Using statistical evidence to judge defensive performance
Nick Bishop’s recent article on the apparent growing divergence between analysis and statistics was surprising and disappointing. The Leinster analyst – who also worked with Stuart Lancaster for England – uses a single metric to support this, and the article seems to suggest only that there is a growing divergence between analysis and the use of tackle percentage as a viable metric of defensive performance. Indeed, some of Bishop’s own work is testament to how using statistics alongside technical and visual analysis can be persuasive and effective.
Bishop’s piece is a classic example of sporting scepticism of statistical study, and is of a piece with articles by mainstream journalists citing a number to support their case and feeling the need to include the standard disclaimer ‘statistics don’t tell the whole story’. A particular instance of poor statistical analysis does not invalidate the entire field, in the same way that a particular instance of poor tactical analysis does not invalidate that field. If anything, Bishop’s piece serves as an excellent piece of statistical analysis itself, disputing any presumption of correlation between defensive performance and tackle completion percentage by using evidence to support his argument. The next step is to work out what are effective metrics for describing different types of defence, and the 2016 Hurricanes are a good case study.
“As the Hurricanes proved with their defensive revolution in last year’s Super Rugby tournament, line-speed is the contemporary killer, and elite defensive coaches will do almost anything to trigger and sustain high line-speed. As Andy Farrell used to say with England, ‘our defence is not a tidy one.’ Frequently ‘Faz’ would be prepared to accept a higher proportion of missed tackles and line-breaks in order to get what he really wanted – pressure on the ball-carrier and ultimately, turnovers. His defence looked horrible on paper but it worked where it counted, out on the field.”
If line-speed over line integrity is a trade off that many defence coaches are willing to make in modern professional rugby, how can we measure the effectiveness of a team’s line-speed? The objective of a blitz defence such as the Hurricanes’ is to cut off connections to the wide channels by making positive tackles behind the gain line, and often tackling above the waist in order to prevent offloading opportunities – the 2016 Hurricanes in particular liked to couple their midfield blitz with the use of the choke tackle close to the breakdown, and TJ Perenara’s sweeping cover defence is distinctive for the way in which he often targets and strips the ball. What follows from this is that teams attacking against a defence prioritising line-speed will likely – if that defence is effective – attempt fewer passes than usual, complete a lower rate of passes than usual and make a higher rate of handling errors due to the defensive pressure exerted on the ball-carrier.
In a basic version of the method used by football analyst Colin Trainor looking at the effectiveness of a high press, we can look at how often the Hurricanes’ opponents passed the ball for every tackle attempt that they made, and how often they committed handling errors, and compare this to team’s averages across the season. As we can see, in the Hurricanes’ best defensive performances – conceding an average of 7.8 points per game from R16 onwards – there appeared to be a relationship between points conceded and the degree to which they were able to prevent teams passing and offloading:
(In addition, we saw that the Chiefs’ handling error rate was higher in games against the Hurricanes than across the rest of their games, and it would be interesting to see whether this holds up across a season as an indicator of defensive effectiveness.)
5 games is not a big enough sample size to draw meaningful conclusions from this data, but this is simply an example of the way in which we can use basic counts in order to attempt to understand how a team is trying to play, and how effective it is at doing so.
As Bishop correctly states, “the intent and tactical doctrine of the teams involved” need to be understood in order to be able to use statistical evidence confidently and effectively, and that is the case with the Hurricanes’ defence above. However, there is a great deal of tactical understanding in the rugby media, and comparatively few attempts made to use even the basic statistical measures of team and individual play (Charlie Morgan and Murray Kinsella aside) to create a more coherent picture. A lot of simple work can be done with the most basic of counting stats, and there remains a huge amount of untapped potential for statistical analysis in the sport.
Beauden Barrett’s attacking game during the 2016 Super Rugby season was polarising. On the one hand, on phase attack he often stood deep, passed with poor alignment and didn’t offer a threat:
These images from R9 and the semi-final are backed up by passing data, which showed that Barrett created fewer tries and line breaks for his team-mates than other NZ 10s.
On the other hand, the way in which he attacked on set-piece and from the backfield on kick return was devastating:
There were some positive indications that he was starting to attack the line flatter at first receiver in phase attack when the Hurricanes had the ball in the opposition 22:
On the whole, however, there is a disparity between the way in which a fly-half like Aaron Cruden works the gainline in attack in other areas of the field and Barrett’s tendency to stand deeper. For the All Blacks in the Rugby Championship and their outgoing tour, he often showed better alignment and committed more defenders when he came onto the ball at second receiver:
TJ Perenara’s second try in Durban is a good example of why Beauden Barrett could be most effective in attack at 15: pic.twitter.com/2z97aseFBr
— The Chase (@thechasesport) November 1, 2016
Barrett will not be displaced as the Hurricanes’ starting 10 any time soon, and his hold on the All Blacks’ jersey should not be threatened either; however, it is interesting to note that there are areas in which the most exciting player in the game could get better, and improved alignment in the Hurricanes’ phase attack structure could lift the holders’ attack further in 2017.
Nehe Milner-Skudder’s injury prevents him from starting at 15 in week 1, although it is not deemed to be too serious. All indications from head coach Chris Boyd suggest that Ngani Laumape will be given a chance to stake his claim for the inside centre berth left vacant by Willis Halaholo, ahead of Pita Ahki and Vince Aso. In the pack, the biggest loss from 2016 is number 8 Victor Vito: Blade Thomson will start there against the Sunwolves in Tokyo.
The Hurricanes have a quite ridiculous amount of talent in their system: even after losing Vilimoni Koroi and Stephen Perofeta to other franchises, they have blue-chip prospects across the park. Front-rowers Asafo Aumua and Alex Fidow, and second row Isaia Walker-Leawere, all played NPC rugby for Wellington as U19s in 2016, and Brayden Iose will likely emulate them for Manawatu in the upcoming season. St Kent’s scrum-half Carlos Price has come to Wellington from Auckland 1A rugby after starting for NZSS at 9, while Kemara Hauiti-Parapara has been involved with Chris Boyd’s first-team squad in preseason and will likely represent NZ U20 in 2017. In the backline, Billy Proctor (brother of Matt) was also brought in to train with the senior team in January as an U18, Tiaan Falcon did well at the Brisbane Global Tens after getting regular minutes for Hawke’s Bay, and as fully-contracted Super Rugby players Peter Umaga-Jensen (nephew of Tana) and Jordie Barrett need no introduction.