Bhagwan and co.

Still working on these figures because they bother me. Reviewed the whole thing. What has happened is.

*Did an initial test of the idea on 60 random races, all races, any races extracted from the Sportsman, May, June races. Wow! An amazing resut. Stacks of winners. Even more placegetters. And paying! A $17.80 winner, for instance.
*Expanded this to 180 races. Same time period. Still great guns. Great strikes.
*Tested it out on a batch of forthcoming races. Did really well.

The trouble was my sample of 180+ included several spectacularly good race days, a couple of really good ones, several bad ones but no shockers.

*Tested it on another batch of forthcoming races. Ordinary.
*Did a bigger sample. For this I've had to turn to old copies of the Herald Sun, whatever I can get. Batches from January, Feb, March, April. Some from 2001. All races, any races.

The more I add the more the overall pattern emerges. You strike bad periods. A couple of race days in a row in April, shocking. Not just bad, but way off. One out of eight at Flemington. None out of five, Victoria Park. Then you hit good days and then amazingly good days. But in my initial 180 I didn't strike any of the shockers the larger sample revealed. The pitfalls of a small sample. If you'd used this mini-system on all races any races in my first sample period, May, June, you'd have struck some great days with a healthy strike rate. Some days stacks of winners. Other days, like today, a swag of placegetters. But the larger samples show this is not reliable. The first sample was even, but its not over time. Can do very, very well or very, very badly. (Definition of a "fun" system I suppose.) But over time the bad patches win. Which is what I expected from the outset really.

But there are spectacular days!

The bigger the sample - trying different batches since they're so uneven - and the closer I look the less enthused I am, but then you get days like today that keep me thinking it is worth pursuing. Beware the bad days!

Hermes