifier based on injuries, regional adder (+50 points for African teams), and my personal assessment of momentum
  3. Calculate a probability factor as 10^(points/400)

The tendancy of this method was to overrate the good teams. I modified step 3 to be closer to what I used in the 2006 World Cup:

  1. Start with the FIFA ranking points
  2. Add my modifier based on injuries, regional adder (+50 points for African teams), and my personal assessment of momentum
  3. Use the points as the initial probability factor

Since this World Cup I am also recalculating the probabilities after each match, I needed to modify how to get the probability factor for each subsequent match. I use Elo ranking to do this. (See Wikipedia for a good article on Elo ranking.) Following the notation in the Wiki article, I use the FIFA points as the team’s initial Q. I back-calculate R. From there I recalculate R after each match then forward-calculate the new Q. Simple, right?

Well, here is how the new method looks diagnostically:

Matches played 23
Cummulative expected model accuracy 58.58%
Actual matches resulting in probable outcome 13
Actual model accuracy 56.52%

Distribution of expected results
Fit of expects results vs actual results

And now the prediction table, with current results added.

Match
South Africa 35% 1 1 65% Mexico
Uruguay 46% 0 0 54% France
Argentina 54% 1 0 46% Nigeria
Korea 40% 2 0 60% Greece
England 53% 1 1 47% USA
Algeria 49% 0 1 51% Slovenia
Germany 55% 4 0 45% Australia
Serbia 53% 0 1 47% Ghana
Netherlands 62% 2 0 38% Denmark
Japan 42% 1 0 58% Cameroon
Italy 59% 1 1 41% Paraguay
New Zealand 35% 1 1 65% Slovakia
Ivory Coast 41% 0 0 59% Portugal
Brazil 85% 2 1 15% N Korea
Honduras 45% 0 1 55% Chile
Spain 65% 0 1 35% Switzerland
South Africa 36% 0 3 64% Uruguay
France 54% 0 2 46% Mexico
Greece 50% 2 1 50% Nigeria
Argentina 63% 4 1 37% Korea
Slovenia 48% 2 2 52% USA
England 58% 0 0 42% Algeria
Germany 55% 0 1 45% Serbia
Ghana 51%     49% Australia
Netherlands 64%     36% Japan
Cameroon 55%     45% Denmark
Slovakia 48%     52% Paraguay
Italy 74%     26% New Zealand
Brazil 65%     35% Ivory Coast
Portugal 82%     18% N Korea
Chile 50%     50% Switzerland
Spain 68%     32% Honduras
Mexico 50%     50% Uruguay
France 67%     33% South Africa
Nigeria 57%     43% Korea
Greece 45%     55% Argentina
Slovenia 46%     54% England
USA 55%     45% Algeria
Ghana 45%     55% Germany
Australia 49%     51% Serbia
Denmark 51%     49% Japan
Cameroon 41%     59% Netherlands
Slovakia 39%     61% Italy
Paraguay 67%     33% New Zealand
Portugal 43%     57% Brazil
N Korea 24%     76% Ivory Coast
Chile 38%     62% Spain
Switzerland 57%     43% Honduras

Anything is possible at the World Cup

With each team having completed their first match, it is a good time to check in on the accuracy of my modelled predictions. As you can see below, the model underperformed. This was driven largely by upsets against the biggest favorites.

Matches played 16
Cummulative expected model accuracy 80.50%
Actual matches resulting in probable outcome 9
Actual model accuracy 56.25%

This chart shows in blue the distribution of possible outcomes assuming the model is correct. The red bar represents the actual model performance. Notice that it is just barely in distribution. Basically, at predicting a winner, so far the model is just a little better than flipping a coin.

Distribution of expected outcomes.

As this model fit shows, the model underperformed primarily at the high end, meaning the biggest favorites ended up being the biggest upsets. Greece, Camaroon, and Spain all lost and Mexico, Italy, Slovakia, and Portugal all tied. It’s like my brother, Daniel, says: “Anything is possible at the World Cup.”

