Analysis: Let’s moneyball some rugby
Our analyst Sam Larner returns to Planet Rugby and this week is crunching the numbers in rugby.
If you are a Newcastle Falcons fan you probably didn’t see this last weekend as a fallow one but for many people during this time of year, if it’s not a Six Nations weekend then it’s not really a weekend. With that in mind I decided to do something a little different this week and share some graphs which each highlight something interesting about the game that we love.
I know that some of you will have lost interest since I started talking about graphs, but stick with it, we’ll uncover some groundbreaking new information. Let’s moneyball some rugby!
Defence wins championships?
We all know that defence wins championships don’t we? Watch any American sports and that cliché will have been forced on to you. Clearly when something is repeated that frequently it must be true, mustn’t it?
Now unless you’re a proper nerd you might need this graph explained. Basically the r2 number is an indicator of how correlated two things are. For instance if you plotted length of time ran with calories burnt those two things would correlate perfectly – the longer you run the more calories you will burn. To pre-empt anyone saying ‘correlation doesn’t equal causation’, we can only draw certain conclusions from two things that are correlated.
Many things are correlated which have absolutely nothing to do with one another, for example the amount of money Orlando Bloom makes from his movies and the number of frogs seen in your local pond might look weirdly correlated but obviously they don’t have anything to do with one another. Also, two related things might be correlated but we might believe that the wrong thing is causing the other. Take ice cream sales and temperature, as the temperature increases ice cream sales also increase but with no context we might believe that ice cream sales cause an increase in average daily temperature.
So, if we go back to r2 the simplest way of looking at this is to imagine a scale from 0–1. If the r2 was 0 then we know that two things are not at all related, anything between 0.3–0.5 is a weak relationship between two things, 0.5–0.7 is a moderate relationship, 0.7–0.99 is a strong relationship, and 1 is a perfect relationship between two things.
If we translate that knowledge into the graph above – the data here is from five years and four different leagues – we can see that both scoring tries and preventing tries from being scored correlate strongly with overall success. However, it is actually the scoring points r2 of .98 which is a better indicator of success than conceding points .93. If the old cliché that defence wins championships was to be true we would expect to see those two switched. What we can say is that it actually looks like the best attackers do better than the best defenders. Of course, the very boring real answer is that it’s neither great defence or great offence which wins championships it’s a combination of them both.
A big pool of playing talent wins World Cups
It makes sense that the more players you have to choose from the better your team will be. Think anecdotally about the games you may play on a Saturday, typically clubs which have five or six teams playing for them would usually have the first team in a higher league. If you are scrabbling for 18 you can be less picky about the quality of player you choose but if you have 60 guys to choose you’re unlikely to pick a weak player. It follows that this would also be true at the national level as well doesn’t it?
I’ve plotted this graph by taking all the registered players in that country and the number of ranking points the team have. We can see a very slight correlation between the size of the national pool and the number of ranking points but this would be considered a weak correlation. Bizarrely the size of the talent pool doesn’t really seem to make much difference to overall performance by a nation.
There is a little more to it though and we can dig deeper into the data. Four teams have won the Rugby World Cup and those four all rank in the top five in terms of registered players. The only other team in that top five is France and they have been runners-up three times, more than any other team. That would imply that pool size is significant and I think we can conclude that having a larger pool of registered talent is never a bad thing but there are loads of different factors which determine your overall success as well.
One final thing we can do is look at which nations punch significantly above their weight by looking at ranking points per 1,000 registered players. Surprisingly Uruguay rank top with 11.5, Tonga follow up with 10.6, and Romania are third with 7.9. The highest ranked Six Nations team is Scotland with 1.7, although if I can fan the flames Georgia are fourth with 6.6. At the bottom end of the table it’s South Africa taking home the wooden spoon with 0.16 with England their main challengers with 0.25 and France completing the trio with 0.28.
Leicester Tigers, where did it all go wrong?
Anyone who has even a passing acquaintance with the Premiership will be familiar with the troubles the Tigers are having. Their position in the league has been falling steadily powered by an increasing number of points conceded. What is interesting though is that their ability to score points has essentially stayed consistent – it’s less than two points per game lower than it was at the start of the data collection phase.
As you can see the paucity of their defence has increased virtually in lockstep with their falling league position. Between the 2012/13 season and the 2015/16 season their points conceded per game increased by six and their points per game decreased by just a single point. Between last season and this season their points conceded per game has surged by 6.5 points and their points for has dropped by 1.7 points – this is the first year that the Tigers have had a negative points difference.
The conclusion is pretty simple, the Tigers’ attack is good enough to build around but their defence has hit almost rock bottom. Unless they can find a way to stop conceding the best part of 30 points a game they’re not going to find their way back to the play-off spots.
Attack, attack!
Anecdotally do you think we’re witnessing more or less attacking rugby? Do you think teams are scoring more points than they were, say, five years ago? I don’t know what you think but my own personal view is that we are seeing slightly more points per game than we were a few years ago but it’s not drastic. It turns out that only one part of that assertion is true, the change has been very drastic.
Let me explain what you’re looking at here. This data is from five different leagues across five different years. I took the top six and bottom six clubs for each year and found their average points per game (PPG) and then averaged that out across all the leagues I chose and plotted it here. Now, what I would have expected to see is either two flat lines with minimal fluctuations or two very gradually increasing lines. That’s not at all what we see though, looking at the top six teams between season one and three the PPG had barely moved increasing just .5 in those seasons, however, between season three and season four the PPG leapt up three points and has stayed at around that level into the final season. For the lowest six teams the picture is similar, between season one and four there was just a .8 PPG increase before it leapt up two points in the final season.
I am going to annoy you and say that I really don’t know what caused such a huge increase in the offensive environment. My initial instinct is a rules change but that wouldn’t explain why the top and bottom six clubs didn’t follow the same path at the same time. I also couldn’t think of a rule change significant enough to cause this. Of the five leagues I looked at there’s not a single one which has a lower season four score than season one for their top six clubs, this is a universal change rather than just an anomaly in a single country.
Finally, once you have such a large data set you can then compare different leagues to it. So, unsurprisingly Super Rugby has the best offensive environment of the five leagues I looked at, +3.25, that means you are likely to see 3.25 more PPG, per side, than the average. The PRO14 languishes at the bottom of the leagues I looked at with a score of -2.25. You can also apply this same logic to different leagues, for example the Japanese Top League has a score of +3.6 and the RFU Midlands Premier has a score of +0.9, better than both the PRO14 and Top 14.
Conclusion
We can learn a lot from individual games of rugby but if you really want to understand what is going on in the game as a whole you need to look at larger sets of data. Once you understand that you can start seeing how those trends play out in individual games and seasons.