Pre-post favs.

[partypooper]

Re kindling an old topic I know, but I still have an insatiable interest in this subject.
So conceding that it doesn't matter if one swaps and changes the publication used to determine the pre-post fav. (can't really belive that but still) I was wondering if any readers are also interested in trying to refine a basic selection method, loosely based on the "hot pots method)
So say rule one
(1) Pre-post fav. pre-requisite, WON last race.
(2) last race within 21 days.
(3) pre-requisite , a win and a placing (i.e.1,2, or 3) in at least 2 of last 3 runs.
(4) last run sat. Metro only
(5) last run within 100m of todays distance.
OK so anyone game to carry on from there with rule 6????? etc

[darkydog2002]

USING WIZARD .WT.RATING 91-100
TIME RATING 92-100.

FIELD STRENGTH -20 TO +1.5

[Shaun]

I use this system for placeing but i changed rule 3 so it is must have won last start and be placed the start before that.

[partypooper]

Ok so revised hot pots selection method now reads:

(1) Pre-post fav. pre-requisite, WON last race.
(2) last race within 21 days.
(3) pre-requisite , a win and a placing in previous run(i.e.1,2, or 3)
(4) last run sat. Metro only
(5) last run within 100m of todays distance.
(6) USING WIZARD .WT.RATING 91-100
TIME RATING 92-100.
FIELD STRENGTH -20 TO +1.5

OK so anyone game to carry on from there with rule 7????? etc, also anyone able to check the stats with what we have so far????

[partypooper]

SORRY THAT SHOULD READ

Ok so revised hot pots selection method now reads:

(1) Pre-post fav. pre-requisite, WON last race.
(2) last race within 21 days.
(3) pre-requisite , a win last start and a placing in previous run(i.e.1,2, or 3)
(4) last run sat. Metro only
(5) last run within 100m of todays distance.
(6) USING WIZARD .WT.RATING 91-100
TIME RATING 92-100.
FIELD STRENGTH -20 TO +1.5

OK so anyone game to carry on from there with rule 7????? etc, also anyone able to check the stats with what we have so far????

[partypooper]

ok, what about rule (7) Delete all 2yo, & jumpers & races for Fillies and Mares only.

Anyone like to write rule 8????

How about:
Distance range 1200/2000metres only
Drawn barriers 1 to 8 inclusive
TAB nos. 1 to 6 only
Price range evens to 2/1 only
Not going up or down in weight more than one and a half kgs. either way
No claiming apprentices
Just some of my thoughts
cheers

Just a couple more thoughts,and I'll leave you to it.

No first uppers,unless win strike rate is at least 50%
Minimum career win strike rate of at least 25%
Minimum career starts 10
Must be in the top 5 ranked average prizemoney

[puntz]

QUESTION
Interesting list of variables. To go through all these listings,what method is used due to time constaraints?

[darkydog2002]

LOOKS PRETTY GOOD TO ME EXCEPT FOR THE 1/1-2/1 RULE.I WOULD SCRUB THAT.

[Shaun]

The time it takes to check these out is very little in deed....the main selection criteria can be found on the race lists those being

1)must be the fav
2)must have won last start and be placed at the 2nd last start

[osulldj]

Following are the results for a series of rules similar to what you have all suggested.

First of all the rules:

RACES RULES:
QLD, NSW, VIC, SA Saturday Metro races, Excluding Hurdles & Steeples and ALL 2YO RACES (very poor value).

HORSE RULES:
Must be prepost favourite
TAB between 1 and 5
Must have won last start within 21 days
Last start must have been a Sat Metro run
Within 100m +/- of todays distance
Staring price must have been 1/10 to 11/2


These rules have produced a 37.4% strike rate and 18.6% Profit from March 99 to Jan 03....425 bets.

I am not a believer in systems for all sorts of reasons, but for those that are believers I hope the above is of interest.

[partypooper]

osulldj, (and all who have contributed)

mmmmmmmm, interesting, the question is does the application of further rules as suggested increase or decrease the POT?????
Also osulldj, did you try this criteria for the place, and are the divies from TAB? if so does the POT increase with bookies price and?or BEST bookies price etc?

Some very interesting statistics I have only just noticed of late.The favourites in races 1 to 4 inclusive have a much better strike rate than the favs in the second half of the program.Here are some results,albeit from last Saturday& Monday only,and only using the major meetings.
Saturday.
8 winners from 20 races.
Monday.
8 winners from 20 races.
16 winners out of 40 races.
Strike rate 40%
Races 5 to 8(inclusive)
Saturday.
5 winners from 20 races
Monday.
4 winners from 20 races.
9 winners from 40 races.
Strike rate a touch over 20%

I think this bears further investigation.It would seem logical(to me anyway)that the favs would be far more successful in the earlier races,mainly because of the lower class factor.Usually a field can be narrowed down to perhaps only a couple of really strong chances,whereas in the latter better class races there may be 5 or 6 strong contenders rated within a length or so of each other,so whilst the winner may not come as a surprise to the form students or ratings men,and it may be in the market,it will invariably not start favourite.How of the general public will race for the phone when one of those clowns on skychannel says "the money has come for this one?"
I think if we can establish a nucleus of criteria,or a basic set of rules to get rid of the "false favs",perhaps we could lift the strike rate to 50/60%,operating on the first four races on each program across the mainland states.Any thoughts on this idea guys,or have I finally lost my marbles?
Osulldj,could you run some figures through your database on this idea to see if the strike rate holds up,say over the last 3 years?
Cheers

[becareful]

Angel416,

I am not at my computer with my database on it at the moment but will have a quick look later today if I get time. Initial thoughts are probably that the average div for favourite in those early races will be much lower than in the later races - so if you have a 40% SR with $2.00 div you get a 20% LOT, 20%SR with $4.00 div gives the same LOT. As I have said many times in the past Strike Rate by itself means nothing - you must look at it in combination with the dividend you are getting.