Model fit

Here are updated predictions for the remainder of group play, with the results so far.

Match
South Africa 1 1 Mexico
Uruguay 0 0 France
Argentina 1 0 Nigeria
Korea 2 0 Greece
England 1 1 USA
Algeria 0 1 Slovenia
Germany 4 0 Australia
Serbia 0 1 Ghana
Netherlands 2 0 Denmark
Japan 1 0 Cameroon
Italy 1 1 Paraguay
New Zealand 1 1 Slovakia
Ivory Coast 0 0 Portugal
Brazil 2 1 N Korea
Honduras 0 1 Chile
Spain 0 1 Switzerland
South Africa 9% 91% Uruguay
France 72% 28% Mexico
Greece 49% 51% Nigeria
Argentina 91% 9% Korea
Slovenia 38% 62% USA
England 83% 17% Algeria
Germany 74% 26% Serbia
Ghana 53% 47% Australia
Netherlands 95% 5% Japan
Cameroon 67% 33% Denmark
Slovakia 37% 63% Paraguay
Italy 98% 2% New Zealand
Brazil 98% 2% Ivory Coast
Portugal 100% 0% N Korea
Chile 47% 53% Switzerland
Spain 99% 1% Honduras
Mexico 44% 56% Uruguay
France 95% 5% South Africa
Nigeria 79% 21% Korea
Greece 25% 75% Argentina
Slovenia 26% 74% England
USA 74% 26% Algeria
Ghana 25% 75% Germany
Australia 45% 55% Serbia
Denmark 54% 46% Japan
Cameroon 12% 88% Netherlands
Slovakia 9% 91% Italy
Paraguay 91% 9% New Zealand
Portugal 10% 90% Brazil
N Korea 3% 97% Ivory Coast
Chile 3% 97% Spain
Switzerland 77% 23% Honduras

World Cup probabilities 1.0

Here are my World Cup predictions for the first round. Thanks to Avi for updates on injuries. Updates forthcoming as the tournament progresses.

Group Match
A South Africa 9% 91% Mexico
A Uruguay 30% 70% France
B Argentina 69% 31% Nigeria
B Korea 13% 87% Greece
C England 65% 35% USA
C Algeria 44% 56% Slovenia
D Germany 74% 26% Australia
D Serbia 64% 36% Ghana
E Netherlands 94% 6% Denmark
E Japan 19% 81% Cameroon
F Italy 89% 11% Paraguay
F New Zealand 11% 89% Slovakia
G Ivory Coast 10% 90% Portugal
G Brazil 100% 0% N Korea
H Honduras 29% 71% Chile
H Spain 98% 2% Switzerland
A South Africa 9% 91% Uruguay
A France 70% 30% Mexico
B Greece 54% 46% Nigeria
B Argentina 93% 7% Korea
C Slovenia 35% 65% USA
C England 81% 19% Algeria
D Germany 67% 33% Serbia
D Ghana 45% 55% Australia
E Netherlands 96% 4% Japan
E Cameroon 73% 27% Denmark
F Slovakia 44% 56% Paraguay
F Italy 99% 1% New Zealand
G Brazil 99% 1% Ivory Coast
G Portugal 100% 0% N Korea
H Chile 53% 47% Switzerland
H Spain 99% 1% Honduras
A Mexico 49% 51% Uruguay
A France 96% 4% South Africa
B Nigeria 85% 15% Korea
B Greece 34% 66% Argentina
C Slovenia 22% 78% England
C USA 70% 30% Algeria
D Ghana 22% 78% Germany
D Australia 41% 59% Serbia
E Denmark 62% 38% Japan
E Cameroon 16% 84% Netherlands
F Slovakia 9% 91% Italy
F Paraguay 91% 9% New Zealand
G Portugal 12% 88% Brazil
G N Korea 3% 97% Ivory Coast
H Chile 2% 98% Spain
H Switzerland 68% 32% Honduras