Hello! Welcome back. So what happened out there this weekend? How well did the model predict the outcomes? And where it didn’t, why didn’t it?
Well, below shows how our predictions compared with how the games actually panned out.
As you can see, we got 7 correct results from 9 (78%), however, this included the prediction of a draw between France and South Africa – a bold call – but one that very nearly came true. Do we get a half point for that? We’ll leave it for you to decide.
The average score difference of just under 15 is a little higher than our year to date performance of 11.79, but there were some unusual results in there. If you remove Tier 2 matches, which are historically a little harder to predict, our average score difference was just under 12.
Our model feeds in games from the previous 4 years and works out an expected score for each team based on location and current ranking. Broadly speaking, this means each score represents how well each team is expected to play against the opposition, relative to their performances over the previous 4 years. Any significant difference to the predictive score could therefore indicate over or under performance relative to their previous matches.
France v South Africa was our closest prediction, despite the wrong call on the result. Unfortunately, we didn’t get to see the match, but it sounded like a tight contest with the lead changing hands several times. This victory means South Africa have beaten France 4 times this year.
Italy v Argentina also went broadly according to our pre-match review. Indeed, had the Argentinians not scored a try in the last minute of the game, our prediction of the score difference would’ve been exactly right.
England v Australia was an incredibly tense match that, despite the large scoreboard difference, was a tight contest throughout. England rode their luck and played well to take a comfortable win in the end but, had certain decisions (or bounces) not gone their way the result could’ve ended in Australia’s favour. But for a brace of late tries, our predicted score difference would’ve been extremely close.
Romania v Samoa also turned out to be the close encounter that was predicted with the Oaks emerging narrow victors. Romania are now up to 14th in the World Rugby rankings, just behind Italy.
Wales and Ireland both made hard work of their victories over Georgia and Fiji. Both teams had made widespread changes to their line ups – something that our model cannot account for without manual override (at the moment). However, despite this, the margins of victory were still far narrower than expected. Both Wales and Ireland will need to expand their pools of players should they wish to compete with the strength and depth of England and New Zealand at the World Cup.
Scotland v New Zealand was the match that probably bucked the expectations the most, with the Scots putting in an incredible effort that nearly earned them their first victory over the All Blacks in 112 years. A desperate last minute tackle from Beauden Barrett on Stuart Hogg deep in the All Black 22 secured the victory for the visitors in the end, but it was a thrilling match throughout.
Japan thrashed Tonga by a considerable margin in the end, a much bigger win than we expected, adding further credibility to their rise as a rugby country to be reckoned with in the run up to their World Cup in 2019.
Uruguay beating Nambia was the only result that was particularly unexpected out of the weekend’s results. Some investigation is required to see whether this reveals a quirk in our model that needs to be corrected for the lower tiered match ups. Regardless of the predictive element, this continues an impressive run for the Uruguayans, who have lost only once this year.
The predictions for week 3 will be up soon, please follow/share/like on all the usual social media to make sure you don’t miss out.