Spectators observe it. Experts talk about it. But how can we actually quantify it?
It is common in sport for TV reporters to pluck statistics from a large dataset to support their conclusions of how they perceived match tactics. A more sophisticated (and more accurate) approach is to assess the data in its entirety, and to search for similarities and trends.
This was indeed the objective of Carl Woods and colleagues. They applied a novel statistical technique to assess for trends in match day data between 2005 and 2016 in the NRL. The statistical technique was called non-metric multidimensional scaling (NOTE: this statistical approach is not novel to science, but it is novel to sport science).
Without going into too much detail about non-metric multidimensional scaling, it essentially identifies (dis)similarities between teams and across seasons. Hence, the authors were able to illustrate which teams displayed similar tactics, and which seasons produced a change in tactics.
There were at least 2 key observations.
There was an abrupt change in tactics in 2012
The authors stated: “Compared to the 2005 to 2011 seasons, the 2012 to 2016 seasons are in a state of flux”.
Between 2005 to 2011, all NRL teams displayed similar team profiles in an ordinal plot (see Figure below). This plot clusters profiles of greatest similarity. Hence, it seems that teams were adopting similar tactics (or, at the very least, similar match day profiles) between 2005 and 2011.
From 2012 onwards, however, there was a clear change in team profiles. To highlight this, the team profiles of the two grand finalists were similar before 2012, but since then the two grand finalists have displayed dissimilar profiles. This might represent the development of innovative and contrasting game plans.
Why was there an abrupt change in 2012?
Well, as the authors state: “The end of the 2011 season led to the largest coach turnover in NRL’s history. Specifically, eight of 16 head coaches either changed clubs or ceased to coach (in the head coaching role or at all).”
The authors made another interesting observation: “The three teams that displayed the greatest observable season-to-season dissimilarity (the Eels, Roosters and Sharks) have had the greatest head coach turnover since the inception of the NRL”.
Hence, the introduction of new coaches seemingly leads to observable changes in team profiles.
Certain teams follow similar trends
Notable observations include:
Storm, Broncos, West Tigers and Cowboys = these teams showed similar profiles. Even during 2012, they all displayed a large change but in a similar direction.
Bulldogs, Rabbitohs and Dragons = these teams showed similar profile paths from 2005 to 2016.
Eels, Roosters, & Sharks = these teams displayed the most dynamic profiles relative to the competition.
Since 2012, NRL teams have displayed constantly evolving profiles. Even in the past 2 years, there appears to be more change. The authors noted that the 2016 season saw changes to the rules. The number of players on the interchange changed from 10 to 8, and a “shot clock” for scrums and drop outs was introduced. These changes were implemented to minimise defensive play. Certainly in the 2016 season there appeared to be an increase in attacking play as evidenced by a rise in the number of off-loads (i.e., passes).
The implication for coaches is clear. Game plans need to continually evolve.
From a sports science perspective, the use of non-metric multidimensional scaling might be a useful method for assessing how similar or dissimilar a teams profile is, which might assist with developing game plans against an opposition team.