The Perilous Pitfalls of Racing Systems Creation

Researching and creating your very own horse racing systems is one of the most fun, most rewarding, pastimes available to those of us who are passionate about horse racing and betting.

It gives us the opportunity to truly steal an edge on the market and our fellow punters and, moreover, it gives us control over our own punting destiny.

Sounds good, right? 😉

Now I’ve whetted your appetite for the task at hand, let me bring you down to earth just a little by outlining SEVEN perils of the systems creation process.

You may well be aware of some or even all of these. If you are, consider this a timely refresher; if you’re not, take heed!

So, without any further preamble, let’s hurtle headlong toward the seven deadly sins of racing systems creation….

1. Not using historic data to support an idea you have

You’ve had an idea. You tested it today against the racing and it performed brilliantly. You checked it against yesterday’s results, and it did fantastically there too. Before you believe you’ve found the golden goose (you know, the one that lays the golden eggs…), let’s step back from the nirvana for a moment.

Consider this:

– hunches are not enough

Obviously, two days of success is a most auspicious dawn to a potential system. But remember, if it seems to good to be true, it probably is too good to be true. Find a research tool (such as Adrian Massey’s site or that will allow you to look at a greater cut of horse racing history to confirm or deny the existence of gold in your micro-survey. It might just be fool’s gold, a substance that as we all know is almost omnipotent in the quest for racing systems El Dorado (enough of the gold analogies!)

– it happened last weekend, and always ‘seems’ to happen

A most dangerous contention is this one. As humans we are programmed on what’s known as the ‘pleasure/pain principle’. In simple terms, this explains why even after being burned for the nth time, we still go back to the scolding pot with bare hands one more time. Let me put it another way… why else would women EVER want a second child?! I mean, that’s got to hurt! 😉

The point in all these seemingly pointless examples is that we quickly forget pain when faced with the pure pleasure of backing a winner. If it happens twice or three times, we’ll conveniently forget the twelve times when it failed to pay off and left us with a bag of chips and bottle of brown ale for supper on Saturday night. (Not that there’s anything wrong with either chips or brown ale…)

– ‘my mate down the pub told me’

Tell me about your mate. Is (s)he known for statements of unadulterated, fully substantiable fact? Is (s)he wildly well off, living a life to which you enviously cast your glance askance? I could be wrong, in which case forgive me and my ‘one size fits all’ generalist assumptions, but I’m guessing the answer is an unambiguous ‘no’.

The key point about Horse Racing Experts is that I want to encourage you not to take anyone else’s word for anything where your own financial investment and personal enjoyment is concerned. Of course, I appreciate this flies at odds with my personal recommendations and, if you want to believe me and trust me, then fine. I’ll never knowingly mislead – if a product is carp, I’ll come out and say it’s a small brown fish. (Or a big brown unit of effluent).

But I want people here in HRE to do their own spadework. It’s fun, it’s rewarding, and it’s very, very cool! 🙂 [OK, so I made that last one up to make myself feel better about being a bit of a geek!]

2. Back-fitting on illogical premises

Happy days, hip hip hooray, and hallelujah to boot. All the ‘h’ celebrations you can muster are appropriate if you’re using system checker tools like those mentioned in 1. above to confirm your theories. But… that’s not enough. Shucks, and you thought this racing systems research mallarkey was a breeze…

Well, it is, but you do need to protect yourself from conveniences and coincidences. Let me explain:

– all days except Thursdays

Why would a system perform poorly on Thursdays? What logical explanation could there possibly be for this? In fairness, I know of people who are averse to betting on Mondays (me included), due to the overly poor nature of most Monday mule races. And there are others who exclude / focus solely on Saturday racing because of its competitiveness. Arguable. Reasonable. Maybe. But Thursdays? No!

The above of course is an example of a badly back-fitted system. Aside from that day losing money, there’s no rational reason to exclude it from calculations. And that day losing money is coincidental, not logical. Leave Thursdays in and tell me how the system performs then.

In any other case where you’re tempted to exclude a subset of the data for no supported reason, think very carefully, as it will very likely cost you cold cash in the very near future….

– these six tracks

Again, if the six tracks are all of similar constitution – such as galloping, stiff, undulating, switchback, whatever – there *might* be merit in taking a system around just them. Probably it’s a case of convenience, and you’re encouraged to isolate more universal filters.

