Play the ball speed and “tunnel ball” has been in the NRL news cycle over the past week, on the back of a story by Adam Pengilly in ~~THIS MASTHEAD~~ the Sydney Morning Herald.

In this article, Pengilly noted that as the Panthers ran down Parramatta during the final minutes of their game in Round 23, they only played the ball correctly once in 11 tackles. Every other time the Panthers did not make a genuine attempt to strike the ball with their foot.

The reasoning is obvious, and Pengilly does a great job of explaining why. The desire for blazing fast play the balls is obvious, as it reduces the time opponents have to set their line and defend following tackles. Not bothering to put your foot to the ball can shave fractions of a second of each attempt.

This led me to check if play the ball speeds were faster this year than prior years. The answer is yes. In 2024 the average play the ball speed is about one tenth of a second faster over the same rounds last year, which was about six hundredths of a second faster than 2022. Tiny numbers, especially at this aggregated level, but it does show a trend that the game is speeding up.

What is interesting about this trend, and the Penrith example from the SMH above, is that every bit of prior analysis has shown that a teams’ average play the ball speed has no effect on winning games. Below is the margin and average play the ball speed metric for every game and margin for the past five seasons.

The R Squared here is 0.002, which basically measures from 0 to 1 how related two variables are. As you can see, 0.002 is almost as unrelated you can get, even if it is positively correlated. At an aggregated match level, a faster overall play the ball metric doesn’t guarantee a result or a better margin.

I’m not presenting anything new here. Prior analysis done prior by Chris Kennedy and Jack Snape on play the ball speed as noted above already showed there was no correlation in play the ball speed and winning or ladder position. Given these were both before the peak set restart era, I thought I’d do a more recent check and see if those truths still held up a few seasons on.

If you need more proof that there’s no link between winning, ladder position and play the ball speed, here’s the average play the ball speed of every NRL team this season.

Canberra has the fastest play the ball speed metric this season, at just 3.36 seconds per attempt. The Dolphins are second (3.39) and Sharks third (also 3.40). The worst team are the Broncos (3.51), only one hundredth of a second slower than Manly (3.50) and South Sydney (3.49). Last year Penrith were first in play the ball speed (3.43), but in 2022 they were 11^{th }(3.67). and 2021 they were 9^{th} (3.52). This year they’re fifth with a few rounds to go.

Even by opponent play the ball, there’s no significant trend.

Melbourne allows the slowest at 3.51 seconds (I’m shocked), tied with the Gold Coast. The Dolphins allow the quickest at 3.27 and the two best defensive teams (Canterbury and Penrith) are mid table at around 3.45 seconds.

And since I know I’ll get asked about it, presented here without any specific commentary is play the ball speed by referee.

As you can see, the aggregated metric of play the ball speed for a game or team has zero correlation with winning NRL games or the size of margin of victory. None.

You will notice how I’ve said “the metric” of play the ball speed. The aggregated metric of average play the ball speed for a game or team means nothing, but that’s not necessarily the case with individual fast play the balls, where they do have some impact in certain situations which we’ll get to shortly.

Part of this is possibly due to how it is measured. Quoting Snape’s article from above:

*“The NRL clocks play the ball speed as the time it takes from the referee’s call for tacklers to release, to the ball being played through the legs.”*

This means that the metric is only measuring how quickly a player can get the ball to the dummy half once they’ve started the process of the play the ball. It doesn’t factor in how long they were held down for, how many players were in the tackle and how slowly they peeled off. If I’m misinterpreting this please let me know. I’m not trying to single out data capturers or providers who do a great and thankless job, just explaining that there’s layers of context needed when using a play the ball speed number as a proxy for something like ruck speed or game pace.

It also assumes that the goal of every play the ball is the same – to play it as fast as humanly possible. That isn’t the case and we’ll look at some numbers shortly that explain why this is the case.

In addition to how slowly defenders peel off a completed tackle, there’s other variables that can influence play the ball speed. The first is the scoreboard, with the chart below showing the average play the ball speed by a team when a game is even, a team is leading, or a team is trailing.

When scores are tied, the play the ball speed is on average 3.37 seconds, approximately one tenth of a second faster than when a team is leading or trailing. When you’re in front you’re possibly trying to run down the clock, so slower play the ball speeds makes sense here.

Play the ball speed is also slower as you leave and approach the try line.

