From a Place, to a Different Place

Before I start, I should apologise for the rather clunky title, but hopefully it will soon make sense, writes Dave Renham. We all know that not all form is equal: a win at Ascot may generally be considered more meritorious than first place at Catterick, for example.

I decided to see if there is any pattern in how form translates from one track to another. To do this, I focused on horses that finished in the first three last time out (123LTO), comparing results between the 'placed run course' and today's track. In other words, I'm trying to find a relationship - positive or negative - between the host track of an ostensibly good (first three finish) run last time and the course next time. Confused? Let's break it down.

The dataset is UK flat racing between 2017 and 2024 with the focus mainly on the turf courses. However, there is some all-weather (AW) course data shared when the other course is a turf one. Put another way, there is no comparison between all-weather tracks. Profit and loss figures have been calculated to BSP less 2% commission on winning bets.

To begin with let's look at the results for all races/courses of horses that finished in the first three LTO:

This is our baseline against which to measure our 123LTO course to today's course results. The win rate overall is around one in every six races with losses of just under 4p in the £ at BSP, which is a fairly solid starting point.

Here are the top 20 win strike rates in terms of today's course to 123LTO course. Here are the numbers (80 runners minimum to qualify):

There are some excellent strike rates here, all well above the 16.4% average figure. 18 of the 20 have seen a BSP profit; and 17 of the 20 have seen an A/E index in excess of 1.00. It is interesting to see that Newmarket has been the LTO course in six of the above, five times horses coming from the Rowley course, once from the July. It is equally noteworthy that very few of the Grade 1 tracks appear in the current course column. I will discuss and dig deeper into this area later in the piece.

Next, here are the 20 lowest win strike rates:

These are a huge contrast with the first group: all bar one has a win percentage below 8% (worse than one win in 12) and huge losses were incurred in 19 of the 20. The A/E indices are generally poor as we would expect given the other metrics, but 16 are 0.60 or lower which is extremely low. What also stands out here is that the left hand column showing the 'today' course where the vast majority are Grade 1 tracks, York and Ascot accounting for 12 of the 20 between them. Meanwhile, the second column of 123LTO courses is largely comprised of lower tier tracks.

My next port of call was to classify each course into grades (It will become apparent later why I’m doing this). There are eight Grade 1 tracks on the flat: Ascot, Doncaster, Epsom, Goodwood, Newbury, Newmarket (both courses), Sandown and York. My plan was to grade all courses in the most accurate way possible. I had two ideas. The first was to work out the average prize money at each course over the time frame; the second was to work out the average race class level. I chose the latter because I thought using prize money could see some course averages get skewed due to the very biggest races offering such huge purses; thus, using the average class of race I felt would be more accurate.

Here are my findings for each course. I have put them into three graphs with the courses in alphabetical order:

Ascot has hosted the highest average class of race, with Brighton the lowest; and there is a significant difference between the two. Ascot averages 2.16 (i.e. between Classes 2 and 3 on average, much closer to Class 2), while Brighton is down at 5.43 (midway between Classes 5 and 6). To help give some context, 55% of Brighton’s races have been Class 6 events, whereas at Ascot 34% of races were Class 1 and 31% were Class 2.

From here I decided to grade the tracks (Grade 1 courses seeing the highest class of races, Grade 5 seeing the lowest):

The average bands I chose were partly based on the individual course averages and partly 'feel-based'. For example, Windsor was on 4.52 and I felt it should be in the Grade 3 group, hence I chose 4.55 as the upper range for that grouping. Having decided upon the splits here are the courses that appear in each group/grade:

As you can see the majority of the courses land in Grade 4 or 5. It would be preferable perhaps to have slightly more even numbers in each group but those seemed to me to be the most sensible divisions.

Having embarked on this course grading journey let me explain my rationale. I wanted to group the 'today' and 123LTO course data into bigger sets then individual tracks to see if there were any useful patterns. For example, going back to the lowest 20-win strike rates table I shared earlier, that table highlighted that horses which finished in the first three at some lower grade tracks performed poorly if reappearing at one of the higher grade tracks. Is that the case generally? Let’s find out!

