October 14, 2019

Predictions for the RWC Quarter-Finals

Team Ratings for the RWC Quarter-Finals

The basic method is described on my Department home page.
Here are the team ratings, along with the ratings at the start of the cup.

	Current Rating	Rating at Season Start	Difference
New Zealand	27.22	27.60	-0.40
England	20.46	19.78	0.70
South Africa	20.22	18.51	1.70
Ireland	12.08	12.12	-0.00
Wales	9.97	12.72	-2.70
Australia	7.60	8.70	-1.10
Scotland	6.14	5.44	0.70
France	4.89	7.11	-2.20
Argentina	3.32	4.15	-0.80
Japan	-5.15	-7.67	2.50
Fiji	-5.54	-5.27	-0.30
Italy	-9.53	-8.57	-1.00
Samoa	-15.53	-11.89	-3.60
Tonga	-15.83	-20.46	4.60
Georgia	-18.08	-17.43	-0.70
USA	-18.64	-16.38	-2.30
Uruguay	-30.89	-35.67	4.80
Canada	-34.63	-32.45	-2.20
Russia	-36.32	-36.78	0.50
Namibia	-39.99	-41.80	1.80

Performance So Far

So far there have been 37 matches played, 34 of which were correctly predicted, a success rate of 91.9%.
Here are the predictions for previous games.

	Game	Date	Score	Prediction	Correct
1	Japan vs. Russia	Sep 20	30 – 10	35.60	TRUE
2	Australia vs. Fiji	Sep 21	39 – 21	14.00	TRUE
3	France vs. Argentina	Sep 21	23 – 21	3.00	TRUE
4	New Zealand vs. South Africa	Sep 21	23 – 13	9.10	TRUE
5	Italy vs. Namibia	Sep 22	47 – 22	33.20	TRUE
6	Ireland vs. Scotland	Sep 22	27 – 3	6.70	TRUE
7	England vs. Tonga	Sep 22	35 – 3	40.20	TRUE
8	Wales vs. Georgia	Sep 23	43 – 14	30.10	TRUE
9	Russia vs. Samoa	Sep 24	9 – 34	-23.50	TRUE
10	Fiji vs. Uruguay	Sep 25	27 – 30	30.00	FALSE
11	Italy vs. Canada	Sep 26	48 – 7	23.10	TRUE
12	England vs. USA	Sep 26	45 – 7	35.40	TRUE
13	Argentina vs. Tonga	Sep 28	28 – 12	24.00	TRUE
14	Japan vs. Ireland	Sep 28	19 – 12	-16.30	FALSE
15	South Africa vs. Namibia	Sep 28	57 – 3	59.50	TRUE
16	Georgia vs. Uruguay	Sep 29	33 – 7	15.40	TRUE
17	Australia vs. Wales	Sep 29	25 – 29	-3.60	TRUE
18	Scotland vs. Samoa	Sep 30	34 – 0	15.60	TRUE
19	France vs. USA	Oct 02	33 – 9	23.60	TRUE
20	New Zealand vs. Canada	Oct 02	63 – 0	61.70	TRUE
21	Georgia vs. Fiji	Oct 03	10 – 45	-7.80	TRUE
22	Ireland vs. Russia	Oct 03	35 – 0	47.10	TRUE
23	South Africa vs. Italy	Oct 04	49 – 3	25.60	TRUE
24	Australia vs. Uruguay	Oct 05	45 – 10	42.70	TRUE
25	England vs. Argentina	Oct 05	39 – 10	15.80	TRUE
26	Japan vs. Samoa	Oct 05	38 – 19	12.90	TRUE
27	New Zealand vs. Namibia	Oct 06	71 – 9	68.40	TRUE
28	France vs. Tonga	Oct 06	23 – 21	26.10	TRUE
29	South Africa vs. Canada	Oct 08	66 – 7	53.90	TRUE
30	Argentina vs. USA	Oct 09	47 – 17	19.00	TRUE
31	Scotland vs. Russia	Oct 09	61 – 0	40.00	TRUE
32	Wales vs. Fiji	Oct 09	29 – 17	18.80	TRUE
33	Australia vs. Georgia	Oct 11	27 – 8	27.20	TRUE
34	Ireland vs. Samoa	Oct 12	47 – 5	24.40	TRUE
35	USA vs. Tonga	Oct 13	19 – 31	-0.80	TRUE
36	Wales vs. Uruguay	Oct 13	35 – 13	45.00	TRUE
37	Japan vs. Scotland	Oct 13	28 – 21	-7.40	FALSE

Predictions for the RWC Quarter-Finals

Here are the predictions for the RWC Quarter-Finals. The prediction is my estimated expected points difference with a positive margin being a win to the home team, and a negative margin a win to the away team.

	Game	Date	Winner	Prediction
1	England vs. Australia	Oct 19	England	12.90
2	New Zealand vs. Ireland	Oct 19	New Zealand	15.10
3	Wales vs. France	Oct 20	Wales	5.10
4	Japan vs. South Africa	Oct 20	South Africa	-18.90

David Scott obtained a BA and PhD from the Australian National University and then commenced his university teaching career at La Trobe University in 1972. He has taught at La Trobe University, the University of Sheffield, Bond University and Colorado State University, joining the University of Auckland, based at Tamaki Campus, in mid-1995. He has been Head of Department at La Trobe University, Acting Dean and Associate Dean (Academic) at Bond University, and Associate Director of the Centre for Quality Management and Data Analysis at Bond University with responsibility for Short Courses. He was Head of the Department of Statistics in 2000, and is a past President of the New Zealand Statistical Assocation. See all posts by David Scott »

Comments

Abhay maheshwari

This prediction will be remain same till the qtr final match start i.e on 19th or might be change in later days.

please confirm

5 years ago
- David Scott
  
  These predictions are for the quarter finals and won’t be updated because there are no matches in between to affect the ratings.
  