This makes perfect sense, as defenders will try to take every advantage they can to keep a team pinned on their line, and also to defend their own. Therefore a team that spends more time inside either 20 metre zone would most likely have a slower play the ball than one that spends more time at midfield.

If you combine the scoreboard data above and field position from the prior charts together you get a good idea of what is happening. Forgive me for limiting the chart axis range to magnify the trend of these lines, I know it’s punishable by death among the charting community.

An even game is a quicker game, and it is the quickest of all in mid field, which is where you can do the most damage with a quick play the ball as defenses have to keep pedaling backwards.

Play the ball speeds also increase at the end of each half, other than in a tied game where they get quicker as teams scramble to slot over a field goal.

The speeds really start to diverge about the 60th minute as match results become more certain.

The impact of each of these variables are smothered by aggregating play the ball speed to at more digestible team or game level. With that in mind, what situations does play the ball speed matter?

For example, in Snape’s article he noted that looking at numbers from 2019 and 2020 showed an increase in line breaks from a quick play the ball. Rather than replicate that, I’m assuming it still holds true. Instead, I’ve looked at the difference between play the ball speeds from regular play the ball to one that results in a try.

Overall since 2020, the average play the ball speed of an attempt that turns into a try is somewhat similar to a regular play the ball, both around 3.55 seconds. That changes however when you look at the location of the play the ball that results in a try. I’ve also weighted the try line in this chart by the volume of tries in each location.

Here you can see just how influential an isolated quick play the ball can be. A play the ball resulting in a try is at least one tenth of a second quicker than a regular play the ball from a similar position downfield.

Even 10 metres from the line, there’s a tenth of a second difference. Once you get more than 10 metres away from the try line it doubles to 0.2 seconds. This is why you’re seeing teams just roll the ball backwards and not try to put their foot on the ball, especially when they’re chasing points.

The overall probability of scoring a try isn’t higher from quicker play the balls, as most tries are scored close to the line where the average play the ball speed is higher.

Generally, they will come from a faster play the ball than usual in the same location, but a faster play the ball alone isn’t a guarantee of creating points.

All of this got me thinking about what could be a better measure of ruck speed outcome than play the ball speed. Why are we so focused on measuring the process of playing the ball over the result that comes from it?

First, let’s establish who the fastest players are by play the ball speed this season (with a minimum of 50 attempts), to see if it might tell you something.

We have a bunch of forwards, plus Nick Cotric, which I’m sure will please Dan at The Sportress. Newcastle’s Brodie Jones has the best number at 2.90, ahead of the straightest runner in the game Josh Kerr (2.935) and Klese Haas (2.969). OK, so we know they had quick play the balls, but what does that mean?

In a vacuum it tells us nothing more than these players get the ball to their dummy half the quickest. Is that the best way to measure the impact of a play the ball though?

Now we’ll check out the slowest average play the ball speeds by player this season.

There are five players above four seconds per play the ball, over one second slower than most of the names on the previous list of fastest players. Reece Walsh at 4.17 seconds per play the ball takes top spot, followed by Isaiya Katoa and Damien Cook. This list is mostly spine players, with a few outside backs and one traditional forward in David Fifita. Maybe Billy was right? (No, he wasn’t).

It makes sense that spine players would have slower play the ball speeds, as they’re likely to be caught with the ball closer to the line where play the ball speeds are slower. In midfield they’re likely directing attack rather than leading it, and less likely to be tackled with the ball. And players who are known to have good play the balls like Cameron Murray, will be targeted and not allowed to get up and play quickly.

So is there a better way to judge the value of a quick play the ball than just by timing how fast it takes? What if we checked on the outcome of each play the ball.

Friend of the site Charles Giess, someone who is far more intelligent than myself and has an honorable mention in the NFL’s 2022 data bowl, had mentioned that I should look at subsequent expected run metres as a way of evaluating the quality of a run and play the ball.

This makes perfect sense. The value in a strong run is not just the run itself, but in laying a platform for the proceeding tackles. If the run is good but the play the ball is slow, the value in that run is lost.

In the end I took Charles’ fantastic suggestion and tweaked it slightly. I looked at which players had the best Run Metres Over Exepcted per run (RMOE/run) following their play the ball this season, to see if there was any thing to be learned. If a play the ball was successful, the following run(s) would generate more metres than expected.

I also checked the percentage of tries scored from each players play the balls, similar to the try scoring probability of this site’s expected point model (ETXP). After all, what we’re looking at here is generating points and winning games, and scoring a try is the best path to doing that.

