Framing Dog Market
 

I was looking at http://forums.ozmium.com.au/showthread.php?t=6234
and I noticed 2 different methods to get odds for dogs based on average time. One was a rating method and the second was using monte carlo analysis. I decided to make an automated excel spreadsheet which when after entering the dog's names in a race. I click a button and will grab the career data of every dog in the race from GRV and give me a dogs career avg mps (Meters per second) and the standard deviation. Then I added another function which would simulate 20000 races and the results would appear on the sheet in about 2 minutes. This would give me a second set of odds and I believe that the monte carlo analysis is more accurate than the 1st rating method using 100*(0.94^hundreths of a second behind winner).

I created the spreadsheet inside 2 hours and if anyone is interested just e-mail me with a request.

Hi kennedy,

I'd love to take a look at it thankyou.

racestats at hotmail dot com

Cheers.

would love a copy

please send to aussiestable @ yahoo dot com dot au

herden10@beagle.com.au
 

I would love a look at the system
garbo10

Count me in!

thebookie @ hotmail.com

Please

onesouthlandman @ hotmail.com

Ok i have sent to the people who have requested it. I sent it to all at the same time so check in the bulk, spam or junk inboxes just in case thats where it ended up.

Thanks Kennedy much appreciated.

Very nice work.

Would also like a copy.
'good4th@hotmail.com'

Thanks for this as it looks very useful.

I have only started investigating the possibilities of how to use this but if anyone does come up with a good staking plan or finds it useful for exacta, quinellas, trifectas, etc, please post it.

I will start playing with this tonight ... thanks again.

Kennedy,

Just want to say love the software and thanks for your effort.

Just framing up the dog markets for tonights races at Albion Park.

I noticed in race 4 that Number 3 Ellies Diamond was given a chance of 6.38%. I looked at Ellies stats and found it has won 8 from 14 and its time is constantly improving.

I then looked at the fav for the race (based on the spreadsheet) and found it had a chance of 40.87% even though it hasn't raced to its best for quite some time (recent form is 68652). I then checked the racing speeds of both runners and found that Number 4 Zinger raced very well when he was much younger (2 years ago).

So based on the above I assume that the anslysis is done on all the races of a dogs career ? Would it be better to just use the last 10 races, or something similar (allow the user to choose it)? Is there a way you could program in the decrease in speed over the life time and plot it to determine if the dog is improving, steady or decreasing in MPS ?

Again, just making sure you know I am not critising the software as I love it. Saves me a lot of work. Just posting some suggestions for enhancements.

Thanks

Umm the data is from GRV. This means that it might not have data for dogs from other states.

I will look into it though and give you an answer.

Well its nice to see that GRV have opened up their form for all tracks in australia so it can be used for any track i think. As for your query you make a very valid point. Zinger in albion park r4 started its career well and has gradually gone downhill since then. For the time being just use the career mps but i will definitely consider changing it so you can decide how many previous starts you want to go back to.

I should point out how does it go over the monte carlo analysis?
I ask this because statistically speaking it is more accurate than the 1st method

Something to watch out for ....
 

Kennedy,

Thanks for the prompt reply ...

One more little thing which I know you can't fix and it is more of something to watch out for.

Race 6 at Albion Park Goliath is in Box 1. When you press "Get Form" it actually retrieves results for 2 dogs ... Goliath that ran in 1990 and Goliath who ran in 2005.

Obviously the dog has the same name and has not been racing for 16 years.

Any good info you can add here on tips/tricks on how to use your software would be great.

Thanks again.

Unfortunately i cannot fix that, out of my control.

I can say the best tip i can provide is simply to use the monte carlo odds and not the 1st method. I leave the 1st method on the sheet to provide odds based on one time. The monte carlo analysis will give a dog a time determined by normal distribution and a random number so the results can vary.

Just use the monte carlo odds and look for good value on the tote.

Kennedy,

I am giving your program a bit of a work out today but I hope I am providing some insights into what I like/dislike about it ( mostly love it ).

For Albion Park Race 8 Number 3 should be by far the favourite along with Number 5 and Number 7. But it has Number 1 as the fav.

I checked the data again ( cause its fun ) and found the reason number 1 was the fav was because it ran a lot of meetings only 366m rather than the current race which was 520m. This meant that it had a much better MPS as it did not get tired. But over 520 it runs only 16.3-16.6 times as opposed to its 366 which is 17.8-17.9 times.

Somethings to think about in the future are :
- Test only over the specified distance OR
- work out a standard drop in the distance for each increased length.

