It’s the quarter finals tomorrow and it’s time for me to predict the outcomes using World Rugby’s rankings. Although I got 33 out of 37 matches right in the pool stages, the 4 errors are enough to change my prediction of who will win.
As of today, World’s Rugby Rankings for the 20 teams taking part in the 2019 World Cup are as listed in order here. Since Japan is the host nation, 3
points have been added to their ranking points which puts them level with France rather than Argentina. It should be noted that Namibia are in fact in 23rd place in the full rankings but 3 teams ahead of them (Spain, Romania & Portugal) failed to qualify.
The biggest upsets in the pool stages came with Japan beating Ireland in Pool A and Uruguay beating Fiji in Pool D. As a result, both teams improved their ranking points considerably with Ireland & USA suffering the largest fall. It’s notable that some teams are unchanged, most notably England. This is because the 3 teams they played prior to the cancelled match against France were all at least 10 points lower in the rankings. When the ranking gap exceeds 10 points, nothing changes if the stronger team wins. Had England been able to play and beat France then they would have gained ranking points.
In my first post on the 2019 World Cup, I explained the two models I am using to predict matches, HighRank & ExpWin. HighRank says that the winner will be the team with the highest ranking points whilst ExpWin predicts the margin of victory for the stronger team based on the ranking points gap. One realisation I had with the ExpWin model is that it could be used to measure how well each team did in the pool stages by comparing the actual margin of victory with the expected 
margin of victory. For example Wales were expected to win all 4 of their pool matches by respectively 23 (Georgia), 2 (Australia), 21 (Fiji) & 36 points (Uruguay). In the event they underperformed slightly and the actual margin of victories were 29, 4, 12, 22, an average underperformance per match of 4 points which is what is shown in the last column of the ranking tables here.
Indeed the table to the right resorts the 20 nations in terms of the difference between actual and expected winning margins. Of course we need to remember that New Zealand, Italy, England, France, Canada & Namibia were only able to play 3 matches due to Typhoon Hagibis. But this does show that South Africa, New Zealand, England, Japan & Australia were outperforming expectations whilst Wales, Ireland & France underperformed. I will come back to these figures later on.
How well have my models performed?
The short answer is that they have performed mostly inline with expectations and therefore I do not propose to change my model for the knockout phase. A more in-depth evaluation can be found here.
The Knock Out Rounds
A reminder that I use dynamic ranking points i.e. the ranking points are modified for the semi finals and finals on the assumption that my quarter final predictions are correct. For each match I have given the probability of winning for the higher ranked team, the expected margin of victory for the higher ranked team and a + column that I will explain later.
It is worth comparing these with what I was predicting before.
Previously I had Ireland sneaking past New Zealand in the final but Japan’s stunning victory over Ireland upset that big time. With Japan topping pool A, the knockouts look very different.
Most notably, Japan & South Africa are neck and neck in the quarter final. My model makes Japan marginal favourite which is quite remarkable but let’s not forget Brighton 4 years ago. I think some people are forgetting that you need to add 3 points to Japan’s ranking due to them playing at home.
The Semis are expected to be close games and lead to a Wales v New Zealand final where I have the All Blacks winning their 3rd successive title by 3 points.
However, an alternative forecast that I could have used would have been to add and subtract the over/underperformance I mentioned earlier in the ranking tables. Take the Japan V South Africa quarterfinal where I basically have the margin of victory at zero i.e. a draw. South Africa have overperformed by 11 points whilst Japan have overperformed by 5 points. The difference is 6 points which could be added to the expected margin of victory and predicts a South Africa win by 6 points instead. This is what the + column in the predictions above show i.e. you can add this column to the expected margin victory to get an alternative forecast.
All will be revealed come Sunday afternoon and next week I will update my forecast for the semi-finals.
