Why Strike Rate is Not the Only Metric in Town
In the articles I write for geegeez.co.uk, I use various different means of evaluating performance, writes Dave Renham. One of them is win strike rate. This is the measure of how often, as a percentage, something wins, be it a horse, jockey, trainer etc. I also typically look at the combined win and placed strike rate which I usually refer to as each way strike rate.
While strike rates have their place, and I especially like them for comparing specific data grouped by year because I believe seeing similar percentages year in year out should give us more confidence that an angle may be replicated in the future.
Winners, or Profit?
However, higher strike rates do not necessarily equate to value or profit. Let me try to illustrate this by looking at the following two scenarios covering a thousand bets.
- A strike rate of 45% at prices of Even money
- A strike rate of 25% at prices of 7/2
The first scenario would see 450 winning bets and 550 losing ones. The second scenario would see 250 winning ones and 750 losing ones. So initially one may think that the first scenario is the preferred one. However, if betting all runners at £1 level stakes, the first scenario would produce an overall loss of £100 (£900 returned on £1000 staked) – that equates to losses of 10%. Meanwhile, betting £1 level stakes based on the second scenario would deliver a profit of £125 (250 x 4.5 = £1125, less £1000 staked) which equates to overall gains of 12.5%. So, the lower strike rate out of these two examples has proved much the better option.
It is a misconception that number of winners is the most important thing when trying to make money betting on horses. Obviously, we need winners to potentially make a profit – zero winners are not going to make anyone any money! I wish I had a tenner for the number of times friends of mine have asked me for ‘winners’ when they go on their annual trip to the races. It happens every time they go. And each time I give them the same answer: ‘if you want winners, back the favourite’. Backing the fav in each race at a meeting will be the best option if your sole intention is backing the most winners; and for once a year 'day at the races' punters, it's the best approach - hope to be lucky.
However, making a profit at this game is about getting ‘value’ prices, more of which later...
Similar Strike Rates
Before discussing the key word ‘value’ I want to now look at what I call similar strike rates, and potential issues therein. Having similar strike rates does not necessarily provide the same return on investment. Below is a table showing a four-year period in the riding career of jockey Hollie Doyle. Profits / losses and returns have been calculated both to Industry SP and Betfair SP.
As you can see, we have very similar strike rates year on year ranging from 14.16% to 15.42%, but the profit/loss and returns columns differ markedly. Indeed, the lowest strike rate of the four years (in 2019) produced the best returns – using either SP or BSP. So, what is happening here? It is simply down to the fact that Hollie Doyle has become a far more well known and popular jockey as time has passed. Being more popular means more punters bet on her rides, which in turn shortens the price of those rides. However, I need evidence to back up this theory.
To do this I am going to present two graphs. This first shows a comparison of the percentage of her overall rides each year using two price brackets – horses priced 5/2 or shorter, and those priced 18/1 or bigger.
This graph shows two things that correlate with each other. Firstly, the percentage of shorter priced rides (5/2 or shorter) has increased steadily over the years; likewise, the percentage of bigger priced rides (18/1 or bigger) has decreased steadily year on year. These stats help to demonstrate that the prices on Doyle's rides have been shortening over time – at least as far as very short and very big prices are concerned.
What happens now if we consider the average price of all her rides during this four-year period? Once more I have broken the stats down by year and the graph below relays the info:
This second piece of evidence that shows the average price of Hollie's rides has been dropping year by year. There is positive correlation between both graphs. Therefore, we have two conclusive 'exhibits' which suggest that her increasing popularity and exposure has shortened the prices of her rides over the years; and, related, has enabled her to secure the mount on horses with better prospects. I am confident that my theory, for once, has proved to be a good one.
This is not the only example of the similar strike rate issue. One of the reasons certain draw biases no longer provide the potential for profit is that as a bias gets better known, the odds on that bias begin to contract/shorten.
