Dutching – A Strategy – Stage 2

I have completed stage 2 of my analysis on how to dutch the top three rated horses.

Once again, thank you all for your suggestions and advice.

My Basic spreadsheet may be found at http://focusratings.com/data/dutching.ods

I'm sorry that this is in OpenOffice/LibreOffice format - it's too big to save in Excel format (although you may be able to open it with Excel.)

Now, to recap...

I am only using data from non-handicap races.

I am only using data from the 19th of November 2013 to the 18th of November 2014 (inclusive.)

I have stripped out a few races where there are silly errors (races called NH Hurdles but run over the AW course at Southwell - due to bad weather - and so forth.)

My original post may be found at Dutching - A Strategy - Stage 1

Class

So, I decided to take a look at how the results varied according to class.

The results are as follows...

Class	Races	Wins	Strike Rate
Class 1	1352	699	51.70%
Class 2	526	277	52.66%
Class 3	463	296	63.93%
Class 4	2420	1472	60.83%
Class 5	3435	1989	57.90%
Class 6	1660	938	56.61%
Unknown	4186	2195	52.44

No big surprises although it seems tat we should concentrate on Class 3 and Class 4 races.

RC1

What is RC1?

RC1 is an abbreviation of Race Coefficient 1.

RC1 starts off as the percentage of horses in the race that I can actually rate.

I then subtract from this, 0.01127 times the total number of runners.

The reason I do that is that the accuracy of the ratings decreases as the number of horses in the race increases.

The 0.01127 number is recalculated, on a monthly basis, as the accuracy of the ratings increases over time.

And finally, if the confidence level of the top rated horse is 200% (the maximum), I then increase RC1 by 5% as the ratings are about 5% more accurate when that is the case.

How to use it...

So, this seems a good thing to look at.

I rounded RC1 to the nearest 5% and got the following results.

RC1	Races	Wins	Strike Rate
0%	680	392	57.65%
10%	3	2	66.67%
15%	21	8	38.10%
20%	30	16	53.33%
25%	123	46	37.40%
30%	345	146	42.32%
35%	767	371	48.37%
40%	602	271	45.02%
45%	922	484	52.49%
50%	901	470	52.16%
55%	1001	566	56.54%
60%	777	442	56.89%
65%	813	485	59.66%
70%	958	560	58.46%
75%	789	517	65.53%
80%	1389	780	56.16%
85%	427	159	37.24%
90%	1853	993	53.59%
95%	1573	1103	70.12%
100%	66	53	80.30%
105%	2	2	100%

Now, it's no good having a good strike rate unless you have the races to use it on.

So, the following graph shows the distribution of races against RC1.

Please bear in mind that the number of races shown is actually 3 times the number of actual races as we have three bets in each race.

The next graph shows the Strike Rate against RC1.

As you can see, the sweet spot seems to be when RC1 is higher than 92.5% as we get a good strike rate and plenty of races.

Conclusion

I've got plenty to think about here and so that's what I'm now going to do.

The next stage of my investigation will be a sideways look at what happens when the top three rated horses are substantially separated from the other horses (in terms of R4.)

This means that I'll be looking at the difference in confidence between the 3rd rated horse and the 4th rated horse.

And, for that, I'll have to write some code to generate the data.

After that I'm going perform a similar exercise but, this time, using the forecast starting price as the criteria.

In other words, if my forecast starting price shows the top rated horse to be evens, the second rated horse to be 2/1, the third rated horse to be 3/1 and...

The fourth rated horse to be 10/1.

Then I'll try to put things together to come up with some sort of testable strategy.

Other things may come up along the way but, I anticipate that, after that, I'll have to perform a similar exercise for Handicap races.

Oh, one last thing - there will be a detour to look at Irish races as more than one of you has mentioned that there might be an angle there.

Once again, if you've got anything to say about Focus Ratings, please wander over to Focus Ratings Review and have your say.

As always...

My kindest regards

One Response to “Dutching – A Strategy – Stage 2”

1. TiggersTaxi says:

   December 7, 2014 at 11:43 pm

   Just run the scenario through the Non Handicap data set
   Which gave 122 races of which 102 produced a dutched winner.
   The 1st pick averaged odds of 3.14, the 2nd 4.03,and the third pick averaged 9.33 (118 races).
   Dutching to get a £100 return would require £67.38 to be staked
   (100/3.14)+(100/4.33)+(100/9.33)on average. 122×67.38= £8220 – the returns from Non Runners.
   Returns =102 x £100 = £10,200 a nice 24% profit over the period

   Log in to Reply