  5 years ago
DL Williams

Have you considered modifying your ratings based on actual team selection. This must have a major influence in one-off games

5 years ago
- David Scott
  
  Not something I would do. These predictions are based on past data, the only subjective components are choice of parameters and sometimes decisions on is there a home ground advantage.
  
  It is possible to do something statistical regarding who is in the team, for example using percentages of games won where a particular player is playing. There are a few problems with that though. First of all the data and analysis required which far surpasses what I use. Gathering data and assimilating it is a big task which I largely automate. The betting companies can do that sort of thing because they have the resources. There are problems with scarce data though: how do you assess George Bridge who has played very few games compared to Ben Smith who has a lot of past data, but may be in decline? You risk adding a lot of variability to your forecasts using poorly estimated parameters.
  
  Incorporating a lot of additional past data increases the risk of overfitting which is the enemy of good forecasts.
  
  My line is that these forecasts are sound giving good performance for the data used. If you want to consider other factors, those are subjective choices which you can use to modify your personal predictions.
  
  5 years ago
Toby Dean

What makes you think Wales will beat France?

5 years ago
- David Scott
  
  Statistics
  
  5 years ago
  - Steve McMillan
    
    Great answer :-)
    
    5 years ago
Christopher Williamson

So… I’m not too sure what the purpose of this is as your predictions are only valid for picking the winning team, not including the margin.

So there is a success rate of 91.9% of predicting the winning team, however that drops substantially when using the winning team and margin (14/37 within a 5 point margin) then its only around 38% success.

Instead one could just look at the international work rankings and make a guestimate and have almost the same success rate at picking the winner.

5 years ago
- David Scott
  
  Yes, it is relatively easy to predict the winner in most of the pool games with high accuracy because of the huge difference in the quality of the teams.
  
  You miss the point a little concerning margins however. The margin given is a proxy for the probability of a correct prediction of the result. Also it is the expected margin only and there is variability about that expected value, the distribution being heavy tailed as I have stated previously. I defy anyone to predict the margin within 5 points with any degree of accuracy. That is only an unconverted try. You only have to think about how many times in an 80 minute game a small mistake or bit of brilliance can result in a try to one side or the other to realise that rugby games are far too random for reliable predictions with that sort of precision.
  
  5 years ago
  - Steve Oram
    
    David, I’ll give you one data point regarding margin prediction. At work, we have a tipping competition underway for the Rugby World Cup. We use the Superbru website / app so can also compare our picks with others worldwide. Superbru assigns points for getting the result correct and then bonus points for predicting the margin within 5 points. I have used your RWC predictions for result and margin exclusively and currently sit 4350th out of 332000 participants worldwide, or top 1.3% worldwide. More importantly, I currently sit 2nd in the work pool and not far off the $200 dinner voucher first prize. Thanks.
    
    5 years ago
Louis Quatorze

Are these numbers calculated based on NZ home games played in NZ? Since its on neutral ground and NZ has only played Ireland twice on neutral ground (South Africa, USA) would this have an affect on it and negate the home advantage?

5 years ago
- David Scott
  
  The home ground advantage is a single value which gives the best predictions when all games between teams in the RWC from 1996 to the present are considered. The data set used includes many games played on neutral grounds, not limited to previous world cups. There are other regional cups where there are many games without home ground advantage. I don’t have a different value for games between different teams—that would mean many very poorly estimated parameters, as your example indicates.
  
  5 years ago
Eggy Smith

In all your various match analyses have you thought about predicting scorelines rather than a straight superiority?

5 years ago
- David Scott
  
  No. Would require different methodology, and would likely be quite inaccurate. Differences are nice to work with statistically since they are roughly symmetric and can also be treated as continuous given the range of values and the number of values. I am not sure what the distribution of individual scores would be. Work I have seen on football scores treated team scores as Poisson, but rugby scores are rather different.
  
  5 years ago
Farrell Nathaniel

Hi

What does the negative denote? Eg SA vs Japan (-18.90)

5 years ago
- David Scott
  
  It’s actually Japan v SA, the first team named being the home team and having home ground advantage, if any. Negative value means the away team or second named team is predicted to win by the given number of points.
  
  So here, SA to win by just under 19 points, so statistically at least, a hot favourite.
  
  5 years ago
Rav gray

Thanks David, pretty nice analysis.
been following your predictions from round 1 and was wondering whether any of the matches in elimination round would go into “extra-time”.

-Rav

5 years ago
Parsant Pillay

Hi David

Do you think any Quarter Final Matches will go into extra time. What are your thoughts on that ??

5 years ago
- David Scott
  
  Draws are very rare in rugby so the chances of going into extra time are quite low. I have seen other predictions where the probability of a draw is typically 1% or less. I never predict a draw in my predictions.
  
  5 years ago
Mitchell Palmer

Hi David,
I really enjoy these predictions. Thank you for them! Have you ever posted the source code (in R, I presume) for the methodology you use?

5 years ago
- David Scott
  
  No code sorry. Has taken a lot of development and I am not willing to give it away (at least at this time).
  
  5 years ago
Jan Coetzee

Hi David,

Well done on your predictions so far! I am really enjoying what you are doing. Any indication when we can expect to see your semi final predictions?! While I am following SA, I am more interested in the NZ vs ENG one!

5 years ago
Henry Ferreira

What are the stats for the semi’s?

5 years ago