To give a classic example of this let us look at Chester five-furlong data going back to 2004. Chester’s 5f trip is infamous for favouring lower drawn horses due to the tight turning nature of the track. Here are the stats for horses drawn right next to the inside rail (stall 1) over 5f showing a comparison between 2004-2013 and 2014-2023. I have used SP only in terms of profit/loss/ROI as BSP was not used from 2004 to 2007:
Strike rates are similar as we saw with Hollie Doyle’s data, a difference of only 1.5% here, but the profit/loss/ROI are markedly different. Yes, the 2004-2013 runners have a slightly higher strike rate but not enough to make a difference of £73 to bottom line, or an ROI swing from 25% to -20%.
If we now look at the average SPs over these time frames, we can see that the prices have shortened:
We have 6.52/1 (decimal odds 7.52) versus 5.56/1 (decimal odds 6.56) - a drop of nigh on one whole point is significant considering how many runners are in the sample.
To corroborate this,I also looked at the median prices as another guide and the median price from 2004 to 2013 was 9/2 (5.5); from 2014 to 2023 it was a point lower at 7/2 (4.5).
Now, if we examine the average prices of just the winners, we see an even sharper drop:
Going back to 2004 punters were aware of the draw bias at Chester, but the understanding of the strength of the bias was far less concrete than it is now. This is the main reason why the prices have shortened over past two decades. Any value that was around in 2004 is long gone – or so it seems.
We see this type of initial profit then loss with some of the best horse racing systems of bygone times. Systems that initially make a profit eventually start to lose money as the betting market eventually cottons on. With the good ones it is not usually down to strike rates deteriorating, but rather than the prices on offer diminish. For any specific win strike rate to make a profit, the prices need to big enough to make that happen: strike rate and available odds are a partnership perpetually operating in tandem.
The similar strike-rate examples I have shared should help punters whose primary focus has been on strike rate to consider that metric in a broader context. It is the price in relation to the actual percentage chance of a horse winning that we all need to evaluate; because to make money at betting, VALUE is the most important thing, not strike rate. We need the percentages to work in our favour.
As an example, if we could get better odds than Even money on the toss of an unbiased coin, we would naturally take the bet because the odds are in our favour. Unfortunately, the odds in horse races are rarely in our favour mainly because bookmakers have a profit margin built in for them. This is called the overround. To explain overround let me share a worked example. Imagine there is a five-runner race with the following prices available for each horse:
I have added an ‘implied probability %’ column which is the percentage chance of each horse winning according to the individual odds that are being offered. Adding the five implied probability percentages the total is 105.56%. A perfect/fair betting book should add up to 100%. In this case the book is not perfect as we have an overround of 5.56% - the difference is the bookmaker’s profit margin. The bigger the overround, the bigger the edge for bookmakers and hence the harder it is for punters to make money.
Finding Value via Actual/Expected
Of course, finding ‘value’ is not easy. Also, we will rarely know for sure if we have got value on any of our selections. We may have a good idea, but horses are not machines and horse races rarely pan out exactly how we expect them to.
One way to assess ‘value’ is by using the A/E index. The A/E index is a type of impact value stat. The ‘A’ stands for Actual whilst the ‘E’ stands for Expected and therefore the A/E index stat is an index of actual winners divided by expected winners. By 'expected' we mean as implied by the odds available. More on that in a second.
If the A/E index stat has more actual winners than expected ones, then we have found a value scenario. If the stat has fewer winners than expected, then we have found a poor value scenario. To calculate the A/E index we need to know the actual number of winners and the expected number of winners. The first part is easy as this is the number of horses that actually won. Calculating the expected number of winners is more time consuming as we need to sum up the odds of all the runners.
To do this, firstly calculate the probability or odds of winning of each runner. This is not the odds of the horse, such as 4/1, but the percentage chance of winning for the 4/1 shot. We use the following formula to work out the percentage chance:
Odds Chance of Winning = 1 / (price + 1)
For our 4/1 horse, then, the percentage chance of winning (probability) = 1 / (4 + 1) = 1 / 5 = 0.2 (20%)
For a horse priced at Evens the percentage chance of winning = 1 / (1 + 1) = 1 / 2 = 0.5 (50%)
Once we have calculated the odds of each runner we then add all those figures up to obtain the expected number of winners.