To begin with I have simply compared the results based on the course grade 'today' linked with the 123LTO course grade. The table below shows my findings:

That's quite a mixed bag of results, with eight of the 25 combinations making a profit. However, in order to get a better ‘feel’ for the data I have grouped the 'today' course grade results into one.

The results at all Grade 1 courses combined are the worst, both from a strike rate and returns perspective. Horses racing at a Grade 3 course after a top three finish have done the best and have snuck into profit. It should be noted that the Grade 3 total of runners is the smallest... but there were still over 10,000 qualifiers from this eight-year time frame.

Having grouped the course data above in terms of course grade, it makes sense to do the same for the LTO course data:

This time we see slightly worse returns at either end of the spectrum, but there's nothing too significant to be gleaned from this grouping unfortunately.

My final piece of digging connected with the grading of courses idea was do some ‘rearranging’. My thinking was to create an up-in-course grade / down-in-course grade idea, similar to the one for class change or distance change. To that end, I subtracted the course grade from the LTO course grade to create a ‘difference’ figure. For example, a horse that raced at a Grade 4 course LTO and now racing at a Grade 1 track would have a figure of 3 (4 minus 1). Positive figures can be deemed ‘up in course grade’, negative figures ‘down in course grade’.

Here are the win strike rates:

This graph neatly shows how the change in grade of course affects the win strike rates. Those going from a lower grade course to a higher grade course (the positive figs) have won far less often than those going from a higher grade course to a lower grade one (the negative figures). This is to be expected of course but the correlation is still positive, and beautifully linear. Hopefully we will see a correlation with the BSP returns. Let’s take a look:

The worst return by some margin is for the largest ‘up in course grade’ figure of 4 (i.e. a horse moving from a 123 effort at a Grade 5 track last time to race at one of Ascot, Goodwood, Newmarket Rowley or York this time); with the second worst being the ‘up in course grade’ figure of 3. Those negative returns can be seen on the right hand side of the chart.

The best returns were with the biggest ‘down in course grade’ figure of 4 (i.e. a horse moving from a 123 effort at Ascot, Goodwood, Newmarket Rowley or York last time to race at a Grade 5 track this time). So, at either end of the spectrum we have something of potential use. Unfortunately, the values in between do not show a clear pattern, which is slightly frustrating given the earlier vastly differing strike rates.

It again emphasises that the betting market is so very efficient and not easy to get the better of, despite the huge amount of data collating and crunching one does!

**

It is time to wind up this piece now. Sometimes, despite how well thought out a research project is, we don't always get the findings we were expecting or hoping for. The second half of this article has been a little like that; I was hoping to find more positive or negative angles using the ideas connected with my grading of courses but little came to light.

The most eye-catching elements might be the top and bottom 20 strike rate tables I shared at the beginning, though it's definitely also worth looking out for any horses dropping from a placed effort at one of the Grade 1 tracks into a low grade fixture. For anyone interested in all of the 'today' course to 123LTO course data I have, I am happy to share it. There were too many combinations to fit in this article (469 with 80+ runners, to be precise). Please just post your request in the comments.

My parting shot is that perhaps I need to re-think the 123LTO course to 'today' course idea by incorporating and classifying the course configuration. What I mean by that is whether the course is, for instance, ‘stiff’ or ‘tight’ or ‘galloping’ etc. The only issue with that is some courses will fit more than one course type ‘descriptor’. Hmm, I need to get my thinking cap on and come up with a plan. If I find a good way to do this I will share my findings in a subsequent article later in the season.

- DR

Other Recent Posts by This Author:

• Comparing Starting Price with Betfair SP
• Favourites on the Flat in April
• Early Flat Season Trainer Form
• Looking at Past Cheltenham Festival Trends
• Specific Course/Distance Analysis: Lingfield 1m2f