Super Rugby Stats Review - Attack - Green and Gold Rugby
Rugby

Super Rugby Stats Review – Attack

Super Rugby Stats Review – Attack

The following article has been written by Eloise Pasteur. Eloise is a former, bad player of rugby while at school (a long time ago) who watches avidly and has a day job teaching science, maths and stats to people of all levels of ability. Occasionally she lets her love of rugby and her love of stats collide.


This article is going to look at the attack stats of the Super Rugby season just past.

If you’ve read my previous article about defence stats, then the following paragraphs aren’t of much use to you. You can just skip to first heading below.

I’ve used simple correlations to try and analyse what stats tell us interesting thing about what is going on. Remember that correlation does NOT imply causation. For each statistic, the blue graph is the whole competition, the gold graph is just the Australian conference, the black is just the New Zealand conference and the green just the two South African conferences combined.

So, let’s attack these attack stats!

More Metres Means More On The Meter

Figure 4Overall there is a pretty strong correlation between points for and metres gained.

Plotting the data this way round makes it hard to talk about what’s going on in terms that are easy to talk about. Reversing the data and plotting that gives the following interpretation: for every 100 metres gained an Australian side typically scores 3 points, and the sides from the other two conferences score 6 points.

NOTE – this is not causation. You can’t say “we made 500 metres so we should have 15 points” but metres gained over a season is probably a not terrible estimate of overall offensive activity. In a particular game you might score a lot of points off short-metre gains from turnovers on their try line and interceptions in their half and a strong defensive effort, but over the season this is a decent measure of your overall offensive effort.

Wallabies v All Blacks 19 Aug 2017 Bledisloe 1 first half

Make those metres

 

The Kiwis and the SA sides are at least twice as efficient at converting this measure of offensive effort into points as the Australians. This is one to keep an eye on over The Rugby Championship, where there will be 6 games. Just divide the points scored by the total metres gained and see how the various sides are doing.

Saying “metres gained is a decent indicator of offensive activity” is true but some types of defence need to be considered. Not all defensive systems are as aggressive as the British & Irish Lions was this July, with some willing to give up metres but not points such as the ‘bend, not break’ system. And then, of course, every team’s defence tightens up close to their tryline.

Why are Australian sides so lacklustre in attack? OK, they played five games against the Kiwis and came up against what the eyeballs would suggest are five of the top few defences in the competition which doesn’t help. But they also played each other twice, and so played four or six games against some of the worst defences in the competition too. It ought to balance out somewhat! It looks like something really needs to be done here.

Just Beat It (Defenders, That Is)

Figure 5

Finally a nice, non-contentious one. Beating defenders better seems to lead to scoring points. The Kiwis do it best, the South African conferences do it the worst, the Aussies are pretty much on the mean.

Calculating the numbers the simpler way round, you score 1.3 points per defender beaten, or, roughly, you score a converted try per five defenders beaten.

This might seem low but, again, bear in mind that although we remember the spectacular tries from deep, if you beat a defender inside your own 22 it often doesn’t lead to points, whereas beating a defender in their 22 is much more likely to lead to a score. Also, beating a defender won’t always lead to a try, points can be scored after you beat a defender and they’re offside at the next ruck or something and you kick the penalty. These statistics are only really a superficial look at what’s going on.

It’s hard to say much more than that. I don’t think you can coach “beat the defenders” nor “don’t get beaten” really but these seem like important things on attack and defence respectively. But this isn’t a new lesson.

It is worth remembering that low rate of converting defenders beaten into actual tries though.

Now for one that really surprised me.

Offloads Don’t Help

Figure 6

There is no good correlation between offloads and scoring points. For the SA conference there’s absolutely no effect, that line is as flat as a pancake.

I don’t know if this because the data just collects “that was an offload” and doesn’t record what happens – success, failure, knock-on, interception etc. or what. The eyeball test suggests successful offloads have a big impact, but a knock-on gives up possession and an interception might well give up 5 or 7 points to the opposition. The Kiwis apparently do it better, but it’s hard to be sure with the small data set so we can’t really trust that.

