Business insurer uses stochastic modeling techniques to predict Sunday's scoreline.
Mark Stafford breaks down the betting trends ahead of this weekend's final on NBR Radio and on demand on MyNBR Radio.
What's the story behind the story? Click the NBR Radio box for on-demand special feature audio.
The jape of financial services firms predicting sports events isn't foolproof.
Witness ForBarr giving England the second-best odds to win this year's World Cup.
But still they come.
The latest is QBE. The business insurer's gnomes have used sports data and stochastic modelng techniques to simulate the entire RWC tournament 5000 times for a total of 240,000 virtual games.
Within each match tries, kicks and conversions scored by each team were simulated separately based on their attack and defence history and factoring in the caps of each captain and a home/away advantage.
For Sunday's final, it's predicting:
- New Zealand 26 – 17 Australia
- New Zealand win probability 70%
- Australia win probability 27%
- Draw 2%
And interestingly, its modeling predicts that the higher the game's total score, the more likely it is that the All Blacks will carry the day:
QBE brags that of the 46 World Cup matches played so far, its model has accurately predicted 37 winners with some uncannily accurate score predictions. When South Africa met Wales for example, the QBE Rugby Predictor estimated the score would be 23-20, the actual result was 23-19, just one point out. Equally, when Samoa played USA, the model predicted a 27-16 outcome, the final result was just two points out at 25-16.
But, no, it hasn't been perfect. Like all-comers, it failed to predict the Japan-South Africa upset or England's early departure.
