A glance at the NRL with Poseidon ratings
If you read my tips for round 1, you would have seen the Poseidon rating making an appearance. The purpose of the Poseidon rating system is to look at each team’s offensive and defensive capabilities separately, as well as their home and away performance, to see if their winning record is concealing strengths or weaknesses.
The underlying principle is quite simple. We look at how many tries a team scores, at home and away, and how many they conceded, similarly at home and away, and see how that stacks up against the league average over the previous twenty-four rounds.
The league average moves over time. In 2013, the home team scored 3.9 tries on average and the away team 3.2. In 2018, those numbers had changed to 3.6 and 3.1 respectively. While 0.3 tries per game may not seem like much, over a 192 game regular season schedule, that’s 57 tries that have gone missing, or about 316 points, just for the home teams.
To keep ratings relative to a fixed point, Poseidon is expressed as a value-over-average. That is, a team with a +22 rating is 22% better than the average, which might mean 22% more tries scored or 22% fewer tries conceded (78% of the league) average. Remember that for this system, positive numbers are better than negative ones and bigger numbers are better (or worse) than smaller.
We can then plot these numbers on a graph. Here’s where each of the NRL teams sat at the end of 2018 for offence and defence:
And for comparison, here’s were they were offensively and defensively way back in 2017:
The trick is to get your team out of the bottom left quadrant, where you have below average defence and offence, and up into the top right.
There’s a couple of interesting takeaways from these graphs for me. The first is the location of particular teams. It’s no secret that I think that the Knights’ performance in 2018 was not reflected in their win-loss tally. They finished the season with the lowest form and class ratings, as well as the worst defensive and home performances, in the league. Yet they rose to eleventh on the ladder with a 9-15 record from the wooden spoon position in 2017, without moving their ratings too much. If the same team had carried forward into 2019, I would have expected to see some reversion to mean.
However, given the huge flux in their starting roster, both from transfers and from players returning from injury, we do more or less have a new look Newcastle outfit. The signs on Friday night were promising for a big season ahead. The same perhaps could not be said for the Titans who are stuck in the same quadrant but have not recruited as successfully for the new year.
The second takeaway is the expression “defence wins premierships” is borne out well here. All of the top teams from each season feature in the top half of the graphs. Look at Souths’ rise from cellar dweller in 2017 to contender in 2018: they improved both areas of their game but defence is a much bigger contributor to their rise up the ladder than improved offence. We see a similar trajectory with the Tigers shooting up defensively but not changing much about the effectiveness of their attack. Contrast that to the Eels and Cowboys, who slid down both the ladder and their defensive ratings.
The third is that teams that are generally good at attacking are good at defending and teams that are bad at attacking are bad at defending. This might seem obvious but it’s not at all intuitive why a team can attack and defend, given that these are separate skillsets. One thing it hints at is that the key to success is good organisation, whether this be expressed in tactics, strategy or the ability to improvise successfully on the fly. If that holds true, then instead of saying “defence wins premierships”, we might say “good organisation creates solid defence, which wins premierships.”
To really drive those last two points home, here’s the Poseidon ratings for every grand finalist and minor premier since 2014:
Only two teams had negative offence, both of which were blown off the park in their respective grand finals, but all of them had an above average defence. Indeed, excepting the 2015 premiers and the two negative offenders, all the teams were substantially above average in defence.
This is a fairly basic level of analysis. The real value of the Poseidon system is in exploring how tries and other events in rugby league matches follow a Poisson distribution. I’ve only started to scratch the surface of what this might mean for predictions, because the system is new and these things take time to flesh out, but the end goal would be providing a probabilistic forecast for try scoring and match scorelines that can be calculated with a spreadsheet.
If you want to have a first pass at it, it’s not difficult. Get into Google Sheets or Excel, type in:
=poisson (x, y, false)
where x is the number of tries that will be scored by a team in a game and y is the average number of tries that the team has scored in a game. So if you want to know what the odds are that the Warriors will be held tryless, x would be 0 and y would be 3.5 and you should get something around 3%. If you were wondering how accurate that is, it’s not bad and gives us something to work on into the future.
You can read a longer piece about the details of the Poseidon rating system here.