Comments invited about applying filters. Here is the outcome of applying various
filters to a set of actual results: you can see that applying the 21 days or less since last
start made hardly any difference to the S/R or ave. divi but did alter the POT
significantly , the winPOT going down and the place pot going UP!!



apply rule (a)
overall 21days
% wins= 33.33 33.3
%place= 71.56 71.87
WinPOT= 8.33 supertab 6.56
plce POT= 6.53 supertab 7.01
ave win= 3.25 supertab inc stake 3.19
ave place= 1.49 supertab inc stake 1.49

Then apply rules (b) (c) (d) individually, then a+b+c, then a+b+c+d and see what happens

(explanation) rule (b) place S/R of at least 50%, (c) = Last Start Winner, and (d) = not
ridden by 3kg claiming apprentice.


(b) (c) (d) a+b+c a+b+c+d
50%S/Rplc LSW noAPP
33.78 40 37.03 41.17 45.16
77.07 80 76.54 82.35 83.87
7.02 30.6 16.91 36.17 49.35
11.93 19.06 15.69 19.2 20.74
3.16 3.26 3.16 3.3 3.3
1.45 1.49 1.46 1.45 1.44

My question is: do you number crunchers find that this is the case in most cases or is it just
unique to my particular set of results???

Well that didn't copy and paste very well, anyone who is interested I will send by e-mail. my address is lumbasakabayo at hot mail dot com

The guts of that is the end result after applying all filters was as follows:
Wins = 45.15%
Places 83.87%
Win POT 49.35
Place POT 20.74%
Ave win Divi = 3.3
Ave. place Divi = 1.44

This is significant improvement in SR and POT yet the Ave divi remained almost constant

[ This Message was edited by: partypooper on 2004-06-29 17:56 ]

Partypooper,

Could be we are on the same track, as the strike rate and POT skyrockets in some systems when you restrict the number of days since last start and filter by losing margin.

I.E. The shorter the winning margin, the less days between runs the better the results.

Partypooper, I find days since last start vitaly important, with 21 days being my cutoff - does hurt me on the odd occasion, like on the 25/06 when Town Watcher got up at the Gold Coast paying $22.00 :sad: .

However I find days since last start really begins to shine when related to distance.


[ This Message was edited by: davez on 2004-06-30 10:18 ]

Davez, yes that's what surprised me when I applied this 21 day rule and found virtually no difference at all to S/R, so maybe I have just got a freak set of results here, or not enough results.
Problem is of course when applying filters is the enormous amount of possible combinations, i.e. if you have say 25 fields in the data base you could apply any of none, 1,2, 3, upto 25 can't be bothered doing the maths but its a lot.

partypooper,

Aah the age old problem......

"What increases the profit or strike rate, is it the filter you applied or something else contained within or indirectly related to that filter?"

I look for a stable pattern when applying a filter.

E.G. if decreasing the number of days since last start, you should see a continuing trend of increasing strike rate, not huge increases and drops as you shorten the number of days, if you do it may be something else which is increasing the result.

The problem is that the strike rate may increase, but you lose all value.

When I find an increase in both, then I investigate further.

I know the discussion re. barriers but also I remember a graph on these pages here somewhere showing a steady decline in SR outside barrier 6 due to number of runners obviously [I think it was 6 ].

However, if applied as a general rule, barrier should improve SR but I'm not sure how it would affect value [?].

Also, I have found 14 days since last start a dramatic improvement over 21 day. Same problem regarding value again I suppose.

Cheers.

[ This Message was edited by: crash on 2004-07-02 09:21 ]

Crash, agree entirely normally, that's what's got me bamboozled this time as these particular set of results shows hardly any difference is average divi no matter what filters are applied, only the POT alters???

[ This Message was edited by: partypooper on 2004-07-02 18:33 ]

Considering most known filters seem to have been throughly crunched, perhaps a variable filter needs to be thrown in to the pot ?

Many systems fall down due to a wild card.
The wild card I'm referring to is race placement of horses by trainers. Either stupidly, or deliberately.

Often a horse picked to win from a system based on past runs/earnings etc., is placed in a race for other reasons than to win, or the trainer is being over-optimistic or just plain dumb.

A fitness run for a different distance, or to reduce weight penalty for a future race it will be placed to win in, are some common reasons.

Unsuitable [meaning as a winning chance] distance/conditions or class etc. are used regularly for these purposes.

I had three runners for a system yesterday and 2 were ruled out for the above reasons. Both were unplaced and the one I backed payed $6.90 for a win [immediate system SR and average divvy improvement].

Apart from when a Jockey is told to run a bad race, we should be able to recognize unsuitable races for a systems pick and as a final system filter, rule the horse out as a bet. Sure, some might salute, but most won't and this should make a big difference to a systems bottom line.

Often, a purely mechanical systems greatest limiting factor is the user, unable or unwilling to throw in a bit of basic reasoning as a final filter.

A simple example: A system throws up a pick that is running at a track that it has started on 6 times for no placings, or has a similar result over a certain distance/track condition.

Bet the animal because the system says so and well who knows, it might win [and so might the 150/1 shot in the race too] ?

I don't think so.

I know such a variable filter can't be measured and used to calculate it's effects on a system, but heck, I'd rather count [money] than measure, any day !

Cheers.



[ This Message was edited by: crash on 2004-07-03 06:23 ]