First of all, these days many of us are doing a lot of number analysis and even if I disagree with your premises, I have to tip my hat for your attempts to analyze the data in a way that we may not have thought of. So while I disagree with some of your findings, it's nothing personal, just a different interpretation of the data that you can take with a grain of salt. (Full disclosure-I'm looking at a horse that doesn't meet your criteria to be a major player at odds of >15/1)

When 3 of the last 5 winners have paid 15-1 or more, I wouldn't be trusting the public's making of the odds as a criteria to toss a horse. If I had a horse who met my criteria except his odds were over 15/1 and because the worst knowledgeable betting pool of the year made him an overlay, (a player's dream) I'd have a tough time tossing him and seeing him pay $35. The fact that the win percentage for slow non-contenders above 3.5 is the 2nd highest win % (13.04%), only surpassed, and not by very much, by fast contenders below 3.5 (14.51%), would lead me to believe that maybe odds aren't a very good criteria to use as a toss item & when you're upgrading higher dosage horses all I think this proves is maybe it's time to just say dosage is an outdated method for picking Derby winners, especially since only 3-4 horses in this field have a dosage index above 3.5.