The breakdown continues to be a topic where there is a great deal of debate around how teams should approach this aspect of the game tactically on both sides of the ball. The prevailing thought is that the foundation of an effective attack in phase play is to generate lightning quick ball (LQB) at the breakdown. The rationale is that if the attacking team can recycle the ball from the breakdown before the defensive line can set, there will be more time and space to attack against a defence that is “on the hindfoot”.
The current conventions for ruck speeds are typically defined as:
LQB = ruck over times less than 3.0 seconds
Regular = ruck over times between 3.0 to 6.0 seconds
Slow = ruck over times longer than 6.0 seconds
NOTE: The “ruck over” times are measured from the moment when the ball carrier is on the ground to when the ball is played.
Since many analytical insights have arisen as a result of challenging conventional thinking, I have always wondered why the threshold for LQB is defined to be 3.0 seconds. Historically, I can confirm from casual conversation with an individual involved in pioneering the concept of LQB that the designation of 3.0 seconds as a threshold was somewhat arbitrary. The line of thinking arose using “baby bear porridge” logic whereby “quick ball” needed to be defined and 2.0 seconds was likely too fast, 4.0 seconds was likely too slow, and 3.0 seconds seemed about right. Given the information available at the time I would not have argued with the logic as the intent was to just lay down a marker and get on with it. However, now that there is more detailed information available around the breakdown, doesn’t it make sense to use the data to refine what is known about an area of the game that is not well understood?
For this study, seven seasons of Super Rugby data (2014-2020) were obtained and a total of 121,266 breakdowns were analyzed. Using the conventional classifications for ruck speeds, if we count the number of rucks that fall into each category we see an interesting trend. If LQB was intended to be an aspirational target, you might expect the distribution of ruck speeds to resemble a normal curve with the majority of rucks falling in the “regular” category and relatively fewer rucks categorized as “LQB” and “slow”. However, the distribution of rucks using the conventional classifications looks very different (Figure 1).
FIGURE 1: Ratios of ruck speed categories using the conventional classification (LQB = 3.0s, Regular = between 3.0s and 6.0s, and Slow = greater than 6.0s).
Can a target be considered aspirational if it is the most common outcome and can be achieved ~50% of the time? Doesn’t this sound like something that corresponds to a mode or a median which are both measures of central tendency? While it’s not the end of the world that the distribution of ruck speed categories doesn’t align with our expectations, what is important is if we can demonstrate that quick ball impacts the game in a meaningful way such as by increasing a team’s chances of scoring. One way we can do this is by looking at a plot of Expected Points versus Ruck Speed (Figure 2A). At first glance, the curve looks promising as it appears that faster ruck speeds correspond to higher expected points (i.e. look at the left part of the graph). If we overlay the ruck speed “bins” we can see if they align with any important features of the graph. Typical features that can be analyzed are peak/minimum values as well as mean values – in this case it would be the mean values for Expected Points and Ruck Speed. The mean value for Expected Points was 0.81 EP across all rucks analyzed. Determining a mean value for mean Ruck Speed is more problematic because there were many outlier values within the “slow” category. This is not surprising as it is not uncommon for teams to intentionally slow rucks down particularly when setting up for an exit in their own half of the pitch (e.g. a caterpillar ruck preceding a box kick). Including these outlier values results in the mean value that is skewed and not representative of ruck speeds in breakdowns that are properly contested. If we try to exclude outlier values by imposing a threshold at which rucks are considered contested versus intentionally slowed down, we will find ourselves making another arbitrary designation - which is something we would like to move away from. For this reason, and for the sake of keeping things simple, any measure of central tendency for Ruck Speed will not be considered in this analysis.
If we look at the curve in Figure 2A, you will see that none of the features really correspond to the conventional ruck speed “bins” in a meaningful way. This really isn’t unexpected because the relationship between Expected Points and Ruck Speed wasn’t known at the time when LQB was being defined. But now that we do know the relationship, wouldn’t it make sense to use the information to optimize what we know about LQB? If we support the narrative that LQB is a good thing and should improve a team’s chances of scoring, why not let the important features of the graph determine the thresholds for ruck speeds?
FIGURE 2A: Expected Points versus Ruck Speed for breakdowns in Super Rugby (2014-2020) using the conventional classification (LQB = 3.0s, Regular = between 3.0s and 6.0s, and Slow = greater than 6.0s).
