How does exposure impact performance?

As I write this opening salvo, I have yet to undertake the number crunching for what will follow so I can be as candid as possible, writes Dave Renham.

Have you ever wondered if there is an optimum or near optimum number of runs a horse should have in a year? I must admit I hadn’t really thought about it until the other day when I was pondering potential new angles for research. My educated guess was that sweet spot in terms of number of runs would be different for each race code, perhaps lower for National Hunt compared with the flat or all-weather. My reasoning for this was relatively straightforward, and hopefully logical, in that National Hunt racing is more demanding and hence horses would need longer breaks between races. Longer breaks between races means fewer races in a campaign. That particular question will not be answered in this article as I am going to focus on National Hunt racing only. Soon I will revisit this idea for the flat and then we'll test the hypothesis.

Thinking about National Hunt only then, I was edging towards around five to six runs as the likely optimum before I started my research. The argument I made to myself was that most National Hunt horses run for a particular portion of a year: a good number will run primarily between October and April which comprises the main NH season. Of course, there are summer jumpers who tend to ply their trade in the off season, as it were. Both types are likely to race for between five and six months of the year; and, working on a premise of roughly one run per month, that is where I came up with my five to six runs prediction.

For this research, data has been taken from UK National Hunt racing spanning from January 1^st 2017 to December 31^st 2023, a period of seven years. Also, to clarify, ‘horse runs in a year’ means the number of runs a horse has had in last previous days. Any profit/loss figures will be quoted to both Industry SP and Betfair Starting Price (BSP).

OK, with that said, let's combine all horses and review by number of runs, focusing first on their strike rate.

N.B. I have excluded debutants from all the findings – hence the cohort of horses with zero runs in the last 365 days only contains horses that had run before, i.e. more than a year prior. Now, this is the only group for which I have made that adjustment. This means therefore that you may get a runner, say, in the ‘three runs in a year group’ that has had precisely three career starts. I have not tinkered with this ‘one run or more’ data in terms of considering career starts for a variety of reasons. One reason was because it was by far the easiest way to collate the data. Doing it any other way would have caused me so many problems / questions it would not have been worth the time and effort. Another key reason was because I did cross check a few parts of the data in terms of considering how many career runs a horse had in relation to their last year's number of runs. It made virtually no difference to the overall strike rates, A/E indices, etc. So ‘if it ain't broke’ etc.

It is always important to be transparent when analysing data – sometimes there is no perfect way, or the route to perfection doesn't justify the additional effort. You just have to go with the method that in your opinion works best.

Reverting to the graph, the zero runs in a year group has comfortably the lowest strike rate at 6.4%. One would have expected this as I am guessing being off the track for so long means most of the runners in this group would have likely had at least a small setback, possibly quite a serious one. The highest strike rate (13.4%) is for horses that had raced five times in the past year, but there is very little difference between the groups of four to twelve runs. At least my five to six prediction lies within this grouping!

It should be noted that the sample sizes start to diminish once we hit nine or more runs in a year. Hence, I have grouped 9 to 10 runs together, 11 to 12 together and 13 or more together. Also, it should be noted that if I had split the 13+ group into subgroups the graph would have continued in a downward direction. Knowing this, if we added a couple of extra bars to the chart we would see a typical bell-curve distribution. As we know, win strike rates are only part of the story, so let's take a look at the win & placed (each way) strike rates as a comparison. Here are the findings:

As can be seen there is excellent correlation with the win only strike rates. One would expect this to be the case, but datasets do not typically match as well as this.

Time to look at A/E indices, which is one of the key indicators of ‘value’. Here is another bar graph, then, this time comparing the A/E:

Given the numbers A/E indices generate, we again have a similar pattern to the two previous graphs. It may be slightly less obvious, perhaps, but the highest figures lie once more between four and 12 runs. The 0, 1, 2 and 13+ indices are again the lowest four as previously seen looking at the Win SR%s.

It is now time to examine the results by runs, wins, profit/loss, ROIs. Here are the breakdowns:

Losses are steepest in the 0 to 2 and 13+ groups. Again, this correlates well with all the pointers given from the previous stats. It is interesting to see the 9 to 10 and 11 to 12 groups edging into BSP profit, but as you would expect there is the occasional huge-priced winner which skews this.

The returns produced again suggest it is best to concentrate on the four to 12 run group, although it could be argued that if focusing on ‘returns’ we could also include horses with three prior runs in the year.

In an ideal world, at this juncture I would have liked to see if there was much difference in terms of whether we were dealing with hurdlers or chasers, but this was too complicated to test thoroughly; the reason being that there were too many horses which switched from hurdles to chases or vice versa within that 365-day time frame. What I could look at, though, were individual trainer performances, so let me share that with you.

In the table below I have compiled the win strike rates pertaining to the number of runs in a year for a selection of trainers. I have also included their overall National Hunt win strike rate (excluding debutants) to offer a baseline, and the tables have been ordered by these individual win percentages (starting with the highest).

I have also colour coded the tables so anything in green shows an above average trainer performance in relation to their overall SR% while considering the average SR% figure for all trainers; anything in red represents a below average performance in relation to their overall SR% while considering the average SR% figure for all trainers.

I have had to split the table into two due to the amount of data. Hence, the first shows the number of runs in the last 365 days of between 0 and five, the second table looks at six or more runs. Where there are empty cells, the sample size was too small to give a meaningful SR%.

This makes for interesting reading for certain trainers. Here are five handlers I have noted for one reason or another:

1. Harry Fry has performed extremely well with horses that have had 0, 1 or 2 runs only in the last 365 days – he bucks the overall trainer trend with such runners.
2. Dr Richard Newland has excelled with horses that have run at least eight times in the last year / 365 days. With this group of runners he has secured a strike rate of close to 24%, compared with his overall strike rate of 17.6%.
3. Peter Bowen’s runners seem to improve steadily the more they run in a year. He has a poor record with the 0 to 2 runs group, the 3 to 5 group hit at his average win rate but, when we get to six runs or more, they generally exceed his average win rate. Horses that have run 11 or more times in a year have won 33 races from 149 runs (SR 22.2%) – nearly 8 percentage points higher than his overall SR% of 14.5%.
4. Paul Nicholls has a poor record with horses that have not run at all in the last 365 days – losses to BSP amount to -£29.21 which equates to 35p for every £1 bet. However, with all the rest, his strike rate is very consistent ranging from 21.7% to 25.7%. Just a 4% differential between highest and lowest.
5. Runners from the Christian Williams generally improve the more they race. Horses that have had 0, 1 or 2 runs only in the last year have produced woeful win figures, losing backers over 75p in the £ to SP; 65p in the £ to BSP. Whereas once we get to seven runs or more in a year we see much better results – much higher strike rate and close to a break-even betting situation.

*

This has been an interesting area to look at; as I mentioned at the beginning this is something I had never properly thought about before, let alone researched. It certainly has uncovered some data I will use in the future especially when it comes to horses that have not run many times in the past year. Also, being aware of individual trainer patterns is surely going to be helpful moving forward. It was not the easiest idea I have researched, however, it has certainly made me want to look at similar data for flat racing – something I will share with you in the future.

- DR

Other Recent Posts by This Author:

• Seeking Future Profit from Last Day P’s
• Top Ten All-Weather Front-Running Biases
• National Hunt Racing Systems for 2024/25
• A Look at All-Weather Returners
• Hat Trick (Plus) Seekers in NH Racing