– odds on to 4/1, and 13/2 and up

In a word, ‘NO’. There is NO reason whatsoever why you’d exclude a mid-range of horse odds. You might only deal with the top end of the market (e.g. up to 7/2 for instance); or quietly fancied horses only (e.g. 4/1-14/1); or even potshot outsiders (16/1+); but any contrivance that attempts to mix and match some of these is… well, it’s just that: a contrivance. Don’t do it if you’re serious about your racing system’s merit potential.

3. Too small data sample sizes

If, once you’ve applied your filters to your starting angle, you end up with a relatively small number of qualifiers, you might well have a problem. This peril is not quite so straightforward as some, because there can be extenuating circumstances. However, for beginners to intermediates, I’d counsel erring on the side of caution.

So how small is too small? Hmm, now you’re asking… Again, not simple to answer, and it depends on the circumstances of the system under scrutiny. Let’s illustrate with some examples:

– A laying system that looks to lay at 8/1 or higher and had only three winners from sixty qualifiers in the last two years.

At first glance, that looks very interesting. But… there’s not nearly enough evidence to support the notion that 8/1+ shots will continue to lose with such satisfying frequency. Further, you might consider that such an odds range opens up the possibility of hitting a 33/1 winner. Just one such beast completely trashes a promising angle.

– A backing system that bets short priced (3/1 or less) horses, and has had a 35% strike rate over the course of 200 selections for a ROI of 12%. Looks promising, doesn’t it? But again, you couldn’t be confident in the sustainability of such a system.

To get a feel for how big a sample size should be, we can look to Paul’s excellent money management article (

His simple ‘ready reckoner’ tool there would readily reckon that backing 3/1 shots could see us incur a losing run of 24 straight. How would that impact your 12% ROI if they were the next 24 selections.

I know there is a series of ‘rules of thumb’ in this article, and this is unapologetically the case as such methods are – at there core – akin to a pseudo-science. The rule of thumb here is that if you can add one and a half maximum losing runs to your system and still see a profit, then it may well have the underlying integrity required to pay you in the mid to long term.

4. Over-complicating things (KISS!)

Often aligned with 3. above, the curse of the over-complicated ruleset is a hardy perennial in the garden of betting systems. This has been far more succintly (if less elegantly) summed up by many before me, as KISS (Keep It Simple, Stupid!).

Think about the filters you’re employing. Do they all make sense? Are they hamstringing your system by reducing the number of potential qualifiers to one a week / month? Do you have filters that overlap each other?

Let’s look into a little more detail at these questions:

– Do your filters make sense?

We’ve discussed the logical mandate against which all filters must be judged: quite simply, if they don’t pass logical muster; if you cannot say, hand on heart, it makes absolute sense that this would be the case, then you might be better excluding that filter and playing with the larger data sample you’d then have to unearth a more credible filter.

– Are your filters hamstringing your system by reducing the number of potential qualifiers to one a week / month?

How do you like to bet? This question must be at the forefront of your mind when you consider how to research a system. If you like high strike rates and few runners, then a herd of filters (I’m not entirely sure what the collective noun for filters is!) might be fine. However, if you crave action, you’re likely creating a system that would never satisfy your personal need to punt. This will inevitably lead you to look for side bets to keep you entertained in the hiatus between qualifiers. We all know you don’t have the discipline to wait for the next real bet! 😉

– Do you have filters that overlap each other?

Not all filters in racing systems research databases are mutually exclusive. By this I mean that certain criteria can have a significant shared subset. For instance, suppose you were researching a flat system in the months November to March. By dint of the racing calendar, you’d be looking at a lot of all weather racing action and maybe half a dozen turf flat meetings. If the turf flat meetings were contributing materially to the ‘bottom line’ on your system angle, you might well be looking at a tenuous premise from which to commence your investment.

In this case, far better would it be to specify all weather tracks / surface and establish the profitability throughout the entire year (i.e. not just November to March). That is not to say that you would definitely need to include the summer all weather months, but rather than it is a better start point than the inclusion of the potentially misleading turf courses in what is almost exclusively an artificial surface racing period on the flat.

Clear as mud?!

5. Misjudging the %age over for laying systems

Laying systems are tricky. Much more so than they first appear. On the face of it, all you have to is get a horse beaten. Simple, right? Well yes, it is undeniable that in every ten horse race, nine are usually beaten (excluding dead heats!). But the frequency with which horses are beaten is only one part of the equation and, actually, not the most pertinent part.