This is information that really needs to be cleaned up to be of use to the casual fan. Successful offloads (going to a team-mate that catches the ball and is tackled quickly and maintains possession, or runs, kicks or passes the ball) would be a good starting point. Proportion of successful offloads would also be good.

Kicking – What Is It Good For?

Figure 7

Finally kicks in play. There is nothing we can be very sure about here and the comments I’m about to make are rather tentative because of that.

It appears that there’s not much correlation between kicks in play and points scored that we can trust but any there is is slightly positive – if you kick more, you score slightly more points, maybe. Thinking about it, that shouldn’t be a surprise, quite a lot of kicks in play are to relieve pressure, when you’re in your 22 for example, and you give up possession so you don’t expect to score points.

Those slopes aren’t really exaggerated, so it’s noticeable to the naked eye that the SA conference is pretty much flat… kicks in play don’t relate much to scoring points. For the NZ conference there’s a fairly strong up slope – kicks in play do relate to scoring points. For the Australian conference there is a weak slope – that suggests kicks in play seem to COST your team the chance of scoring points.

Duncan Paia'aua gets a clearance kick away

Duncan Paia’aua gets a clearance kick away

Applying the eyeball test Beauden Barrett, Lima Sopoaga, Aaron Cruden and, to a lesser extent, Richie Moanga all kick for some of their players in an attacking way (I haven’t seen enough Blues games but I wouldn’t be surprised if it happened for them too). It’s not 100% successful but it is certainly a dangerous part of their attacking arsenal. Defensive kicks are too from most of their players and various full backs. But how often do you see a New Zealand player make an aimless kick? It does happen, as do poor kicks and kicks out on the full, but it’s a scatter of them for each player across the season.

The impression is that, although there are these attacking kicks from Australian 10’s there are a lot of poorer kicks, kicks to players, kicks out on the full, kicks that invite counterattacks and the like. This stat would tend, maybe, to support that impression.

Conclusion

What could the Wallabies, and the Super Rugby teams, learn? Or what can we, the fans learn and try to look out for?

Australian teams don’t very good job of gaining meters (although that could be down to good Kiwi defences), nor do they do a good job of kicking the ball. As we all know, we’ve seen the Kiwis make mistakes with the boot, but the fact is that they don’t make many. But those Kiwis make more kicks, and also don’t just let their flyhalves kick – their wingers kick too! How often do we see the Wallabies make bad kicks?

I briefly played with the data from the 8 sides in the finals against the rest – what seem to be the characteristics of the successful sides compared to the unsuccessful ones seems like a reasonable learning point.

There are some common things that come through: their metres gained is a better indicator of points scored (and they’re more efficient in the points per metre stat). However, given there are 4 Kiwi sides and 3 South African sides but only one Australian side, while ‘the others’ are nearly half Australian, how much does this just reflect the fact that this split could be down to the national styles of play rather than the successful teams? The better linking between tackle efficiency and points against is not there in the data – that seems to be emphasised more here, but the NZ and SA are more efficient at the metres gained and scoring points linkage and dominate that best eight and probably pull the graphs that way.

The clubs and international coaches have far more data and there is far more analysis that can be done. But it seems some of the data we’re shown in official statistics is also a poor summary of the match so it’s worth being wary what we look at and consider important.

You can find Eloise’s previous article over here.

  • MST

    Eloise thank you. We really appreciate you taking the time to put the articles together. Both of your articles have been fantastic and the information has been really interesting. I was really interested by the attack metres made statistics and correlation to points. It would be interesting to track the progress of teams (especially the Aussie teams) next year. Maybe we can get you to help us with some stats for the Top 5 next year!

    • Sure, happy to help. And compiling and organising the stats was fun.

  • mikado

    Thanks once again for that Eloise. The lack of correlation between offloads and points scored is interesting. There’s absolutely no correlation at all between offloads and turnovers conceded, so it’s not as if high-offloads teams are consistently coughing up the ball to their opponents. Perhaps it’s simply that a lot of offloads are low-value. Often a player offload the ball to a team-mate who is static and just about to be tackled himself.

