#7 – Green is good…..Yes?? – Part 1

Hello, it’s been a while since I have had a chance to sit down and download my thoughts to paper: House moves etc are now complete and almost all the moving boxes are empty.

Several additional features have been added to The Bet Machine since my last blog so I want to have a quick look at how you can use them to narrow down to a profitable system.

The main change has been to the data presented in the ‘Additional Breakdown’ screen as you can see on the right. 

Further to the above additional criteria filters have been added. These are Claiming, DSLR (Days Since Last Race) and WeightLBs. However today I will focus on the Additional Breakdowns.

For this example I’ll choose data from a single course to try to highlight some of the things to look out for and interpret; Firstly I’ll use data from Southwell with no filters as below :

As you’d expect following all selections in every race at 1pt results in a loss but hopefully by applying some filters to the data we can pull some profit.

So clicking on will open a new window.

NB. How you configure this section,

will influence the data presented in future steps; If you find a system that appears to give an incredible Strike rate and phenomenal profit just check that you are looking at the markets that you intend to bet in.

For now I will concentrate on ‘Win Single Bets’. Clicking Additional Breakdown button opens a further window defaulting to Course.

For most Courses there will only be 1 line of data but with Southwell we have both Turf and All Weather racing, and if we were looking at a non-course specific system multiple courses would be listed.

So what do these numbers mean, remembering that at this point we have no other filters applied?

All we can deduce really is that there were an average of about 8 to 9 runners per race across both surface disciplines (100% / 11.xx% = 8 to 9-ish).

This is confirmed looking at the RaceRunners remembering that these are declared runners.

So what we are trying to do is find where the value lies within the datasets. It can be a little like a ‘needle in a haystack’ scenario but with a little logic and observation the trends can be seen.

With all these numbers we need to think about what they mean:

Do They;
Define the race conditions
That is things like:
Course, Surface, Class, Code, Type, Age*, Ratings*, Sex*, Handicap/Non, Distance, Going etc

Or Do they
Differentiate between horses
That is things like:
Trainer, Jockey, Age*, Ratings*, Sex*, Form, Stall Draw, DSLR etc

* Several of the criteria can define a race and be a differentiator between contenders

Trying to find a system with a broad range of race conditions can be very difficult so it makes sense to aim at a more defined target.

Most of my activity recently has been to find Course Specific systems so lets start with Southwell (AW)

You can see from the trend line on the mini graph there are about 6 big price winners kicking the trend up slightly.

This is where you can vary the path you take:
You could exclude all data above a certain odds value
Try to find a trend in the ratings etc data

Checking the data show winners starting at 170BSP with large gaps between winners down to 25BSP so for me being the cautious type and wanting to preserve my betting bank from large drawdowns I’ll restrict the odds cap for now. In the image below I have added max odds as 25. 

Scooting through the Additional Breakdown I don’t see any obvious trends in the ratings. For example RPR

There are some reasonable strikes rates and trends (even though negative) in the mid table but I suspect that there are some other differentiated criteria affecting the Profits.

Jumping back to Trainers you can see some good strike rates from reasonable numbers of bets and apparently good profits

Conversely the mid and lower table shows both outright poor performance and poor returns on reasonable strike rates
Mid Table:

Lower Table:

There are many here that either haven’t found the right Horse, Jockey or Training style that suits the Southwell all weather track.

Using the List Generator in TBM and a spreadsheet, I’ve narrowed the Trainers down to those that have

Raced more than 2 entries

Have a strike Rate >=10%

Achieved >5% ROI

These are limits that feel about right for me, you may want to adjust to suit.

Now obviously this is to a degree back fitted data and future performance cannot be guaranteed however choosing fairly wide criteria minimises the back fit.

So now we have a (historical) profit. Lets see if any trends pop up in the profitable Trainers.

Jumping back to RPR we can see some good strike rates mid table, though there is not much opportunity to reduce the range of RPR in the system criteria to achieve a better fit.

Age may give a better lever to pull: Strikes rates are good until we get to 7 year olds. Excluding those older than 6 years reduces the number of bets by >250

Dropping the data into our sister software – The Staking Machine –  this shows a reasonable performance:

Win data

Place Data

There was no plan to the route through the data today: It was completely ‘off the cuff’ in order to demonstrate the basic process. Repeating this another day I might choose another route.

This hopefully highlights several steps in the process of finding systems:

Trends can be hidden, use some logic and a little filtering to try to reveal them.

Don’t go too far! Backfitting is easy.

Find a system then run with it either paper trading or using minimal/Test stakes.

Expect variation: Systems using Trainers/Jockeys can and will alter. The string of horses they have will age/alter/get injured etc. Take time occasionally to review performance of the system.

Research with a subset of the data then apply the rules you find to the whole data set and see if it stands up.

That’s all for now, I need to pack for a short break up in sunny Northumberland and I’ll post a new episode when I return.