In his book The Science of Winning Bertrand P Fabricand discusses favourable races in terms of long term betting outcomes based on, for want of a better word, how contentious the race is. His belief is that when the race makes it difficult for the crowd to form a favourite, late information unknown to the mass market can be a deciding factor in forming a favourite and given the relative low losses on favourites, could be a route to a profitable system. He then go’s on to describe the most convoluted set of rules for determining bets that conjure up images of a new punter to racing, high on crack cocaine and just introduced to RSB.
The whole idea did however remind me of a discussion on the email group some time ago where I put forward the idea that ratings or at least the set I was looking at, performed best when the race was competetive. This met with some agreement from one member of the list and a commercial ratings provider. Examining the ratings I have for top rated performance to BFSP and trying to crudely differentiate less competetive races by using the price of the favourite, I came up with the following results for flat handicaps and top rated.

Fav less than 6/4
Top rated 2458 bets PL +19.7 roi 0.7%

Fav less than 2/1
Top rated 5247 bets PL +455 roi 8.6%

Fav greater than 11/4
Top rated 7336 bets PL +650 roi 8.8%

Fav greater than 3/1
Top rated 5435 bets PL +501 roi 9.2%

Fav greater than 7/2
Top rated 3052 bets PL +172 roi 5.6%

Fav greater than 4/1
Top rated 1781 bets PL +43 roi 2.4%

The above seems to suggest that contentious races are best but I am not sure whether the decline for larger priced favs is simply due to lack of sample size at these bigger prices or whether there is a sweet spot somewhere in the middle where races are competetive enough but not to competetive.
Perhaps other rating users or compilers could comment?.