Hermes,

Trust and Betrayal was one of my winners from yesterday but through a different selection procedure! Paid $10.80 on NSW :smile:

Might have to have a closer look at the QTab ratings - I do store them in my database but have not been able to come up with a reliable system based on them. Might try some of your suggestions over a longer period and see what comes up.

__________________
"Computers can do that????" - Homer Simpson

Keep up the good work Hermes.

There`s some very interesting observations being shared here ,great for those system research pundits out there, to research further & possible refine.

__________________
Cheers.

Hermes (and anyone else interested):

I did some analysis on my database using the QTab ratings and thought you may be interested in the results. I have the QTab ratings for all races run this year (since 9/1/02) so I looked at all races since that date (quite a few races - I need a faster PC as it took some time for each query!).

First I just looked at 96-100 raters that started at $10 or more which gave 6094 qualifiers for total divs of 5291 (87% return). I then tried various price ranges (eg. $10-$20, over $20, $8-$12, etc) which gave various results but nothing profitable. Restricting the selection to 99 or 100 raters made a very small profit (about 1% POT).

I then tried looking at the data by the day of the week and found that weekdays were better than weekends - Sunday was particularly bad. So I cut out weekends but was still very marginal. I then looked at it by the race state and track condition and finally came up with some profitable figures.

The best criteria I came up with was:

99 or 100 raters only
Approx div at jump between $11 and $20
Ignore ACT, NT, TAS & overseas meetings
Weekday meetings only
Only bet on Good or Fast tracks

This gave the following results based on which TAB you use for the selection price and dividend:
TABCORP: 351 qualifiers paid $559 (59% POT)
NSW: 325 qualifiers paid $556 (71% POT)
QLD: 342 qualifiers paid $451 (32% POT)

Including the 96, 97 & 98 raters still resulted in a profit (with the other conditions) but POT is significantly lower so wouldn't really recommend it.

Applying the same criteria to Saturday meetings gave a small profit on NSW (around 15% POT) and break even on Tabcorp/Unitab.

So there you have it - a very simple but profitable system based on Qld rating. Can't say whether results will hold up in the future but it could be worth a shot!

PS. This is only for gallops - Harness racing shows a marginal loss with the same criteria and Greyhounds a big loss.

__________________
"Computers can do that????" - Homer Simpson

Very interesting results there!
Use the QTAB ratings for free to return a profit. Nice little earner and provided you use filters it seems to work on a steady basis. Becareful, I would say that the sample you used was broad enough to make a good judgement providing QTAB doesn't change anything in it's method of ratings - without telling anyone. That could be a disaster!

Would be interested to know what the average price and strike rates are.

[ This Message was edited by: Equine Investor on 2002-08-24 22:41 ]

Great stuff, becareful. Can't argue with a database like that. You wrote:

I need a faster PC as it took some time for each query!).

I'm doing it by pencil and green highliter! Gotta get me a program that sucks html data off the TAB site!

In your initial sample - 6094 qualifiers for total divs of 5291 (87% return) - you don't distinguish between runners per race? So if there are three qualifiers in a race, three bets? I don't suppose there's a quick and handy way to cut it down to one per race in your database, eg. lowest TAB number or best last start. If the sample counts all qualifiers, even several per race, then 87-90% return isn't too bad, I think. In many races there are two or three or more $10 jobs rated 95+. I wonder what proportion of races have two or more qualifiers? If ALL qualifiers net around 90% return there's hope of finding a mechanism to find the best of them where there's two or more in a race.

In my (midget) sample I found weekday country meetings best, metro Sat. meets worse.

You wrote:

Restricting the selection to 99 or 100 raters
made a very small profit (about 1% POT).

Again, if that includes scenarios of two or more qualifiers in a race, that's not too bad.


You wrote:

Including the 96, 97 & 98 raters still resulted in a profit (with the other conditions) but POT is significantly lower so wouldn't really recommend it.

How significant is significant? I'm wondering about strike/out patterns. Personally, I'Ii accept a lower POT if a system is more stable with shorter runs of outs and more happy evenings with that winning feeling.

