Darkydog, your stats are undeniably right and my reason for using the 1.5 filter is not an irrational attraction for lightweights but purely because applying it resulted in a profit wrt favourites.
ie Pure curve fitting over the past 18 months data which is why I am keen to test for longer and frankly sceptical it will hold up.
It does however remind me of an article sourced by the Inracing people, whom I know you respect, titled The Principle of Maximum Confusion where from memory they contend that if a horse is made favoutite without the obvious credentials (eg not near the top of the weights where most of the winners come from) they are favourite maybe for more subtle but often no less valid reasons and are so relatively underbet and more profitable than those obvious highly-weighted favourites which are then overbet.
I am extrapolating from the strict application of the article but I think am in keeping with its principle.(for what its worth)