Just to jump in with an observation on the 17/28 figures that might help explain it a little, Mike
TheBluesBrother has obviously spent an inordinate amount of time working out each part of what he does now. It's easy, when starting out, to assume that the job can't be beyond human wit to handle, after all we are looking to calculate how fast a horse has run, we thing it's simply a case of speed = distance/time, it's only when we try to run with the idea and produce something that we hit problems - for example:
Okay, I calculate the speed of the winner, but how do I calculate speeds for all the other runners... how fast was the guy who finished 16l behind in 7th?
How do I compensate for, or calculate the effect that the varying weights carried might have had?
I have standard times, but we all know they run faster on firmer ground and slower on softer, how much do I have add or subtract to compensate for this?
Is there a reason most handicappers rate the runners in pounds?
As each of these, and many other issues. is dealt with there's often a long period of trial and error - because horses do not run like clockwork every time, so when A beats B by 2l and you decide that equates to A being 5lb better (for arguments sake) than B, the next time they meet B beats A despite being 3lb worse off.... how does that affect your calculations? (Short answer - like a grenade going off). Gradually a calculation is produced that seems to produce the best results over time - you know that your 'x lbs per length beaten' figure isn't perfect, but it's shown itself to be the best value to use compared to any other.
Then when you've tied enough ideas, estimates, and occasionally inspired guesses together you get a method that works as well as you seem able to manage, and if it picks winners often enough you realise that is as good as it is likely to ever get, because the fastest horse often doesn't win the race, you are actually picking the one that has the best chance of taking advantage of a good position and good ride - provided it didn't fall three fences back.
In other words, he tried lots of figures in the attempt to get his ratings to line up well enough with the official handicap ratings so that he could use his ratings to spot when the BHA handicapper had been too lenient or harsh and provided an opportunity to profit, and 17 and 28 are the figures that did the best job overall.
Dave