Traditional rugby statistics (e.g. metres gained, tackles completed, etc.) are commonly referred to within analytics as descriptive/counting statistics. This is because they identify various game events and total them over the course of a game. Recently there has been a trend to identify and count more detailed aspects of the game (e.g. double tackles, LQB, phases with 3+ passes, etc.) but is more of the same better? While these are the performance measures to which most people have grown accustomed, they are somewhat problematic in that they often lack context and don’t always align with results. How many times have you looked at game summaries where one team dominates the statistics but actually loses the game? This happens more often than it should and suggests that perhaps the traditional statistics that are currently used might not be the best predictors of a winning team performance?
One way to add context is through the use of Expected Points (EP). This is nothing new in terms of advanced analytics as all of the other major sports already have some version of this metric. To give you some idea of how it is calculated consider a scrum on the 10m line. If over the course of a season 200 points are scored from 100 scrums the resulting EP would be 2 points per possession. Now expand this calculation out for all possessions and for all locations on the pitch for multiple seasons and you should be able to construct an EP curve (Figure 1). This curve basically represents which team is likely to score next based on factors such as field position, source of possession, and phase of play.
FIGURE 1: Expected Points curve for rugby
It’s important to note that above figure is a generic curve. The shape of the curve along with the EP values along the y-axis will vary depending on the level of competition. For professional rugby EP curves are specific to each competition and the differences reflected in the curves can likely be attributed to differences in strategy and other influential factors such as weather.
One of the conventional measures of team performance in rugby is metres carried. In the advanced analytics world it is considered a flawed measure because not all metres are created equal on the pitch. If you look at the EP curve, you will see that there are portions of the graph that are linear and other areas (e.g. close to either goal line) where the growth is exponential. Thus the metres gained in the red zone are worth more than metres gained around midfield.
Expected Points provides a useful framework that can provide some context to many of the traditional metrics utilized in rugby. Is it perfect? No - but it can provide additional insights that may not have been previously known when analyzing traditional counting statistics. In future posts we will incorporate this framework to demonstrate how EP can be used to evaluate decision making in areas such as team performance, player evaluation, and strategy optimization.
How do you think Expected Points can be to evaluate performance in rugby?