Follow @FocusRatings |

I am now ready to start moving forward and define a new strategy based upon the performance of the different race types over the last 28 days.

I have created a spreadsheet and would like to walk you through the process that I shall use to identify a profitable strategy.

The spreadsheet is available at Recent Results Strategy and...

Just to tantalise you (or convince you to read this boring post),

Here are some interesting numbers...

Bet Type |
Races |
Wins |
Strike Rate |
Profit |
POI |

ISP |
299 | 149 | 49.83% | 46.11 | 15.42% |

BSP |
299 | 149 | 49.83% | 66.942 | 22.39% |

I'll start off by explaining the concept.

**The Concept**

Now, we know that Focus Ratings does better with some race types than others.

We can calculate the performance of each race type over the last year (or, indeed, the last three and a half years.)

We can calculate the performance of, lets say, Flat 2yo Maidens and we might come up with a strike rate (for the top rated horse) of 22% (this is a made up number.)

However, I know that we can't expect that sort of strike rate in the first week of April; the 2 year old horses have no real form and not much (if any) experience of racing.

My theory is that, instead of basing our selections on the historical strike rate for that race type, we should just look at the performance of that race type over the last 4 weeks (28 days.)

Taking Flat 2yo Maidens as an example, in the first week of April there probably won't have been many Flat 2yo Maidens run over the previous 28 days and, for those that were run, the strike rate will have been far less than the historical strike rate of 22%

Thus, probably a race type to avoid.

Let me clarify race types...

**Race Types**

Let's say that, on an average day, there are 35 races.

From those races there might be 10 or 12 different race types.

In other words, there may be 3 non handicap Novice Hurdles; there may be 2 handicap chases etc.

I should, at this point, clarify that a non handicap Novice Hurdle is a different race type to a handicap Novice Hurdle as far as the system is concerned.

Now, every day, I calculate 2 new numbers for each horse in every race.

These are...

1). Recent Race Type Performance - this is shown on the spreadsheet as rpp but I'll get to that later.

This equates to the strike rate (for the top rated horse) for that race type over the last 2 days. All horses in a non handicap Novice Hurdle will have the same rpp; all non handicap Novice Hurdle races will have the same rpp.

I then rank the race types in descending order of performance (based upon the rpp - the race type performance over the last 28 days, that is.)

2) Race Performance Rank - this is shown on the spreadsheet as rprank - I'll get to that shortly.

This just shows the ran of the race type in terms of its performance over the last 28 days.

So, as an example...

Lets say that over the last 28 days we have had a 40% strike rate for Hunter Chases, a 32% strike rate for non handicap Novice Hurdles, a 30% strike rate for non handicap AW sellers and a 29% strike rate for AW Handicaps (and so on) this would be ranked as...

Hunter Chase - rprank = 1

Novice Hurdle (non-handicap) - rprank = 2

AW Selling (non-handicap) - rprank = 3

AW Handicap - rprank = 4

And so on.

I have shown, on an earlier post, that performance in the real world varies in line with rprank.

Here is what I said...

Now, bear in mind that we had an average strike rate for all race types (over the entire year 2016) of 25.42%

If we restrict ourselves to those races which are ranked 1 to 3 in terms of their performance over the last 28 days (only taking into account those race types that are running today)

we increase our strike rate to 35.28%Here is a graph showing strike rate for each rank...

You can ignore the silly result for the lowest ranked race type - there are only 3 races and the result is a blip.

In order to prove a strategy we need to look at a spreadsheet...

**The Spreadsheet**

The spreadsheet (which is available at Recent Results Strategy) starts off with just one sheet.

This is the tab named Data.

Data comprises of every top rated horse in every race in 2016.

I have removed all races where the top rated horse was a non runner - this just makes the calculations easier.

The spreadsheet contains the usual columns but has some additional ones at the end (the right hand side - you'll need to scroll over to see it.)

This new column is named rprank and shows the rank of that race type as explained above.

The first analysis that we need to do is to test what happens if we blindly back the top rated horse in every race where the race type has an rprank of 1 to 3.

To do this I shall copy and paste all the data in the Data tab into a new sheet/tab called rprank 1 to 3.

I then delete all races where the top rated horse had an rprank of more than 3.

I now need to do some basic analysis on this new data (just those races where the rprank is 1 to 3.)

**Analysis 1**

I did a pivot table and showed the results on a new tab which is named Analysis 1.

I have hidden some columns to make it easier to read.

The first visible column (column B) shows the number of wins.

The second visible column (column D) shows the average ISP for the winning horses.

The third visible column (column F) shows the average BSP for the winning horses.

The fourth visible column (column G) shows the number of races - bear in mind that races where the top rated horse was a non-runner have already been taken out.

We can see, from the basic analysis, that by blindly backing the top rated horse where the rprank is 1 to 3 is slightly unprofitable.

However, it is slightly less unprofitable to BSP (I have calculated and removed the 5% commission.)

Now is the time to do the same thing but with races where the top rated horse has an rprank of 1 to 4.

**Analysis 2**

I did a pivot table and showed the results on a new tab which is named Analysis 2.

I have hidden some columns to make it easier to read.

The first visible column (column B) shows the number of wins.

