The growth in fantasy gaming and online gambling in recent years, while not necessarily a good thing, has correlated with advances in statistical analysis and modelling in rugby league.
The same thing happened in American sport years ago.
Like American sporting franchises, NRL teams have their own data analysts and, one assumes, their own proprietary databases and performance metrics.
Rugby league isn’t as advanced when it comes to statistical modelling and probably won’t ever be. It’s not just the later starting point. Baseball is structured, situational and very measurable. Rugby league, less so.
While I’m fascinated, I must admit that things like probability modelling and ELO ratings are a long way over my head.
I’ve never been good at maths. I make my living summarising written material for people who don’t have the time to read all of it.
I suspect advanced performance and probability metrics are unlikely to intrude on mainstream rugby league coverage in the foreseeable future – outside the gambling advertisements of course.
Perhaps we can find a middle ground between Bryan Fletcher and Pythago NRL? I mean no offence to Bryan Fletcher; I’m referring to his deliberately obtuse television persona.
The following is an NRL player value model by a maths dummy for other maths dummies.
First, what am I not doing here? The methodology is sufficiently general to be applied to players across all positions but is not a means of comparing players in different positions – for obvious reasons.
Performance is a function of opportunity. Most middle forwards spend much less time on the field than other players. Wingers and fullbacks tend to operate in a lot more space than players in the middle. Spine players touch the ball more.
Somebody much smarter than me might be able to figure out how to make positional adjustments.
Kicking is excluded because not many players kick in general play and even fewer attempt conversions, penalty goals and field goals.
Stats Insider has developed a true kicker rating system that ranks goal kickers based on their overall success rate and the degree of difficulty they’ve encountered. I can’t improve on that.
So, what am I doing? I’m measuring the impact of NRL players based on three broad and interrelated measures. Through impact I capture much of a player’s value.
As mentioned, some aspects of performance have been excluded. There are also aspects – effort, decision-making and positioning, for example – that can’t be quantified.
The first measure of impact is engagement in the game. This involves calculating how many times a player runs the ball or attempts a tackle per minute of their playing time. This is then extrapolated to total engagement over 80 minutes, with errors, penalties and missed tackles deducted to reach a net positive engagement value.
What about players who only play 25 to 35 minutes at high intensity? Do you apply a greater weight to the performance of players who impact a game over more than 60 or 70 minutes? If so, where is the cut-off and what are the weightings?
I say it’s better to extrapolate the available sample over the same time frame and remember we’re comparing players within categories. It’s just an estimate of value, not an absolute measure.
The second measure of impact is facilitation: tackle busts, line breaks, offloads and try assists.
It’s here we meet the POPE: the proportion of positive engagement rating.
This is straightforward. It’s how many times a player facilitates a potential attacking play, with the POPE rating being a proportion of the player’s net positive engagement.
The third measure is finishing. Basically, tries. The finishing POPE rating is calculated the same way as the facilitation POPE rating.
The quality of a try can vary from a player catching the ball and falling over the line to something like Manly’s Reuben Garrick-inspired 110-metre odyssey against North Queensland.
Either way, the scorer of a try benefits from the facilitative work of others. Therefore, facilitation carries more weight than finishing. Each line break and try assist is worth 25 per cent more than other facilitation and finishing values.
What does all this mean? All data is from the 2021 NRL season and sourced from the Fox Sports Lab.
Dally M Medal winner and all-round man of the moment Tom ‘Turbo’ Trbojevic has a POPE rating of 7.11 per cent. In other words, 7.11 per cent of his involvements have a significant impact on a game.
It’s not very catchy though. Here I borrow from the ingrained tendency of baseball fans to refer to percentages as whole numbers. For example, a player who gets on base 35 per cent of the time is verbally referred to as having an on-base percentage of 350.
I know it’s mathematically unsound, but we’re moving from the realm of math into that of popular culture.
Tommy Turbo is worth 711 POPEs.
In the realm of the very, very good is James Tedesco at 640 POPEs.
Further back, Ryan Papenhuyzen comes in at 472, with Clint Gutherson at 407.
If you were about to say, ‘I don’t need a spreadsheet to tell me that Turbo Tom’s the best,’ hold that thought.
Latrell Mitchell is worth a heavenly 731 POPEs, primarily because of his off-the-charts run metres per carry and because he makes fewer mistakes.
There are a couple of other things we can do with the POPE rating.
First, we can establish what constitutes a replacement-level player. At fullback in 2021 replacement level is somebody like Caleb Aekins or Blake Taaffe, who between them played 15 NRL games in 2021. The average of their combined performance is 254 POPEs.
Tom Trbojevic is therefore 65 per cent better than a replacement-level player.
Second, we can establish the opportunity cost of mistakes. This is expected POPE, or the x-POPE.
Tom Trbojevic’s 711 POPE rating could have been 721 had he not made any mistakes.
Penrith’s premiership-winning fullback Dylan Edwards is worth 290 POPEs, only just above replacement level. His x-POPE rating is 292. He might not make much impact with the ball, but he rarely makes mistakes.
I intend to do a few more pieces applying the POPE system in other ways. Any and all constructive feedback from Roar readers would be most welcome.