Quote:
|
Originally Posted by michaelg
without knowing the criteria of your selection process someone might suggest the distance of race, barrier, days from last start, place/unplaced last start, etc but these might already be in your elimination rules.
|
Michael I take your point. Without going into too much detail we're basically just playing the percentages here. There's no classic form fundamentals involved.
By way of example and in very simplistic terms it works like this.
First I take a very simple view of the race eg. there are ten runners in the race therefore every runner has one chance in ten of winning. I then gather some "facts" about the runners the most important part being that they are independant of each other i.e. one "fact" doesn't influence another. I crunch these "facts" together for every runner and get a probability at the other end. I look at the top four, subtract their probabilities from my simple one chance in ten view of the race and those with a positive result are the lay selections.
eg. Runner 4 is the lay selection
Runner 1 .10 - .31 = -.21
Runner 2 .10 - .15 = -.05
Runner 3 .10 - .11 = -.01
Runner 4 .10 - .07 = +.03
---
Runner 5 .10 - .05 = +.05
Runner 6 .10 - .02 = +.08
.
.
.
You'll occasionally get a group of runners that are too close to seperate into a top four. In this case I include them all. That's the reason that some races have three or four selections. Some races have no selections, and I've a personal bias against 2yo, 3yo, Maiden and Jumps races so I ignore those, but I've no reason to think the method wouldn't work with them. Big fields can also be a problem and I'm still experimenting a bit there. And yes, the runners with the negative result in the top four do win more than their fair share of races.
Getting the "facts" right is the hard bit and the bit I'll keep to myself but they should be very simple and proven to hold true over many years. There's a myriad of them though. It doesn't necessarily hold that the more you use the more accurate you are either. I've got it down to three that I use in Australia. The method works in the UK too, and was the basis of my "Beat the Donkey" thread. Interestingly what works on a Saturday in Australia doesn't work during the week where I think private information is of most benefit.
You'll get more of an idea here.
http://www.flatstats.co.uk/ppp/viewtopic.php?p=7441
That's enough from me the rest is mine