Anyway will have a look later today.

Becareful,added up the divvies of the 16 winners,divided by 16,average dividend $2.57.
Perhaps by culling the pretenders and lifting the strike rate to approx. 50%,and maybe using Malcom Knowles "Power of Ten" staking method a good quid could be made!!Will be most interested to see what figures you come up with over the long term.By the way,did you ever remember to run that doubles method thru your database,or did you not recieve my email????
Thanks for your help mate
cheers

[becareful]

Angel416,

I ran it through yesterday (sorry for the delay) and emailed the results to you. I will resend the email when I get home in case it didn't get through. Sorry to say the results were probably not what you were looking for.

[osulldj]

Partypooper,

The historic results are based on NSW TAB dividends.

My own betting experience says that you can increase your POT by approximately 15% over SP prices through use Betfair / Best Bookies price.

The average odds of my winners since the launch of betfair has been 24% better than SP...which again at the lower price range is much better than the TAB.

You could play around all day with different rules that positively indfluence the results. The trouble is that it doesn't mean anything going forward...it's all just backfitting based on statistical boundaries that are never consistent.

For example, I used the rule of SP from last run because it is a very powerful indicator of a horses future performance. There is a clear pattern in strike rate based on the SP of each horse at its last start. On average horses that started 1/1 last start win more this start than horses that start at 5/1 last start. If I get time I might post some figures on this.

Anyway, back to my point...I used an SP rule for that reason and selected the limit of 11/2...why? Not because there is something magical about that price that influences a horses chance...purely because horses at 6/1 lost a significant amount.

Horses at 8/1 actually won a little...horses at 10/1 lost. You have to draw the line somewhere.

Those boundaries aren't consistent from one record set to another.

This is the whole problem with the concept of developing a mechanical system that can produce profits each year. I can develop a system that shows profits for the last 4 years...but it's all based on rules that happen to fit the statistical boundaries for those particular records. It's not valid to assume those statistical boundaries will remain in future record sets, that's why systems fall over. Just look at PPM's.

If you are into system development or any form of automated rating process, the most valid approach to take is to split your available historical data into two record sets. Develop your approach on one set of data, without even considering the other. Get it to a point you are happy with and then witout further refinement run it over the unseen data and watch in amazement.

If the results hold up or improve then you may have something of worth. It's highly likely though that if you are honest with your use of the seen and in particular unseen data the results won't hold up.
The temptation is then to make a few adjustments including the unseen data but then you are back to square one.


The harsh reality is that mechanical approaches are not winners for any reasonable length of time of real betting on future, unseen races.

If there was, why would professionals all across the world spend so much time and effort on form study, ratings, pricing etc.

I'll get off my soap box now :smile:

P.S. I forgot to mention one thing...with the rules I posted, if you change the distance factor from -100 to 100 last start to "-1000 to 99" (i.e. horses going down in distance or up in distance but less than 100m) then the strike rate becomes 38.9% and the POT 26.9%. The number of bets reduce to 385. There must be something magical about that extra 1m hey? :smile:

<font size=-1>[ This Message was edited by: osulldj on 2003-05-21 12:04 ]</font>

[partypooper]

osulldj, thanks for all that, and please step back onto the soap box, all good stuff!
My aim was always to produce a QUICK EASY SELECTION METHOD, with not too many rules, as I take your point about influencing PAST results, or making em fit shall we say. I find that going down that path just reduces the amount of qualifiers anyway, and if you keep going there is 2 bets in 6 months or something similar. Somebody once told me that if you take the pre-post fav, (last start winner) and apply one rule only i.e. it is a bet if riden by one of the top 3 jockeys for that meeting, (as per table)that this will produce a profit consistently, I've never been in a position to check this out but it did spark my interest in pre-post favs.
But I'll have to admit I've never considered
the SP from the last run (in this concept) VERY interesting. My thoughts were originally to start with the basic HOT POTS rules, and say maybe increase the list to say 10 rules (using the best from other interetsed parties) not sure if this will materialise but still, it has sparked some interesting discussions.

If you really want to get fussy I would add in.
A) Must be racing at own track
B) Must have good jockey
C) Must have won at that distance

Just a couple of suggestions which might help with only 14 runners or less.
(sorry if I repeated over somebodys idea already printed)

[partypooper]

osulldj

Just a thought, do you think it makes sense, when one thinks of a plan to work it over a couple of years past results, i.e. say use 1999 & 2000 whilst working on the idea, then when satisfied, run through 2002 & 2003 (eg. only) and see how it fares?

[osulldj]

On 2003-05-23 00:03, partypooper wrote:
osulldj

Just a thought, do you think it makes sense, when one thinks of a plan to work it over a couple of years past results, i.e. say use 1999 & 2000 whilst working on the idea, then when satisfied, run through 2002 & 2003 (eg. only) and see how it fares?


Hi Partypooper,

Yes that is the only sound approach to take. You need to work with what is called a "seen" set of data to develop your method and then test it over "unseen" data without change and see how it fares. When most do this they are confronted by the harsh reality of mechanical selection methods.

As I said in my post above

"If you are into system development or any form of automated rating process, the most valid approach to take is to split your available historical data into two record sets. Develop your approach on one set of data, without even considering the other. Get it to a point you are happy with and then witout further refinement run it over the unseen data and watch in amazement.

If the results hold up or improve then you may have something of worth. It's highly likely though that if you are honest with your use of the seen and in particular unseen data the results won't hold up.
The temptation is then to make a few adjustments including the unseen data but then you are back to square one. "