Figure 2B proposes new ruck speed thresholds based on the insights obtained from breakdown data. The peak EP of the graph corresponds to ~2.0 seconds, thus by achieving this standard at attacking breakdowns you are maximizing your team’s chances of scoring points. Given the implications of this feature of the curve, would it not make sense that this should be the identifying ruck speed for LQB? Those of you with a keen eye may argue that the peak of the curve is actually closer to ~1.8 - 1.9 seconds and this is a valid point. However, in my experience, when trying to explain rugby concepts to players/coaches at all levels of the game it’s better to avoid referencing numbers – particularly those that involve decimals. Thus to avoid blank stares and ubiquitous millennial eye rolls I think it’s worth trading off a bit of accuracy for greater understanding by saying “we need to generate 2 second rucks” as opposed to “it would behoove us to facilitate the winning of possession of the ball in precisely 1.8 second increments”.
FIGURE 2B: Ratios of ruck speed categories using the conventional classification (LQB = 2.0s, Regular = between 2.0s and 4.0s, and Slow = greater than 4.0s).
The second feature we should investigate is the average EP value for all rucks, which is identified in Figure 2B by a horizontal orange line. You will see that the curve intersects this line at ~4.0 seconds. This means that for ruck speeds less than 4.0 seconds a team will score a greater than average expected points and for rucks speeds greater than 4.0 seconds a team will score a less than average expected points. This makes 4.0 seconds a great candidate to differentiate between a regular ruck speed versus a slow ruck speed.
Thus, based on the findings from an analysis of data from the breakdown, a revised classification for ruck speeds could be:
LQB = ruck over times less than 2.0 seconds
Regular = ruck over times between 2.0 to 4.0 seconds
Slow = ruck over times longer than 4.0 seconds
If we take these new classifications and look at their distribution, we obtain ratios that are more in line with our initial expectations (Figure 3). Rucks classified as “regular” would be the most common, and LQB would be harder to achieve as 18.5% of all rucks would meet or exceed the new threshold.
FIGURE 3: Ratios of ruck speed categories using the revised classification (LQB = 2.0s, Regular = between 2.0s and 4.0s, and Slow = greater than 4.0s)
Outside of the features that have already been identified, there are also some other interesting insights that can be inferred from the Expected Points versus Ruck Speed curve. Since 2 second rucks result in greater expected points compared to 3 second rucks, does this mean that faster rucks are generally better? Well….not necessarily. If we look at the area of the graph where ruck speeds are between 0 - 2 seconds we see that the expected points actually decrease as we get closer to zero. This might be due to the ball squirting out of the side of the ruck just as it forms, or a player may decide to pick the ball and go without waiting for support and get stuffed by the defence. These are possible scenarios where very quick ball could have an adverse result. Thus, it’s good to generate quick ball, but not too quick!
It is also interesting to see what happens when you look at the right side of the graph. This is the portion that corresponds to slow ball. In general, slow ball results in fewer expected points when compared to LQB/regular ball, but does this necessarily mean that it’s a poor platform from which to attack? If you look closely at the right side of the graph, you will notice that data is quite noisy (i.e. widely scattered) and in fact some points actually lie above the orange line (average EP). Though this might appear to be counterintuitive, the wide variance of expected points resulting from slow ball could be the result of intentional strategies in the Red Zone. It is already well known that many tries are scored from setpiece originating inside the opposition 22m, and that these scores typically occur within 3 phases. Beyond 3 phases, most teams will resort the “hammer ball” where teams will try to punch it in with a series of tight phases. Thus, many slower rucks result in scores due to teams playing hammer ball in the red zone – so slow ball isn’t necessarily a bad thing.
Given that there is a movement from World Rugby to speed the game up, there may be some appetite from some progressive individuals to adopt these new classifications for ruck speeds. This will be tough though as the LQB = 3.0s convention has likely been the acknowledged standard for the past 10+ years (estimated). It is important to note that even though these updated conventions may have been derived using means that are a bit more sophisticated than simply “guessing”, it is still not known whether or not the using the LQB = 2.0s standard will provide better/more accurate insights than the current convention (LQB = 3.0s).
Earlier in this post I alluded to the fact that the breakdown is not well understood. While we can confirm that data does exist, any meaningful insights – if they even exist at all – are not widely known or shared. Aspects of the breakdown that we would like to look at in the future are:
Do winning teams generate more LQB than losing teams? Will a team that generates 65% LQB from all rucks win more games than a team that generates 55% LQB from all rucks?
Does the sequence of ruck speed matter? Is it better to generate consecutive LQBs? If on a 3 ruck sequence a team generates 67% LQB is it better to have a Slow/LQB/LQB sequence versus a LQB/LQB/Slow sequence?
Is there a part of the pitch where generating LQB leads to more scoring? Red Zone? Danger Zone?
Is the Expected Points versus Ruck Speed curve different for competitions in the Northern Hemisphere compared to the Southern Hemisphere?
Do the results for the above questions differ significantly when using the LQB = 3.0s convention versus the LQB = 2.0s convention?
What other insights about LQB and breakdowns would you be interested in knowing? Let me know – it might become a future post!