The NRL is reaching the business end of the competition, and fans are now facing the distinct possibility of some unfamiliar teams contesting for the trophy at the pointy end of 2015.
Since the competition expanded to an eight-team finals series, there have been some winners and grand finalists very much unfancied at this time of year. Most notably the Wests Tigers in 2005 and Parramatta Eels in 2009.
We can look at previous years to see how likely your team is to be there or even win on grand final day based on the season they had.
2015 NRL GRAND FINAL KICKOFF TIME
To do this I have used a stat called PyEx, which determines the possibility of each individual team winning the grand final since 1999 with seven rounds to go in the competition.
PyEx is calculated by dividing the square of the points scored by the sum of squares of the points scored and points conceded.
It works better than for and against because it corrects for high and low-scoring teams with the same for and against. A team that concedes less will win more often.
Its predictive capabilities, despite the uncertainties of predicting this far out and the fact that it depends on who you’ve played so far, aren’t too bad.
I have uploaded a histogram of PyEx values plotted against the probability of winning the grand final,
or making the grand final
on Twitter.
As you can see, the better your PyEx, the farther you’ll go in the season. Teams that have a PyEx of 0.75 or higher winning the grand final 28 per cent of the time and making it 57 per cent of the time.
So, should you buy grand final tickets?
Buying tickets to the grand final can be a costly experience at this time of year if you only really care about your team. It can be much worse if you have to travel South from Queensland or even across the Tasman.
I had to fit a very rough model to the histogram (in red on the graphs on Twitter) to predict the chances of each team to make the grand final, or win it, this year.
It suggests that fans of the ladder-leading Brisbane Broncos should multiply the cost of their tickets by 2.7 to reflect their 37 per cent chance of making the grand final, and the North Queensland Cowboys by six 5 to reflect their 15 per cent chance. Disappointingly, the Sydney Roosters are a better than even chance of making the grand final.
Roosters and Broncos in the box seat
The Roosters are the team that is most likely to be victorious the grand final with a 28 per cent chance of winning. They are the highest ranked side on PyEx, which has actually won half of the grand finals in this time period.
The Roosters sit ahead of the Broncos (18 per cent) due to their impressive defence, but the Cowboys (8 per cent) are holding their lofty position on the ladder. Unfortunately, the 2014 premiers South Sydney Rabbitohs (7 per cent) have to really improve their form.
Outside of the current top four, the Melbourne Storm are the best chance (10 per cent), and as good as Billy Slater is, I would not be ruling them out this year either.
The Canterbury Bulldogs, St George-Illwarra Dragons (both 5 per cent) and the New Zealand Warriors and Canberra Raiders (both 4 per cent) have outside chances, while the Cronulla Sharks are less likely to win (3 per cent).
The one caveat is…
If you think this has been an even year, you’re right. The standard deviation (a measure of the spread of data) is very low this year, second only to 2006 since 1999.
Whilst teams in the bottom half of the eight have to improve to have a shot at the title, it is not an insurmountable task to match it with the teams right at the top. Whoever wins the grand final will be playing better footy in September and October than anyone has so far this year.
prince igor
Guest
The NRL are in real strife if the poor crowds keep happeningThe AFL are that far ahead its not funny and the NSW govt wants to waste 1.6 Billion on stadiums for a dying sport! The AFL had a bumper 235,328 fans pour through the turnstiles at its four matches last weekend, completely dwarfing the NRL’s 75,361 fans who attended three Sydney games and one in Melbourne. The television figures weren’t any better for the NRL as their total combined average of 3,332,000 was well behind the AFL’s 4,974,000. The numbers shoot down the argument that rugby league fans are staying at home because it’s easier to watch it from the couch.
pete bloor
Guest
Yep but then by seasons end they've typically been two spots lower on the ladder than they were in Pythagorean (which is why I pay attention then for all of the reasons listed above)
Sleiman Azizi
Roar Guru
If they are playing better they are.
Nathan Absalom
Roar Guru
This time last year North Queensland were 10th with a Pythagorean Expectation of 61.0%, and in 2012 were 6th and 61.3%. I know they're getting a lot of media attention, but are they a better team than last year?
pete bloor
Guest
I hear about it in the NFL all the time? It's predictive capabilities aren't as great but they are still better than ladder position
pete bloor
Guest
the north QLD is a strange one typically they've had the opposite problem when I've paid attention.
