I am having problem accepting the Harville method for calculating Exactas.

I have 2 animals from the Field I wish to calculate the Chances of the Exacta occuring

#1 is $2.50
#2 is $11

The Exactas work out as 1 - 2 is $16.80
and for 2 - 1 is $25.20

It seems to me that the Chance of the Rougher animal beating the more Favoured animal is too close to the Chances of the Favoured animal beating the Roughie.

If I could calculate each animals Chances of winning the race, and also calculate each animals chances of running second, is there a Formula I could use that would give me a more accurate assessment of the Exacta occuring.

Moeee,
You really need to change the payoff to Natural probability (100%) then take out the track take for which ever exotic.
I have attached a sheet left hand side is the proof side, enter your probs. % in the right hand side of sheet.

Attached Files

harville.xls (22.5 KB, 537 views)

Moee the harville formula is very inefficient when compared to what MCsims can produce, when using the right mix of std dev and distribution

I been playing with My Ratings in the MCSim.
They are figures based on expected finish times for each animal.
The Figures are up around the 90 Mark.
For the Standard Deviation I have been putting my own figure, and it seems to run best when I make it 3.
Thats a 3 for all animals, regardless of how consistent or not the animals actually are.

What annoys me is that some animals turn up as having a lesser chance of running 2nd than they do 1st and 3rd.
It shouldn't be possible that an animal can be 20% Chance of Winning 14% Chance of running second, and 18% Chance of running 3rd.

I'm running the Sim 20,000 times.
Its done by Kennedy91210.

I'm wondering if it might be better if I can store all 20,000 Exacta results in a loop thingo - You know what I mean.
Then just divide the Number of times each permutation occurred into the 20,000 to get the Odds.

I'm sure your meaning .300, anyway each box really has it's own distinct std dev. Box 1 you'd have to agree with it having a tighter std dev say .240.
than say box 4 which has a naturally wider st dev. If you have enough data for a dog out of the same box say more than 10 starts you'd use that data and st dev (as long as no outliers occur). Sometimes you have to make synthetic changes to your data, to how you think the dog will perform.

Kennedys sheet is based on using a normal distribution, which doesn't happen each box for each distance will have it's own basic distribution, the only times that changes is say a wide runner starting in box 1 for example.

Only when dogs and for that matter horses race on their own do you get a normal distribution of times.
I remember conducting many trials of dogs on their own on a track, at that time was a replica of Wentworth park with the help of a great late trainer

That's your basic requirement Mooee, your looking for exact 1-2 finishers for each box, not how many times it ran 2nd or 3rd. Just doing that will improve your ROI many points then any tweaking of your handicapping could ever acheive.

Here's an Example of my Ratings for Warragul Race 6 tonight.

1 FREE DREAM 80.9 $5.6
2 TEN YEARS GONE 82.0 $3.3
3 GO VICTOR 80.9 $5.6
4 COSMIC SNOW 79.6 $10.5
5 FATHER OF MINE 78.0 $22
6 CHEWY'S GIRL 78.5 $18
7 DEMON JIM 79.5 $10.8
8 DR. WOWZER 78.4 $18

The Ratings followed by MY Estimate of the Win Odds.
If I pump those Ratings into a MCSim , I need to use 3 as the Standard Deviation figure to get a similar Market.
What the Simulator does is adds upto 3 and subtracts upto 3 randomly for each dog - I think

Thanks for your help Woof43.
I'll catch you tomorrow when I start being fair Dinkum about Winning on the Greyhounds after massive losses in the past few days.