Much more important is the odds at which you can lay the recalcitrant beast in question. And the level of market maturity is staggering. That is to say, over time, the odds at which one must lay horses on Betfair bears remarkable comparison with the frequency with which they win.

Exemplified, horses laid at 2/1 on the exchanges win very close to one in three races. Note, the non-exchange odds may be anywhere between 6/4 and 9/4 – that is a tiny pool of cash compared with the vast oceans of Betfair liquidity.

The key point here is that it’s not easy to lay horses consistently at odds akin to their non-exchange prices. If you could lay every horse returned at 6/4 in Ladbrokes, at 6/4 (2.50) on Betfair, you’d be able to retire this time next year. But you can’t. Often, you’ll have to lay the horse showing as 6/4 with Lads at 2.76 (just over 7/4) or something like that – fully 17% greater than the SP. And then, if you are lucky enough to get the horse beaten, you’ll be obliged to part with 5% of the strangers’ stake you laid to Betfair for brokering the trade.

The lesson here, specifically for laying systems, is that whatever you think you’d be able to lay the horse at (on average) on the exchanges, add 5% if 3/1 or lower; 10% if 3/1 to 6/1; and probably 20% if laying above those odds. I’m not saying that you’ll always have to lay at, say, 22% over SP in the 3/1 – 6/1 bracket. Rather, I’m saying that when researching laying systems, I strongly urge overinflating the over-SP you’ll have to lay at. If the system still shows a profit, you’re less likely to lose money. And, as any forex trader will tell you, the first priority is not to carelessly lose money. If you happen to make more profit because you inflated the lay odds, that can only be a good thing, huh?

6. Too much / too little ‘action’

OK, so maybe I’m repeating myself here. But it’s of paramount importance that you design a system to suit your betting habits. This will include the odds at which you back/lay (and of course whether you back or lay), the strike rate / losing runs you’re comfortable, and how much ‘action’ you need.

– A system with 15 qualifiers a day may be ripe for refinement to make it more profitable and less bet intensive

Most people are happy enough to bet every day, so long as it’s just two or three bets. More than that and it may be too much of a strain on available funds, or simply too much going on. If designing a system for possible commercial purposes, bear this in mind as it will directly impact how successful your system will be at market, regardless of reported profitability.

If you’ve got lots of qualifiers to generate a decent potential profit, you might well want to consider adding a further refinement criterion. Or maybe you just want to get stuck into all of them! Your call… 🙂

– Too little action makes it time consuming to establish an RIO and can be associated with 3. and 4. above

The flipside of too many picks is, of course, not enough. I’ve touched on this in 3. and 4. above, and again if you’re happy with waiting for days between runners, then that’s fine. It’s your system and your choice. If, however, you wanted to think about sharing your find with others, then you’d need to be aware that for most people (who love to bet) two or three runners a week may be too selective for them.

7. Unacceptably low strike rate / ROI

Finally in this seven stage cruise through the perilous pitfalls of systems building, let’s consider the impact of low strike rates or returns on investment.

For seasoned professionals, low ROI is a part of what they do typically: these shrewdies rely on turnover of capital and low percentages to garner an overall profit.

Moreover, for the likes of Eddie ‘The Shoe’ Fremantle, whose specialism is in backing longer odds horses, the barren spells between winners can stretch for weeks, and psychology is a critical success factor. Having courage and conviction in such times is a real challenge both of the system and the person following it.

Ask yourself, how long are you prepared to go without a winner? How long can you tolerate ‘treading water’ in your quest for a small percentage increment?

Depending on your nature and your appetite for risk / experience, you may need to factor this into your systems creation.

The easiest way to do this is by focusing at the sharp end of the betting market. These horses win the most often and, whilst it’s not easy to make a profit in the most studied trade sector, the higher strike rates will smite most punters’ fears of interminable losing streaks.

But remember this if you’re planning a commercial aspect to your system: when I surveyed NagNagNag readers about their primary reason for betting, almost half gave a non-financial reason (either to solve the puzzle, or for the interest / pastime of it).

Obviously, these people still want to win. But they may tolerate a low ROI on a high turnover of bets in their question for a puzzle solution and / or some entertainment. (TTS is a great example: typically 1000 or so bets for 60 to 80 points profit. A 6-8% ROI is perfectly acceptable and, when combined with the regular interest, and occasional decent priced winners, it’s a winning combination for bettors).