Here's a simple example based on 100 horses sent off at 4/1. The percentage chance (in decimal terms) of a 4/1 shot winning, as shown above, is 0.2; so adding all 100 of those together sums to 20. Thus, with 100 horses priced at 4/1 we would expect 20 of them to win. If 25 of them actually won, then the A/E index would be calculated by dividing 25 by 20 to give an A/E index of 1.25. Anything above 1.00 can be said to represent ‘value’ so a figure of 1.25 suggests very good value.
We need to be aware that by calculating A/E indices in this way does not take bookmaker’s overrounds into account so one could argue they are not 100% perfect. However, they are still a very useful barometer and focusing on A/E values of, say, 1.1 or 1.15 will account for most if not all of the bookmakers' margin. As an aside, if you are comparing A/E indices ‘like for like’ – e.g. one trainer versus another – then overround is not an issue.
Regular readers of my work will know I mention A/E indices in virtually every article I write. This is because they are an important piece to be considered in any profit/loss/value consideration.
But the A/E index is not the ‘be all and end all’. You can be profitable with an A/E index under 1.00, likewise you can lose money with an A/E index of above 1.00. However, it is a very useful metric and one we should try to use when analysing results.
Essentially then, if we can obtain value prices when we bet, we will make money in the long term. This can be achieved with a strike rate of 50% but could also be achieved with a measly strike rate of 5%. It should be said that from an emotional / discipline perspective, it is far easier to keep your head in the game with a winner every second bet than one in twenty; this is not a mathematical notion but very much a harsh reality for all of us when we're staring down the barrel of a losing streak!
Another point worth making is that it is important to be aware that profit/loss figures, like strike rates, can be misleading. Imagine a trainer has had 1000 runners over a five year period and has made a £150 profit to SP and a £400 profit to BSP. Do you decide to back this trainer in the future based solely on this evidence? Hopefully your answer is ‘no’. Ultimately you need to do more digging. If during your digging you find that the trainre had two winners during this period both priced at 100/1 SP (paying 200/1 and 250/1 BSP), you will appreciate that when removing these two winners the remaining 998 runners would have made a loss to both SP and BSP. We can see that ‘no’ was the right answer!
Let me share a couple of real life examples. Here are two trainer/jockey combinations that have made a decent profit between 1st Jan 2016 and 12th Feb 2024 (when I penned this).
On the face of it we have excellent figures for both. It is rare to get these sort of profits using Industry SP. However, look at what happens when we split their ROI%s into different price brackets. First Moffatt and Jones:
As can be seen, the partnership have been profitable across all price brackets. The 14/1 or bigger results are the highest and it will come as no surprise that they have had a few big priced winners together. However, these bigger priced winners are not the reason for the overall profit. It has helped, but they have produced excellent returns in every other price group (all above 40p in the £).
Now let’s look at the Lacey/Sheppard splits:
Here is a much more scattered distribution. Three of the price groupings have made a loss. In fact, the 15/2 to 12/1 price bracket produced dreadful figures – losses of over 90p in the £ due to just one win from 99 rides. As we can also see, the bulk of the profits have come from the bigger priced group. Indeed, digging deeper I can realte that the pairing enjoyed a 200/1 winner which has obviously skewed the 14/1+ profits.
Therefore, if you were considering backing one or other of these trainer/jockey combinations in the future, the Moffatt/Jones partnership looks to be a far more reliable proposition. Whether their performance will be replicated in the years to come is impossible to say, of course, but this type of extra analysis should pay dividends for those punters who are prepared to go the extra mile. For the record the A/E indices for the combos were 1.48 and 0.97 – the better value by far was for the Moffatt/Jones results.
*
It is an obvious thing to say but making money from horse racing is not easy; and finding value is not easy. Ultimately, we as punters generally want to make a profit – and ideally a long-term one. Strike rates are not the answer as we have seen although, as I mentioned in the first paragraph, they do have their place in certain circumstances.
I believe the main takeaway from this piece should be this: it is useful to know roughly how frequently something happens; but it’s crucial to know where the value lies.
The A/E index can help with ‘value’, but it is still only one piece of a rather complex puzzle!
- DR
hi dave gd read is there any chance of doing stats for 1st, 2nd, and 3rd starts with trainers with the best strike rate.