    Apart from that, I noted that there’s virtually no correlation between points scored and set piece success (lineout% R2=0.16, scrum% R2=0.10). So teams with weaker set piece are clearly compensating in other ways (but not to the extent that teams who have a strong set piece are missing out). There’s also no correlation between success at kicks for points and points scored (R2=0.03) so, again, teams who pick weak kickers are clearly compensating in other ways.

    As you’d expect, there’s a fairly strong correlation between points scored and clean breaks (R2=0.80) and a weaker correlation with carries (R2=0.64).

    • I was surprised about the offloads – but I don’t collect the data so I’m still wondering really about the quality of the data there.

      I think with set pieces, they take place all over the field, so I’m perhaps not that surprised there’s a relatively low correlation. Also, does your data include not-completed (ends with a scrum-reset, free-kick, penalty etc.). I’m not sure what the percentages for successfully completed scrums and lineouts are, but certainly less than 100% (I’d guess less than 70% for scrums, higher for lineouts) which will affect that correlation too. That might be disguising the impact of the stronger set piece teams. Alternatively, the teams with a stronger set piece tend to score that way, while others score more from open play, the driving maul or whatever not being an option. So it balances out.

      The success at kicks also doesn’t surprise me. I’m not quite sure what the range is but I’d be surprised if any team was much under 65% or any team was much over 80%. So you’ve got quite a narrow range for the kick success. And you can rack up a lot of points by scoring tries and missing your kicks, so it’s not the whole story. (in something like AFL or soccer/football kicks on goal to scoring is a more useful measure because it’s your only scoring method).

      Clean breaks and carries are both fairly decent measures of “offensive activity” as well. Clean breaks obviously make for more scoring opportunities, both by eyeballs and stats from what you’re reporting. Carries, well we’ve all seen teams carry 20 times, go sideways (or backwards) and kick the ball away, as well as carry once and run the length of the pitch and score. So a weaker correlation is to be expected.

      I think with all these things, it’s a choice of what to look at. Metres carried per point is easy, and asking why it’s so different for the different conferences raises some interesting questions. Asking why offloads don’t work also asking some interesting questions… but it’s more about what’s being collected I think (we see offloads work, and fail so there is something to them, clearly).

      • mikado

        The set piece %s are just from the Saanzar super rugby site. I like you am suspicious as to what exactly they mean. Does a scrum that ends in a dubious penalty award count as “won” or “lost”, or is the stat just for cases where the ball remains in play. The stat doesn’t reflect set piece “dominance” of course, in terms of how a team was able to use the ball they’d won. Also the set piece success %s are across a narrow range (scrum 81.5% [Cheetahs] to 91.1% [Chiefs] and lineout 83.4% [Rebels] to 92.2% [Crusaders]) so there’s a greater chance perhaps of random variation removing the correlation. However given the way that some teams place high priority on strong piece whilst others neglect set piece in order to strengthen their loose play, it’s interesting (and encouraging) that either approach can work.

        The kick % success varied from 63.5% [Reds] to 87.2% [Cheetahs]. Obviously the % success is not just due to the skill of the kicker, it’s also down to what options teams take with penalties. And possibly also whether teams tend to score tries out wide (I don’t know if that’s thing, but if it is it’d make conversions harder). Again, given the way that some teams place high priority on a good place kicker whilst others neglect place kicking in order to strengthen their try scoring, it’s encouraging that either approach can work.

        • I’m surprised the top kick rate is quite that high… I wonder if they have a brilliant kicker or only take a smaller percentage of kicks and mostly score tries under the posts?

          I think, from what I watch rather than analysing the data, Super Rugby tends to score more points from tries that some of the NH competitions where the weather and the pitches can be worse for long stretches. There might be more value for a kicker (even with the relatively narrow range) in the Aviva, Pro 14, Top 14.

