Thanks – this is really helpful and suggests that my Excel formulas may be off (I was trying to do the Neil route – had seen the video) as my numbers aren’t converging at over 1000 iterations. I have Winston’s Mathletics Least Squares (and weighted LS) sorted – but would like to have a different method for rating things (I like putting a number on things) and like what the SRS has to offer.

I noted SRS and Winston’s Least Squares have fairly similar number but like what SRS does with SOS being part of the rating. Started reading the Carl Meyer book (think SRS is derived from the Massey thesis paper of 1997 possibly – I think Neil mentioned this) and noted he had to put a constraint in to his matrix and wondered if this was the case.

Using the Neil method, no constraining needed. I do not know if that’s the case with the Mathletics model.