After all that, you’d be forgiven for thinking that there’s waaaay too much to account for when looking at racing systems. But really, you’ll probably acknowledge that you already know at least some of these and factor them in as second nature when thinking about how you might build a betting system.

If not, when you present your idea(s) to the System Proofing / Angles area, there are plenty of fellow systemites who’ll be happy to help you onto the straight and narrow.

Ultimately, know your punting style and heed the warnings here, and you’ll be well on your way to creating the perfect system for you… and maybe others too…

Good luck!



10 Responses to “The Perilous Pitfalls of Racing Systems Creation”
  1. Paul says:

    Great article Matt – very educational too!

    ~ Paul

    • Toby Drysdale says:

      Nice article Matt, definately some things to consider.

      I’ll just go back to point 2 though – Depending on illogical premises.

      Depending on the data size, I would say it’s worth looking “further” into the reasons why a particular system may work, by excluding a particular day. If your system is backed up by 2-3 years worth of results and a decent sample size for each year – then there may be a genuine reason – but it isn’t instantly obvious. To me, these systems can be quite valuable as most people will miss them and potentially give you a nice niche in the market. Sometimes you just have to think a little outside the box 🙂

      • Chris says:

        Excellent post Matt! 🙂

        I have a question very much linked to point two and the above post made by Toby. Where for example we include/exclude specific courses – would this be allowed in the event we show years of positive statistics backing up this move or would it still be frowned upon by the majority of punters?

        Thanks Chris.

        • Paul says:

          It depends on what the rest of the rules are Chris. Matt’s TrainerTrackStats targetted specific courses, but that was logical because certain trainers target specific races at specific courses. Another example would be if courses are similar in style and conditions… ie. flat courses versus courses like Epsom with lots of ups and downs. Or jumps courses like Uttoxeter or Chepstow that have a habit of getting waterlogged. If the rest of the rules were pointing in the direction of these factors that could make sense.

          ~ Paul

          • Chris says:

            Paul, do you know of anywhere that I can look into courses styles and conditions further?

            These questions are mainly to gain a better understand of Matt’s article.

            I just wondered also for example if there were ever exceptions for saying having a 5 year profitable picture by excluding a few courses or more than ten courses without any real knowledge behind the results being seen by the user. Well, there could be some, but not explanations for all of the courses. I was just thinking up various scenarios could potentially face in future unless have these clarified now.

            However, what seems clear here is there has to be logic behind any rule/decision even with the selected courses and so I assume this wouldn’t be a wise move?

            Thanks 🙂

      • Paul says:

        I don’t think there’s a logical reason for excluding a particular day either to be honest, not even a Monday – although the racing is typically low-grade on a Monday. As for a Thursday, maybe it rained more on Thursdays than any other day over the last 5 years, but would that be a good reason to remove it from the results? It wouldn’t… just because more rain fell on a Thursday over the last 5 years, doesn’t mean it will happen again in the future.

        It doesn’t really matter what the sample size is, if there isn’t a logical reason for something happening, then I don’t think you should use it as a filter.

        ~ Paul

  2. Paul says:

    Hi Chris,

    I needed to reply here as the comments only allow so many in a thread.

    The SportingLife website is quite good for this…

    Although for a more thorough reference guide you might want to try Amazon…

    The Channel4 guide is probably one of the best.

    I don’t know about 5yo’s, not on the flat at least, as they’re mature by that age and I can’t think of any reason why that age group would perform better at different courses. However, a possibility might be 2yo’s and 3yo’s on testing courses, because the conditions would be tougher for them. Looking at breeding might be the best angle into that as the offspring from different sires will develop at different times, just like humans really. Mark Johnson’s juveniles would be a good example… quite often you’ll see him run something in a maiden that’s built like a horse, literally :), without even seeing a racecourse before, whereas everything else in the field looks about half the size. I don’t know whether there’s a system to be made out of that particularly, but looking at the breeding will often give you a good idea of what type of course and what type of race a horse will be suited to.

    ~ Paul

  3. Chris says:

    Hi Matt and Paul

    Thank you both for your explainations, great examples and useful resources provided.

    These have really helped to clear up some of my concerns and given me some great new ideas. 🙂

    Cheers Chris 🙂

Speak Your Mind

Tell us what you're thinking...
and oh, if you want a pic to show with your comment, go get a gravatar!

You must be logged in to post a comment.