          I think the lineout stat that really matters, in terms of point scoring is penalties taken as lineouts converted into tries/converted tries/penalty tries, and compare that to the penalty likely outcome. Lineout steal % might also be interesting. Everyone has a decent lineout… 83.4% isn’t terrible. But even if you win, if you play on and lose the ball after a knock-on say, or a bad maul, then you’ve lost the points. If the penalty was on the 15m line, what chance did your kicker have of scoring? Did you do better, over the season, by opting for the lineout or do you do better opting for the lineout? It’s very nerdy, but that’s an interesting one (that we can’t really tell from the data we get as fans).

    • Graeme

      Why use R2 instead of just R. For instance R2 of 0.16 means there is a 0.4 correlation between lineout% and scores, which is actually pretty high. With only 18 teams it is probably too few to have statistical significance, but would still seem to suggest lineout% is an important factor.

      Also R2 of 0.1 for scrums% equates to a correlation of over 0.3, again fairly high.

      • mikado

        Yes could have used R, I guess. As you say, the correlation wold have to be very good to be statistically significant.

      • Using r^2 gets rid of the sign (you can have positive or negative correlations, but that value can be determined from m, the slope of the best fit line), so you can determine if the correlation is significant or not simply from whether r^2 is greater than a given value. That value varies, depending on n, the number of data in the set you’re running the correlation on. (Pendantically it’s on the dof in the dataset, but you can usually find it presented as n as well.)

        “Fairly high” doesn’t actually matter – it’s more a matter of is it statistically significant or not.

  • MalachyBernard

    Thanks for the analysis Eloise. One stat that would be fascinating to see is “average metres gained from clearance kicks from own 22″. For me the obvious difference between NZ teams and the rest is the ability to keep the other side pinned in their own half. If a clearance is shallow or does not make touch, more often than not there is a successful counterattack by NZ within a few phases. Kicks that make touch, go deep and cannot be thrown quickly by the opposition are the ideal.

    • Agreed, and that’s probably the sort of stat the teams have that’s just not provided easily to the fans I’m afraid.

    • Thinking a bit more, net gain on kicks from your own 22 is the stat I would like to see. You can obviously kick it straight out, so that’s simple, but if it stays in field, how far do you gain before the next breakdown or stoppage, whatever it is? Perhaps that needs to be in two parts, the kick and the return.

      We all think that Australian teams kick badly from hand and NZ teams run back brilliantly. This would let us test that hypothesis.

  • Sorry for necroposting on my own post… but the metres per point stat I said was a good measure of offensive effort. I haven’t done the whole competition, but this weekend we saw Australia and Argentina in what we might call a game of two halves, and if you look at that stat, both sides were getting around 7 points per 100 m, Australia a bit more at 7.5 and Argentina a bit under at 6.4. But quite close, and both around the average the NZ and SA conferences were getting in the Super Rugby.

    New Zealand finally put it together for the full 80 minutes and their scoring rate sneaked up to 10.1 points per 100m gained. Even in their “bad” matches they have been around the 7/100m mark and in Bledisloe 1 they were at 9.7/100m. They held the Wallabies to under 7 (under 6 in Bledisloe 1). We all know, from watching, their attitude seems to be “we can score points whenever we get the ball, if you want to beat us, you have to score heaps of points against us” but these stats illustrate just how efficient they are at scoring, and how efficient they are at stopping the other side from scoring too.

Rugby

Hopes to play David Pocock in the inevitable biopic. Lifelong fan of whoever Jarrad Hayne is currently playing for.

More in Rugby

  • Monday’s Rugby News

    Monday’s Rugby News looks at the Wallabies major task of securing the Tri-Nations, the future of the...

  • Friday’s Rugby News

    Friday’s Rugby News sees New Zealand Trying To Make Amends,  Argentina Prepares , Europe Wrap, and Caslick is All In...

  • Thursday’s Rugby News

    Thursday’s Rugby news looks at the All Blacks on the war path, as the Wallabies look to...

  • Wednesday’s Rugby News

    Wednesday’s Rugby News looks at the Wallabies changing up their processes, Will Genia’s big idea to revamp...