Do you ever? My thoughts on the racing "market" being at or near its limit of efficiency stem from the following observations;
1; North Horseplayers (and owners/breeders) appear to obsessed with "Speed" ,almost to the exclusion of other factors.
2; Prior to 1990 most players used only the DRF for PPs which contained only raw times and daily track variants. Some would pay for "speed figs" or analysis from various sources such as Beyer figs or Brisnet (or any of the various "sheets") In the early 90s, much of this information became widely available either by the "Beyer " figs being included in the DRF or the "Brisnet" figs in those PPS. There may also be others of which I'm unaware .
3: Winning favorites (and second favorites) only slightly increased in their winning percentage) but "The Age of Speed Figs" was born and it wasn't long before a seemingly endless variety of Figs for both speed and pace hit the market. The winning figures in 1992 for the top 2 in betting were
roughly 32% for favorites and 18% for second favorites.
4; By 2002 ,interest in computer analysis for handicapping was gaining support and the Tracks started offering rebates, which paved the way for the "whales" . Although other players had also already started using computers ,the majority were still a bit wary of it . Still "The Age of Computer Handicapping" and the "Age of the Whales " were born.
5: By 2012, computer analysis continued to get better results (as did the whales) and analysis of all of the thousands (or tens of thousands)of possible permutations of all of the speed and pace figs as well as class ratings and jockey/trainer stats (and any other factors that were quantifiable) was not only possible, but likely tested in every possible way. The quantity of data ,however, is a (large) but finite number and ,of course, so is the number of permutations.
6: Now,in 2022 ,we see the winning percentage of winners at roughly 38% and second choices at 22% .( An increase of 18.75% for favorites and 22.22 % for second favorites) For the pair this is ONLY a collective 20% increase in 30 years including 20 years of extensive computer study. In the absence of significant new information,I don't see any significant change over the rest of this decade.
While my exact numbers may be off a bit, I think it's clear that we're near peak market efficiency and need to alter our methods significantly.
If everyone is using the same data ,they'll get the same or similar results and since that data is largely "speed oriented" data , it'll be "baked into the odds" and yield low rewards .