My Simple Ratings Method, Part 3

This is the third article connected with my attempt at creating simple ratings for certain horse races, writes Dave Renham. The first piece looked at how you could create a simple ratings method; the second tested this simple idea using some historical results. As the results were quite promising, I thought I would extend the number crunching to more past races, and in this third article I will report back my new findings.

To give some context, I was looking for a method for rating handicap races and, as far as weighting was concerned, I decided to give each factor a similar one. In order to do this I used the PRB (Percentage of Rivals Beaten) metric.

The rating method I came up with involved five factors or variables – these were:

1. Draw – splitting the draw into thirds;
2. Most Recent form – for this factor I used last time out (LTO) finishing position;
3. Recent Market data – LTO price was used for this one – so the Industry Starting Price the horse was returned in its most recent race;
4. Long term form – for long term form I used career placed percentages in handicap races.
5. Fitness –I used days since last run to create the PRB figures for this final variable.

For more ‘meat on the bones’, the first article explains in depth what the individual PRB values were within each of the above categories. Here's the link.

Essentially each horse ends up with five PRB values and therefore when rating a race, we simply add up the five PRB figures for each horse giving them a final total or rating score. The horse with the highest total becomes top rated, the second highest total becomes second rated and so on.

In the second article I did some back testing on the ratings, although for this stat-busting exercise I decided to ignore the draw factor by focusing on longer distance races. ‘Time’ was the main reason for ignoring the draw – it was something that was going to take far too long to collate the necessary information. Hence my ratings were ‘trimmed’ – now they would be created by using the four other PRB figures produced from LTO position, LTO price, career handicap placed percentages and days since last run.

I back tested 324 races with the following rules:

1. Year - 2018 (UK racing)
2. Age group - 3yo+ / 4yo+ handicaps
3. Distance - 1m1f or longer
4. Runners – 8 or 9

The results showed promise, and you can look at the in depth findings here.

For this next batch of testing, I kept the first three rules mentioned above, but changed the runners rule to 10 to 12 runner races. Before sharing the results of this second phase of testing, let me quickly share my thoughts on what I perceive to be the most important finding. For any ratings to have the ‘potential’ to be useful, they need to show strong similarities with the actual betting market. The first phase of testing did see this happening. For example, the top rated runner started favourite in over 41% of the races rated. Just 3.7% of top rated runners started 7^th, 8^th or 9^th in the betting. Ultimately if your ratings do not mirror the betting market that well, then the chances are they are going to be dud.

Of course, as punters, we are looking for value, and the hope is also that the ratings throw up value selections. The first set of results shared in the second article offered some promise in that regard.

OK, it is time to look in detail at the 10 to 12 runner 2018 results. For the record this comprised of 362 races in total.

Firstly, let's review how the top rated horses matched up against the actual betting market.

 

 

This graph perfectly shows the type of sliding scale one wishes to see. It is similar to the one we saw when analysing the 8 to 9 runner race data.

Over 35% of the top rated runners started as the favourite, while more than 72% of the top rated runners started in the top three in the betting. Compare this to 8^th or bigger in the betting which accounted for just 5.1% of all top rated runners.

Let's now look at the second top rated runners next in the same way:

 

 

Again, this graph gives the type of results that suggest the ratings are fairly accurate in terms of assessing potential chance of winning. We would expect the highest bars in the graph to be on the left hand side once more with a sliding scale going from left to right. 57.2% of the 2^nd rated runners ended up in the top three of the betting compared with 2.3% ending up in the bottom three of the betting (10^th or worse).

So once again the top section of the ratings are looking good. Time to take a look at the lowest rated to see how they fit against the market. Firstly let us look at the market rank percentages for the lowest rated runners. To begin with let's review the 12^th rated runner in 12 runner handicap races:

 

 

This graph is effectively a mirror image of the first two we saw, which is exactly what I would hope to see. Over 66% of the bottom rated runners ended up 10^th or lower in the betting.

Now a look at the 11 runner races (bottom rated):

 

 

An even better set of figures here in terms of correlation. This is probably due to the fact that the 11 runner sample size was around 33% bigger than for 12 runner races.

I don’t see the need to show the whole graph for the 10 runner races as well, but the results were similar once more. The bottom rated runner appeared 8^th, 9^th or 10^th in the betting market over 63% of time, while only 12% of them ended up in the top three market positions.

It is very pleasing to see that the results we got for 8 to 9 runner races are being replicated here. Essentially these simple PRB based ratings are looking like providing a relatively sound framework in terms of forming our own market – if nothing else. I discussed some ideas about how to form a betting tissue/market in this article which preceded this series. These ratings could be used in conjunction with that – or even be used in a stand-alone manner.

