The Roar
The Roar

Mark McGrath

Roar Rookie

Joined June 2017

546

Views

1

Published

4

Comments

Mark's passion is analysing sports performance using statistics and in particular, rugby league. Although he did stats at uni, he didn't learn much because that subject was on during Friday arvo happy hour. Back in the 90's, he spent 7 years as a handicapper with the AJC. Years later and looking for some new brain challenges, Mark decided to make up for lost time and teach himself statistical analysis and tackle rating and forecasting NRL teams. Away from the spreadsheets, Mark still plays cricket in a local 1st-grade comp (just), despite being on the older side of 50 and is a 4th generation South Sydney supporter.

Published

Comments

Good question.

I plan to do an amended version of this analysis on all-time try scorers in the NRL. Being a backrow forward with a high try scoring rate you would have to think he would be right up there.

But he has some hot competition like Harold Horder, who scored 102 tries in 86 games; a try scoring rate higher than Ken Irvine.

Who really is the best try scorer in the NRL this season?

BMoz came in just outside the top 25 (#26) at a rating of +3.92%.

Who really is the best try scorer in the NRL this season?

Small correction; in my explanation above I said:

GMP = Game Minutes Played (total game minutes divided by 4800)

This should have read:

GMP = Game Minutes Played (total game seconds divided by 4800).

Who really is the best try scorer in the NRL this season?

Thanks guys,

Following the editor’s guidelines, I didn’t want to bore people with too much data and statistical nerdy stuff. But for those that are interested in this side of things…

The formula
The formula was derived from a multivariate, linear regression on the 2013-20 dataset. The formula extracted from this regression was:

Expected Tries = – 6.627 – (0.0184 * TT) + (0.357 * GMP) + (2.039 * PV)

where:
TT = Team Tries
GMP = Game Minutes Played (total game minutes divided by 4800)
PV = Position Value (a factor derived from the % distribution of tries across positions, with the Interchange position =1)

The Adjusted R Squared value for this formula (a statistical measure on how good the formula fits the data) on the 2-13-20 dataset was 74.70%; a pretty good result considering that the formula includes no skill factors.

Extra results data
I would like to post extra results data, that will enable you to make more sense of the ratings, but the content management system running this site provides no easy way for posters like me to publish this data. I’ll contact the eds about this.

Potential improvements to the model
There are two areas where I think the model can be improved:

1. Adjusting Team Tries to the total number of tries that were scored when the player was on the field.

80-minute players are not affected by this in the current results. But less than 80-minute players are because tries that were scored when they were on the bench are counted as an opportunity against them, when in fact they weren’t (you can’t be expected to score a try when you are on the bench).

2. Using Tackled in Opposition 20 data

I think this would be a strong opportunity variable to use. But until I get the data and test it, I don’t know if that would really be the case or not.

When tries are scored
Daniel, you would first have to calculate the distribution of tries across game time, to see if there were any potential advantages or disadvantages here for players playing less than 80 mins. If there were differences, then you would have to do a significance test to see if these results were just a fluke (ie randomness) or if there was something more to them.

But potentially yes, you could have something to use there to improve the model.

Further updates
If people are interested, I’d be happy to post an updated table of results after each round.

Who really is the best try scorer in the NRL this season?

close