On 2003-11-10 07:53, crash wrote:
I was wondering how many systems you run concurrently ? I understand the logic of it but not sure what you do when two [or more] systems come up with different selections in the same race. Do you still place the bets ?

Hi Crash, I am currently running 16 different "systems", 13 are Saturday systems, 7 are midweek systems, and 4 are all days systems. 6 of the systems are all based around Price Predictor Better Ratings, so they'll always pick the same horse, and some of the systems are very selective, with only one bet per week on average.

The problem of what to do when I have two runners in the same race presented itself the first day I ran multiple systems, and I decided then that I would bet them all. My reasoning was that if one system didn't win, the other might, and I don't expect my systems to win each week. I don't dutch, I just bet the amount determined by each system's bank. I place about 40 bets on a Saturday, and will bet multiple horses per race in about 3 to 5 races each day.

This is not as bad an idea as it sounds, as out of the 7 "discrete" systems I run on a Saturday, 3 of them are "mainstream", and the other 4 are longshot systems. I have absolutely no problems betting an "in the money" selection for a win and a longshot selection each-way in the same race.

The other thing that needs to be determined is what to do when multiple systems select the same horse. Do you double-up or not? I have decided that when totally different systems select the same horse then I will double-up, but if systems based around my Price Predictor ratings select the same horse, then I don't double-up.

...snip... but more importantly because a lot of sane filters just can't be configured so are replaced with meaningless ones that lead to error. Within 2.5 lengths from winner [as opposed to a finish position, except first of course] is an example that I have just discovered can be checked. What about 'must be no worse than 8th. at the 800m', which establishes being on the pace ?

Price Predictor stores the sectional times if available in the form, but these are not available for system testing unfortunately. However, sectional times are only available for mainly metro tracks, so are pretty much useless for midweek betting. On Saturdays, there are a lot of horses in a lot of races that are racing at Metro tracks for the first or second time, so sectional times might also be misleading here as well. Don't get me wrong, I like the idea of being able to select on the pace horses, especialy at certain tracks. And if you're restricting your betting to the better class Saturday races then they can be very advantageous, it's just that I don't restrict my betting to that area.

The whole business of systems is very exciting but I have never had much faith in them. Working out software solutions that fit decent filters seems to have more merit that force fitting filters to suit the software [most systems around].

I agree wholeheartedly. I find that working out the systems is more fun than betting the systems. I actually have two databases, the Price Predictor database, and a database that I have created myself which links to the Price Predictor one.

I have done this for two reasons, firstly my program is several orders of magnitude faster than Price Predictor. I like to test individual variables over their entire range, so will run thousands of analysis. Secondly, I can include variables that are available in the form database but not available for the system selector module. This means I can actually test on any variable that is available in the form... This latter has allowed me to create some very good (admittedly backtested) systems, but these are actually performing reasonably well at the moment.

For instance, one thing that I have played around with quite a bit was selecting each runner's "best run" out of their last 5 runs, and comparing elements on those runs. So, for instance, rather than specifying "won last start", you might specify "won best run", or rather than "distance change from last start of -100..100", you might select "distance change from best run of -100..100"..... I haven't finished this particular area of analysis yet, but it is looking quite promising.

I haven't played with sectional times yet, mainly because I haven't figured out what to do about horses that don't have this information available due to where they've been running.

Beaten lengths is something that I've tested a lot more extensively than can be done in Price Predictor, as I've looked at beaten margins over the past 5 runs, whereas Price Predictor only tests last start.

I believe that designing a system which goes beyond what is capable in a normal system builder gives me an edge that a lot of other punters won't have. So far it seems to be working.

Cheers,
Chris.