Thanks again .. still love the software and I am just modifying the data to only use the data I want tested again . For example I am removing the records for the 366 races and recalculating the MC analysis.

Now to see how the results go....

Thank You

Id love a copy, evan_c14 at hotmail.com.

cheers

modification ... success ? ?
 

Kennedy,

I have updated the program slightly. Would you like a copy ? ( you could just as easy make the change yourself it was pretty simple ).

I have changed it to only get the past results for the current distance being run. Primarily I don't want to bet on dogs who haven't run this distance before.

So Far I have run it on a couple of races and it has picked the top dog every time. It also picked the trifecta.

This is a top little simple program you have written.

Thanks

can i get a copy thanx
ccisdabest@yahoo.com.au
or is it too late?

Yeah send me a copy of it

kennedy22384@hotmail.com

I will try to change it but not for that exact distance but in 3 different groups, sprint, middle distance and stayers

Kennedy,

I changed the URL to have the following modification ( rather than send you the spreadsheet).

DIST=ALL&

is now

DIST="&Cells(12,1)&"

It was a simple change. Just testing the both against each other and there isn't much difference.. not enough to worry about anyway, as your version picked the same dogs anyway on most races.

I decided as my own judgement to only include races up to 100m from todays distance on either side

Hi, from my own studies you will have to create a database of track to track adjustments, and then try an quanitfy the difference in Classes from one track to another.
Then you will be able to move onto partioning data based on distances.

I think is see what your getting at. Use one track & distance as a standard and use other tracks and distances and divide it by the standard to get a rating for that track. Then multiply the rating by the dogs MPS to get a scaled MPS

can someone then help me acheive this. Someone provide me with a list of many greyhound names and I will automate the process. Then I will come up with the track ratings for each track and rating. I will use Sandown 515m as the standard and so it will be rated at 1.00

After Stdev the next important function is Linear regression, if you seperate tracks by their physical layouts then group by distances then using the Slope an Intercept function within excel you will start to get a feel on the adjustments.

Then the next step is to analysis individual dogs performances from one track to another an quantify the the time differences do this over an over an you will start to get a better feel re the adjustment.

How would i go about this

Can you give me formulas or information about this as such

Some results...
 

All,

Here are some results from today :

Cessnock: 17 units bet, return of 35.7 units. Profit 110%
Bendigo : 8 units bet, return of 19.2 units. Profit 140%

I know that 25 bets is not enough to go off but I have only started testing this.
So far so good.

Good Luck

You know i have just realised if I scale each track and distance to a specific rating then there is no need to only pick races that are the same distance as all races will be scaled to today's race.

Thanks for the email Kennedy.

All very interesting...

Ok i got the data i need to scale the dog's MPS by track and distance but it will not be easy and i will update on progress.

data
 

Once you have sufficent data, reprocess each dog's data take particular note of the correlation in the moving avg of each dogs stdev and the number of starts. This variation will give you an idea as to how dogs progress thru their racing career the same applies to horses.

After making the track adjustments to the data the LTD Stdev will become tighter or smaller.

Then your next step is to breakdown each dogs data by track, distance, Box etc and make slight adjustments to the LTD Mean an LTD Stdev based on sufficent data for each dog or by General analysis of the data.
Then run your MCsims once again, your next goal is to rank the dogs and make sure the MCsim probabilities Vs Actual stats are smooth, you don't need them to match the actuals just to have the same variations thru the probability bins

Once you have acheived the smallest variation possible your finished with doing anymore Handicapping...

I got my data and am now doing the odds by scaling the dog's mps by the track, distance and the box they were running from. The standard deviation is much smaller now. But the data I have will only suit victorian tracks, which isnt a worry for me cause victoria has many meetings (at least one per day).

I tried the slope and intercept but I also wanted to see the r squared that it would produce. The results were not good, r square was very low which mean the line does not fit very well so it is of no use.

I know what you mean with the coefficient correlation, try testing each tracks distance combo's, I'm assuming your using Track records as the Fixed point of reference for each track and distance, which isn't the best, but its a good starting point.
You will quickly find distances where they may have been clocking something other then the dogs.

WOOF43,are you rich?
I don't got a clue what your talking about.
It is a foreign language as far as I am concerned.
You must be rich!

You will quickly find distances where they may have been clocking something other then the dogs.[/QUOTE]
Woof,
No one on here will believe you that they sometimes clock the bunny, too many sceptics.

Moeee,
Check on how many runners Gilbert has now and how they finish.

Cheers.