Your findings on good and fast tracks was surprising, but there it is.

Your findings on greyhounds was very surprising and have saved me heaps of work and probably a few bucks as well.:smile:

Among those rated over 95 starting at $10 today I snared Maron Chevel ($13.90) and Idealistic ($18.50). Marstic, Calming and Senate Lease good placegetters. Ahead on the day.

Great analysis. Many thanks becareful.

Hermes

There must be some sub-clause of Murphy's Law that says when you start taking samples from old races you get spectacularly good results which then get worse from there. This ensures that people like myself waste countless nights chasing false hopes.

I had another idea (based on some statistical hunches): - selection is any horse rated exactly 3 points below the favourite on the QTAB. I thought it might follow a hole in the figures.

So I started a sample, turning first to Aug 21, 2001. Five qualifying races that day. First race - $1.40 placegetter. Ok. Second race - $34.50 winner. Third race - $8.10 winner. Fourth race - $4.20 winner. Fifth race - $5.80 winner!!! Amazing!

So I moved on and took a stab at 22nd July 2001 - Sixth race - $35.20 winner. Race seven - $2.90 placegetter. Race eight - $12.70 winner. Race nine - out!

No kidding. Amazing run.

Since then my figures have settled down considerably. I don't think it will pay in the long run.

Hermes

Now that's dedication hermes!
I find it time consuming enough doing it by computer and excel etc. Manual entries would take you forever.

Hermes,

No I didn't distinguish between runners per race with any of the analysis so if there were 2 qualifiers in a single race I assumed that we backed both of them. Personally I don't think there is a problem with backing 2 or even 3 $10+ runners in a race (different if you are looking at $5 shots) and trying to eliminate runners can get rid of just as many winners as losers.

On NSW figures with 99/100 raters we had 325 selections for 556 in divs. If we include the 96,97,98 rated runners we get an additional 878 selections (so 1203 in total) and an additional 965 divs (so 1511 in total) so the POT on the 96,97,98 selections is about 10%. Not too bad but way below the 71% POT on the 99/100 raters. Again this is with no elimination of double/triple selections.

I don't know how the QTab ratings are calculated so I don't know if the track condition filter is surprising or not. If the ratings do not consider track condition then it is probably not suprising they are less accurate on wet tracks? Personally my own system works better on Good/Fast tracks (although it also works fine on Dead but not very good on Slow/Heavy) so it seemed a reasonable filter to me!

Just had a look at the "selections" for last couple of weeks based on 99/100.

For last week there were 13 selections but no winners.

For week beginning 12/8 we had 18 selections (0 Mon, 1 Tue, 8 Wed, 3 Thu, 6 Fri) with 4 winners. There was 1 double selection (SR8 on 14/8 - 7 Quadri rated 100 and 10 Think On rated 99. Think On won so eliminating based on TAB No or lower rating would have got rid of the wrong one). The winners were:
14/8 SR8 10(Think On) - $12.00
15/8 BR3 10(New Method) - $18.50
16/8 BR3 8 (Toobanna) - $11.50
16/8 BR8 2 (Credit Limit) - $17.40

The previous week (5/8) we had 15 selections and 3 winners. There were 2 double selections and in both cases one of the selections won! The winners were:
5/8 MR2 4 (Limerick Belle) - $11.70
7/8 BR1 10 (Kananga) - $13.00 (also 6 Diamond Zoff)
7/8 WR7 5 (Great Fingall) - $12.10 (also 3 Mikis Prospector)

_________________
"So certain are you. Always with you it cannot be done. Hear you nothing that I say?" - Yoda

[ This Message was edited by: becareful on 2002-08-25 10:29 ]

Sorry EI - missed your question when I read the thread this morning. Strike rate/ave div for the 99/100 raters is as follows:

NSW: 39 winners, 12.0% SR, $14.30 Ave Div
Tabcorp: 40 winners, 11.4% SR, $14.00 Ave Div
Qld: 33 winners, 9.6% SR, $13.70 Ave Div

__________________
"Computers can do that????" - Homer Simpson

Thanks becareful,
Seems the strike rate is not as low as I would have thought.