Actual W-L vs "Expected" W-L

[MightyMagpie]

I'm not a serious statistician, but just for giggles I have been playing with algorithms that use Points For (PF) and Points Against (PA) to calculate an expected W-L record.

The difference between actual and expected records can be regarded as "luck".

This only considers PF and PA - it does not consider "luck" regarding injuries for example.

The formula that has best correlated (0.93786) to actual results over 2012-2015 is pretty simple (and very close to the original Bill James formula used in baseball):

win % = 1 / (1 + (PA / PF) ^ x

The variable exponent "x" is at ~5.6 right now to best correlate with the period 2012-2015.

FWIW in 2015:

FRE +1
WCE -2
HAW -3
SYD 0
RIC 0
WBD +1
ADE 0
NM 0
GWS +1
GEE +1
COL -3
PA -1
STK 0
ESS +1
MEL +2
CAR +2
GCS 0
BRI +1

From this it could be concluded that HAW and COL have been "unlucky" not to have had more wins given their respective PF/PA.

Using Champion Data's premiership level benchmark 100 PF and 80 PA per game, it results in a W-L record of 17-5.

[RudeBoy]

Taking 11 (ie half off the 22) wins as a bare pass mark, I've multiplied it by our current percentage of 114.9 to derive our expected number of wins for the season, which is 12.64. Rounding it to the nearest whole number gives us 13 wins for the season, which will get us into the top 8.

I'm not sure if this methodology actually means anything, but nevertheless I like the result.