MikeTV
Guest
After Round 19, the hypothetical NRL ladder based on Pythagorean Expectation would be: 1 Sydney Roosters 76.3% 2 Brisbane Broncos 69.7% 3 Melbourne Storm 61.4% 4 North Queensland Cowboys 57.4% 5 South Sydney Rabbitohs 55.5% 6 Canterbury-Bankstown Bulldogs 51.5% 7 St George-Illawarra Dragons 51.1% 8 Canberra Raiders 50.7% 9 Warriors 50.4% 10 Cronulla-Sutherland Sharks 46.3% 11 Parramatta Eels 41.5% 12 Newcastle Knights 40.5% 13 Manly-Warringah Sea Eagles 39.6% 14 Wests Tigers 39.4% 15 Penrith Panthers 39.1% 16 Gold Coast Titans 34.9% When you compare this ladder to the actual ladder, we can see that the Canberra Raiders and St George Illawarra probably should have won a few more games in 2015, while the North Queensland Cowboys may have been very lucky at times.
pete bloor
Guest
You also will get the issue that teams with the same Pythagorean expectation don't finish on the same spot on the ladder, so identical Pythagoreans for say a team in 5th versus 4th changes the probability of winning as team 2 gets a pass mark in the first round and plays at home if it does need to paly in the second.
pete bloor
Guest
When I did a crude model a few years back I just pretended the week before origin didn't happen and weighted the past 5 games double and it did okay against the spread and back testing. The problem you'll have with 19 v end of season is probably a combination of the size of the data set again with origin, suspensions and injuries carrying more weight (and quality of opponent being more varied). Also the team construct over those last 7 weeks is probably going to more reflect the team playing the finals than the one in the first 7
Nathan Absalom
Roar Guru
You're right, the historical datasets are limited as the finals series have changed dramatically, and some reweighting would be very useful. However, another major problem for doing any sort of analysis like this for the NRL is how to handle the origin period. It's striking that Pythagorean expectations at the end of the season are far better predictors of the final series than round 19, but I haven't worked out the best way to see if this is caused by the origin period itself.
Nathan Absalom
Roar Guru
I'd say the probability dropping off is an artefact due to a relatively small number of teams having that good a stat. Also, it's not corrected for years when there are very poor sides (such as Wests in 1999) that lead to some years having several teams with high stats due to blowouts against low-ranking sides.
no one in particular
Roar Guru
While Pythagorean expectation creates a good talking point it is not relevant to the NRL due to sample size. It is why you hear pythag exp mentioned a lot in baseball, but very rarely in the NFL. Also, the constant rule changes mean the exponential can never be consistent
V.O.R.
Guest
A great sense of humour is far more life giving than a burgeoning intellect. I think you will be o.k. ;)
pete bloor
Guest
I've done it a while back for previous years (not current year) so at the start of the season was last years ladder, Pythagorean or finals finish the best indicator of likelihood of making the next grand final and Pythagorean was pretty darn useful though a reweighting to emphasise defence proved more telling as it is more consistent from year to year. The problem with using the historical is that there aren't a lot of data sets and it doesn't account for the differences of a Sydney vs non Sydney team (effectively home ground), it probably has more merit to look at the Pythagorean expectations (accounting for home field) vs the outcomes over the same period for every match, that would give a more reliable data set and then run scenario analysis that gives a % chance of each team based on how the finals play out.
Jay C
Roar Guru
Great start. Give us more. Can you do an article on how well this model would have predicted outcomes at the time of yeast. Statistical question. Why does the probability drop off above .75?
Darren
Roar Rookie
I had to Google Pons Asinorum... I think i must have a congenital form of it
Dogs Of War
Roar Guru
I would like to see some data for each year at the end of season (regular rounds) compared to what actually happened that year. Ie who was said was likely to win and/or make the grand final from the data analysis.
nopuritan
Guest
WTF!!! I feel like I've just fallen off the "Pons Asinorum "
Sleiman Azizi
Roar Guru
When rugby league changed from 3-point tries to 4-point ones, the number of fingers per fan was not commensurately increased. This severely curtailed the capacity to enjoy the game until a communique was sent out informing fans that bigger numbers equals bigger fun. Balance was restored to the rugby league universe.