The second visible column (column D) shows the average ISP for the winning horses.

The third visible column (column F) shows the average BSP for the winning horses.

The fourth visible column (column G) shows the number of races - bear in mind that races where the top rated horse was a non-runner have already been taken out.

We can see from the basic analysis that the strike rate goes down and we lose even more money.

Thus, the graph (further up this page) was right - we should just concentrate on those races where the rprank of the race is 1 to 3.

**Where are we? - One**

So, at this point, the rules are...

1). Top Rated Horse

2). Rprank of 1 to 3

Our results are...

Bet Type |
Races |
Wins |
Strike Rate |
Profit |
POI |

ISP |
1851 | 653 | 35.28% | -213.13 | -11.51% |

BSP |
1851 | 653 | 35.28% | -108.369 | -5.85% |

Another filter is required to make us some money...

**Confidence**

Now. at this point, I should say that (on another spreadsheet) I did vast amount of analysis to test the effects of things like Race Class, LTO, Number of Runners etc. but...

The only filter that showed some results was the obvious and logical one, my fallback - Confidence Level.

I determined that if we only concentrate on races with a rprank of 1 to 3 and where the top rated horse had a confidence level of 120% and above, we started to see results.

Thus, I took the data shown in second tab (rprank 1 to 3) and put it into a new tab which I named rprank 1 to 3 - conf >= 120%

I then deleted all those races where the top rated horse had a confidence level of less than 120%

**Analysis 3**

I did a pivot table and showed the results on a new tab which is named Analysis 3.

I have hidden some columns to make it easier to read.

The first visible column (column B) shows the number of wins.

The second visible column (column D) shows the average ISP for the winning horses.

The third visible column (column F) shows the average BSP for the winning horses.

The fourth visible column (column G) shows the number of races - bear in mind that races where the top rated horse was a non-runner have already been taken out.

The strike rate has gone up to 40.69% and the losses have reduced.

**Where are we? - Two**

So, at this point, the rules are...

1). Top Rated Horse

2). Rprank of 1 to 3

3). Confidence of top rated horse of 120% or better

Our results are...

Bet Type |
Races |
Wins |
Strike Rate |
Profit |
POI |

ISP |
988 | 402 | 40.69% | -87.99 | -8.91% |

BSP |
988 | 4.2 | 40.69% | -47.863 | -4.84% |

Another filter is required to make us some money...

**RC1**

It was at that point a member sent me an email to suggest another filter (thanks RC) and so I took a look at RC1.

RC basically shows that amount of information that we know about a race (mainly the number of rated runners, although it's a bit more complicated than that.

I did some analysis and determined that the sweet spot is those races with an RC1 of 80% or better.

Thus, I took the data from the tab named rprank 1 to 3 - conf >= 120% and put it into a new tab called rprank 1 to 3 - conf >= 120% - RC1 >= 80%

I then deleted all those races where the race had a RC1 of less than 80%

**Analysis 4**

I did a pivot table and showed the results on a new tab which is named Analysis 4.

I have hidden some columns to make it easier to read.

The second visible column (column D) shows the average ISP for the winning horses.

The third visible column (column F) shows the average BSP for the winning horses.

The fourth visible column (column G) shows the number of races - bear in mind that races where the top rated horse was a non-runner have already been taken out.

The strike rate has gone up to 49.83% and we are making profits.

**Where are we? - Three**

So, at this point, the rules are...

1). Top Rated Horse

2). Rprank of 1 to 3

3). Confidence of top rated horse of 120% or better

4). RC1 is 80% or better

Our results are...

Bet Type |
Races |
Wins |
Strike Rate |
Profit |
POI |

ISP |
299 | 149 | 49.83% | 46.11 | 15.42% |

BSP |
299 | 149 | 49.83% | 66.942 | 22.39% |

**Conclusion**

So, by using just 4 simple and logical filters (one of which - rprank, is a new way of looking at the ratings) we have a system that gives us about a race a day. You need to remember that I've taken out the races where the selection was a non runner; also, there are a few days of the year where there isn't any racing.

We have a strike rate of a smidgen under 50% - that should ensure short losing runs.

We make a 15.42% Profit on Investment if we use the bookies - there is a possibility of doing better if you use BOG (best odds guaranteed.)

We make a 22.39% Profit on Investment if we use Betfair (remember, I have deducted the 5% commission) - there is a possibility of doing better if you can get a bet on at a higher price than BSP.

There are other filters that could be added but they would reduce the number of races whilst potentially increasing profits - I'll leave that to you to implement.

The spreadsheet is available at Recent Results Strategy -please feel free to have a play with it and, if you find that killer missing filter, why not keep it to yourself unless you really want to share it?

I am now going to write the code to produce a PDF to display any selections for this system. I'll probably need to send it out manually for the next few days but, ultimately, there will be link on the morning ratings for this.

I can add this to the builder (the rprank, that is) and display it on the ratings pdfs and CSV files but...

That will take me many, many hours and, as you get the information anyway, pretty much goes to the back of the list.

I'll be automatically reporting on the results of this strategy on the Morning News (just as I report on the First Page Strategy and the Place Betting Strategy)

As always...

My kindest regards

## Leave a Reply

You must be logged in to post a comment.