Neither of these two metrics correlate with fast play the ball speeds (R Squared values of 0.004 and 0.0005), yet tries and run metres have two of the highest correlations with winning matches. Another nail in the coffin for aggregated play the ball speed.

Looking at RMOE/run, here are the top 15 players whose generate the most expected metres following their own play the ball.

Chris Randall from the Titans takes first with players gaining +2.54 metres more on runs following his play the balls, well ahead of second placed Brandon Smith at +2.19 metres above expected. Wade Egan places third at +1.378 RMOE/Run, which indicates to me that a dummy half run sets the defenders on the back foot more than a standard hit up and should be used more regularly with the right player.

Melbourne’s Josh King is the top middle forward at +1.36 RMOE/run, followed by Mathew Feagai and Sione Katoa. I’ll also note Jack Howarth in this list, who has become an integral part of the Storm this season, and it’s not surprising to see a smaller middle like Canterbury’s Bailey Hayward thrive under a quicker pace of play.

The interesting part about these names is that it’s not props that are generating more metres than expected following a play the ball. This might be due to the subsequent runs being shift plays which are more likely to generate negative metres against expected than a regular hit up.

The player with the quickest play the ball this season, Brodie Jones, ranks 57^{th} here at +0.464 RMOE/Run, or nearly half a metre above expected. Still a good number but every single player above him has a slower play the ball speed, including Egan who is right at 4 seconds per play the ball.

Here’s the bottom 15 by RMOE/Run on subsequent runs.

The one thing in common with the above 15 players is that their play the ball speeds are all quite different, from Damien Cook at over 4 seconds down to Leo Thompson at nearly three seconds. Yet all 15 of them generate at least one metre fewer than expected on subsequent runs.

Now let’s check the top 15 players by the percentage of tries following a play the ball. These numbers will be skewed somewhat by the tackle number they’re playing the ball on, but there’s some really interesting names here.

First place is Jarome Luai at 11.6% (a try scored on 8 of 69 of his play the balls), which would indicate his attacking the line does a great job of setting up a try. Next is Dogs half/utility (like that narrows it down) Drew Hutchinson at 10.3%, with Randall popping up again at 9.4%.

Sharks middle Royce Hunt is the highest traditional forward on this list at 8.6%, with 5 tries being scored following his 58 play the balls. Karl Lawton also shows up again here in 5^{th} at 8.2%, behind Felise Kaufusi at 8.5%.

Kaufusi’s position here is interesting when you compare it with Lawton, who ranked third above. The Dolphins forward is 6^{th} in percentages of tries scored following a play the ball, but has a negative RMOE/Run number at -0.056. This would indicate that Kaufusi does an excellent job close to the line at setting up a score but isn’t as valuable in yardage situations.

Going back to the quick play the ball chart prior, we noted Kerr (2^{nd}) and Cotric (5^{th}) on this list as having some of the fastest play the ball speeds in the competition. Yet if you look at the percentage of tries scored from their play the balls, they have resulted in a combined **zero **this season. Speed matters in rugby league, just not always from play the balls.

Narrowing it down to goal line play the balls only, Tui Kamikamica has one of the better rates with 4 tries following his 10 play the balls inside ten metres. Hunt has had 14 play the balls inside 10 metres this season, with five of them resulting in tries on the following play. Jared Warea-Hargreaves the exact same strike rate of five from 14, proving he can still create an impact in short bursts. Another Shark in Thomas Hazleton has a similarly good record, with nine tries following his 29 play the balls inside 10 metres.

I’m reasonably happy with the results of these methods, even if they aren’t perfect. It shows there is some value in these numbers. There isn’t anything here that looks specifically wrong at first glance, at worst there’s some unexpected names. If it passes the standard eye test then I’m probably on the right track, and will see how these track moving forward.

As established, the metric of aggregated overall play the ball speed is virtually useless in evaluating a team or player. In specific situations you can quantify how a quick play the ball impacted a single possession. However by changing the focus to look at the outcome of their play the balls, rather than the process itself, we can better understand the players and positions that can generate better production from a quality, but not necessarily fast, play the ball.

Love this analysis, I have been noticing a lot lately that tries are consistently coming of the back of good dummy half service, as opposed to a shit load of dummy half passes going to the receivers shoulder. What is the actual impact of dummy half passes out in front compared to the opposite?