The week before State Of Origin fills me with a mixture of sadness and happiness. I’m happy because every day is one step closer to one of the greatest sporting events of the year* but sad because it means weeks of disjointed and unpredictable NRL matches.
*Every year I see more problems with State Of Origin and it is starting to lose its lustre. Still a great event but there are some major issues that need to be addressed soon. But that’s a topic for another post.
Like a true data nerd, I have found that executing some fun data analysis is a great cure to ease my bye-induced NRL sadness. This time around I’ve chosen to analyse my NRL ranks by comparing them to a few of the common or socially acceptable methods people have for ranking the teams. And finally, for a bit of fun (have I mentioned I’m a nerd yet?) I invented a newish method for ranking teams, based on one of those socially acceptable methods, which I think serves as a decent first order approximation of which teams are playing well this season.
My NRL 2017 Ranks
Here is the current ranking of teams, ordered best to worst, based on my custom NRL ranking algorithms.
|My NRL Ranks|
If you want to see (kinda)pretty graphs for how these have changed over time this year than check out the NRL 2017 Rank page.
These ranks are actually a combination of four separate ranking algorithms I have that, when combined, provide a fairly decent estimation of which teams are doing well for the season and how their form is trending (good, bad, or…flat).
Comparison: My NRL Ranks vs NRL Ladder
The first and most obvious comparison is to look at how my ranks compare when put alongside the actual NRL ladder which rates teams by competition points (ie who has won more matches). IThe tiebreaker is the overall “points difference” (i.e. the difference between points scored and points conceded per team.)
My NRL Ranks
Interesting. They are similar enough to give me confidence but also different enough to give me confidence. Let me explain that contradiction.
The NRL ladder is a very basic method for ranking teams but as a first order approximation (that’s twice I’ve used that phrase now…I must think it sounds cool) it does a decent job. As such, I don’t want to see my algorithms spitting out ranks that are wildly different. If they did it would mean I’m heading in the wrong direction.
However, seeing some differences is quite positive too because it means that my algorithms, which are hopefully more advanced, are finding subtle methods for differentiating teams that the other simpler mechanisms cannot do. At least that’s what I tell myself.
Enough waffle, lets do some analysis on the key differences between my NRL ranks and the NRL ladder.
- I rate the Titans! – I have them 8th when they currently sit 13th on the ladder. Five places apart is massive, especially when all others teams have a difference of two places or less between to the ranking methods. (Personally, I don’t rate the Titans at all.)
- Sea Eagles round out my top four. Whilst only two places above their position on the actual NRL ladder, I thought this was important because of the home ground advantage the top four teams get in the finals. A stronger team that just misses out on a that fourth finals place will be worse off compared to a weaker team that sneaks in to the top four.
Comparison: My ranks vs Betting odds
A fairly simple idea here. Odds are a representation of probability so therefore I could look at the NRL 2017 Premiers betting markets and use their odds (or converted to what I like to call implied probability) as a method for ranking the teams.
I like doing this for two reasons – you assume the people putting their money on the line have a fairly decent idea about what they are doing, and if you find a few outliers they just might be worth a bet 😉
My NRL Ranks
Once again, a few similarities and a few discrepancies. If I were to assume my ranks are awesome and 100% correct (arrogant muc?) then I would say the betting market has it wrong in the following ways:
- Roosters are overrated – 6th in my ranks, 3rd in betting
- Panthers overrated – 10th vs 6th
- SeaEagles underrated – 4th vs 9th
- Dragons underrated – 5th vs 8th
- Titans underrated – 8th vs 13th
There is a simple observation here. The betting markets move slower than the weekly ups and downs of an NRL season so there will be some skewing towards the original betting value at the start of a season.
Say what? Basically, teams that were heavily favoured at the start of the season will not suddenly drop out of contention after a bad start. Instead they will slowly drift out as people come to terms with their earlier wrong predictions. But my ranks are a bit more judicious and these teams, if they are playing badly, get booted to the bottom quicker.
Side note: from that above list, the underrated teams are potential ‘good value’ bets. As I mentioned earlier, I don’t rate the Titans (but that could be human error/bias) so I wouldn’t be keen betting on them. But both the Sea Eagles and the Dragons are looking decent after 11 rounds so I do see potential there.
Comparison: My NRL ranks vs Points Difference
The third comparison is still fairly obvious if not as common. Here I ignore the current competition points and sort the teams solely based on their Points Difference (PD).
If you’ve ever checked out an NRL table online or in a magazine (if any still exist) then the chances are you’ve seen a PD column alongside each team. It is so prevalent because it is deemed to be a significant indicator of a team’s performance. Take all the points they have scored for the season and subtract all the points they conceded and you get a rough indicator of strength. Obviously better teams will/should have higher PDs and the worse teams will have negative PDs.
My NRL Ranks
Points Difference (PD)
This crude methodology actually seems fairly close to my algorithmic outputs. The biggest discrepancies are:
- Those bloody Titans – 8th (my ranks) vs 11th (PD). This is a noticeable commonality in these comparisons. Mental note: My algorithms may be overrating the Titans.
- Dragons in top two? – I have the Dragons at 5th, which is higher than the previous two ranking methods, and yet based just on PD they are second only behind the Broncos. Very interesting…
Comparison: My NRL ranks vs My Adjusted Points Difference (ADP)
Finally, I unveil my new mechanism for a simple(ish) approach to ranking teams based on the PD method but adjusting for the relative strength of opponents played. The idea is that the PD for each team does not account for teams playing stronger, or weaker, teams. That is, just because you can rack up a big scores against the cellar dwellers doesn’t necessarily mean you are a worthy of a high rank.
If there is interest, and perhaps first and foremost if there is any merit in the idea, I can write up a more detailed description later. For now, let’s just look at the tables:
My NRL Ranks
Interestingly there is not much difference here. Two main observations:
- Titans are 11th (vs my rank of 8th). You know my views, I wont beat this dead horse any longer.
- The Storm are 5th! This is the lowest any method has ranked the Storm and on the surface it seems a little crazy but when I cast my mind back to the Storm games I’ve watched this season I have noticed they have lacked their usual polish. Sure, they are still winning games but they are not dominating them as much and sometimes they are grinding out low scoring wins with grit rather than panache. Or maybe I’m just remembering things with some confirmation bias now?
Summary: All Ranks Side by Side
Here is the full table of all ranking methods, side by side.
Points Difference (PD)
There’s lots of potential for analysis and I could easily spend the next few hours going down rabbit holes debating the pros and cons of each ranking method but I’ll save that for another time. Or perhaps for my wife who loves nothing more than hearing my inane dribblings about sports, ranking algorithms, and data analysis. At least I assume that’s why she married me.