Time to see how the ratings fared in terms of winning – their win strike rate. In the 8 to 9 runner results, the higher rated runners comfortably out-performed the lower rated ones. Obviously, I'm hoping for the same scenario here:

 

 

These results are reasonably positive – the top rated runner has done extremely well and we do get the type of sliding scale one would hope for. In truth, the 8 to 9 runner data looked stronger, but when you analyse win and placed data, the picture looks more clear-cut:

 

 

This graph gives excellent correlation with higher rated runners hitting higher win and placed strike rates; lower rated runners doing the reverse.

It is time now to look at the performance of the individual rating positions in terms of profit/loss to BSP. At this point it should be noted that in the whole data set for this article, there have been some unusually big-priced winners. The ten highest BSP prices of winners during this study were:

 

 

Within the type of sample size used for this piece, huge priced winners are a common problem when trying to use BSP as a value metric. That is why in other Geegeez articles, where appropriate, I have quoted BSP on shorter priced runners only. Unfortunately using this shorter priced idea will not work effectively on the rating positions data due to very small sample sizes (for the lower rated positions in particular). Therefore, in this case, the BSP profit/loss figures for individual ratings positions shown below may confuse matters for some readers, but hopefully you'll still get the gist at least:

 

 

We do have to take the profit/loss figures here with a pinch of salt however; especially the lower rated ones. Ordinarily strike rates of 4.1% and 5.2% are not going to produce stunning returns of 78.6 pence in the £ or 57.2 pence in the £.

Thus, instead of dwelling on these skewed figures, it makes more sense to dig deeper into the top-rated runner results as these prices as a whole are much shorter. In fact 88% of all top rated runners were priced 12.0 BSP shorter.

I want to look at two main areas when it comes to top-rated runners. Firstly I want to delve into profit / returns, so here are the profit/loss stats for top-rated runners in terms of their market position.

 

 

It is interesting to see that the top-rated runner has made a decent looking profit when actually starting as the favourite. From a ratings value perspective though, I would have liked to have seen slightly better figures from the poorer market positions. Having said that, 362 races is too small a sample to see a potential pattern emerge such as that, especially when just 63 top rated runners started 5^th or higher in the actual betting market. It is heartening to know that top rated runners that were 5^th or bigger in the betting did make a profit of £11.15 from these 63 runners, but we need much more data. Not just on top rated runners but on other rating positions too.

Secondly, I wanted a breakdown of how far clear of the 2^nd rated the top rated was. This is something that I omitted to think about when penning the last article. Hence for this next table I have combined the relevant stats from both articles to include all 8 to 12 runner handicap results. This gives us a bigger data set for analysing the gap between the top two rated runners. Here are the findings:

 

 

Before commenting on this, it must be stressed that despite expanding the sample size, it is still a relatively modest one. However, one could not have dreamed of a much better set of results (well, I suppose I could have, but you have to have some sense of realism!) The bigger the gap, the better the results – both from a strike rate and returns perspective.

My next port of call was looking at ratings position versus betting market position. I wanted to compare the performance of horses that are rated better than their odds position, compared to those who are not. Here are the results:

 

 

As with the 8 to 9 runner data, horses ranked better than their price ranking have done best from a profit/loss perspective. There is a big differential here, but as I have already stated, the BSP data for all these races is not too reliable, and hence I would not read too much into this.

Before finishing, I have one more set of figures I want to share. As I did in the previous piece, I want to look at the actual rating scores and group the lower rated runners as a whole, and compare them with the higher rated runners. Tthe lowest possible rating using my PRB scores is 1.64; the highest possible is 2.39. The groupings I have used are horses that were rated 1.64 to 1.84, and horses that were rated 2.18 to 2.39. These are exactly the same groupings I used in the 8 to 9 runner article. Here is the comparison of wins, runs and strike rate for our two groups in 10 to 12 runner handicaps:

 

 

These stats are what one would have expected based on all the previous data shared in this article. However, it is always nice to have expectation validated in black and white.

I have not added the BSP profit figures as the 1.64 to 1.84 group had two of those huge priced winners I alluded to before (353.78 and 137.3). Such winners totally skew the profit/loss column making a comparison a mockery (as we have seen twice before in this piece, with the profit/loss figures for individual ratings positions, and with the rating rank v market rank BSP comparison). For the record, the 2.18 to 2.39 group, which did not have big priced winners skewing the results, lost a modest 20 points to BSP equating to an ROI of -4.7%.

So that’s currently where I’m at. There has definitely been further promise in this latest piece of research. I will decide where I go next with this over the next few days and any new ratings research will be written up and shared with you in the very near future.

- DR

Other Recent Posts by This Author:

• National Hunt Racing Systems for 2024/25
• A Look at All-Weather Returners
• Hat Trick (Plus) Seekers in NH Racing
• Sires on the Sand: All Weather Stallions To Note
• Statistics for National Hunt Racing in November