We spoke before about "Betting Without Validation", meaning that if we see something anecdotally happen that fits our built-in bias, we believe it to be true, eventhough when we look at it with some backing numbers it might be totally wrong.
This struck me again reading this forum on Paceadvantage.com. A poster asks what a good estimation of what a horse might pay to place is. Dave Schwartz, proprietor of Horse Street Software, replies around "40% of win odds"; meaning that if a horse is 5-2 (2.5-1) the horse should pay around $4 to place (2.5 x 0.40 = 1.00-1 or $4). Someone replied to Dave that he checked 10 days of results and that is wrong. Dave came back with hard data. He checked 25,000 races, and for horses under 7-1 the ratio is 0.38 and for all odds ranges it is 0.366.
It is something that has struck me in my handicapping life that never ceases to amaze me. I still trust, but now I look to verify.