mathare
1st April 2007, 22:30
I recently came across a few systems in a new gambling publication and they struck me a great examples of how NOT to create a system. The figures look quite impressive I'll admit but then they would when you see what the system creators have done.
I'm pretty confident I'm doing nothing naughty here as the systems are already in the public domain as they are printed in the latest edition of said publication. I have paraphrased as necessary to avoid copyright issues.
The magazine (who will remain nameless for now as there is no need to either promote or embarrass them) lists four systems, one for each of the courses they are covering in detail. I shall list the system rules for one of the systems and give the summary stats before offering my opinions on why this is the wrong way to go about things.
System 1
Redcar only
Fav with SP of evens to 11/4
Flat races exc 1m and 1m1f
2-15 runners
Soft, good or good-firm
Exclude selling stakes
Not Wednesday or Friday
Not March, August or November
33 winners from 76 races (43.4% SR)
124.8% ROI at SP
24.8% profit may seem like a decent return but take a look at this system in more detail. The key to system building is to find a profitable trend and to exploit it. These rules look like they have been fitted to the data and as such any profit is more likely to be coincidence than anything else.
Redcar only - I don't have a problem with this rule. Some trends show up better at certain tracks. However, it is worth noting that when building a system the performance at each track should be considered a filter rather than a major rule. That is you should be looking to identify a trend that makes sense for all racing and then see how the profit breaks down by track. It may be that for some reason some tracks don't fit the trend as well. This may be down to the course characteristics, e.g. galloping, undulating, left-handed etc.
Fav with SP of evens to 11/4 - concentrating on the market leaders is not an unusual ploy. But including SP in the rules is a rookie mistake. Starting price is not, and can not, be known until the race has started. And you can't generally back a selection when the race has started. If you need to use a indicator of the market either compile your own forecast odds or use those provided in a newspaper or online at Sporting Life or Racing Post. You can't include SP in a system rule.
Flat races exc 1m and 1m1f - these distances are odd ones, normally systems apply up to a certain distance or beyond a certain distance. To exclude the 1m and 1f races strikes me as baffling, unless there is something strange about the part of the course over which these races are run. The course is a left-handed oval of just over 1m with a chute providing a straight mile. So perhaps the problems arise with that straight mile and races run once round the oval. I dunno. But it doesn't sit comfortably with me.
2-15 runners - less than 2 runners and you don't have a race but you don't want fields to get too large. OK. Of all the rules this is probably the one I have least problem with. Horses can run into a lot of trouble in big fields so this one at least makes sense to me.
Soft, good or good-firm - using the going as a filter is not a problem, but the chosen goings are a little odd. Soft is OK, good is OK, as is good-firm. But not good-soft so why not? System rules should quote a range of goings because the official going is often slightly wrong. Soft ground may be riding good-soft, for example. But that throws this system out of line. Had the rules stated the going had to be soft through to good-firm I would be happy. Not too muddy, not too fast with plenty of scope in between if the official going is slightly out. But that's not what it says, and that makes me wonder how this rule came about.
Exclude selling stakes - some trends are not seen in some race types. Maybe it's the quality of the animal running in them. This seems a little too specific though. Claimers are OK, as are maidens. But not sellers. And a selling handicap would qualify. Does that not strike you as odd?
Not Wednesday or Friday - and now we are in to the plain ridiculous. Redcar hosts 17 flat race days this year. There are two Wednesday fixtures in that lot and two Fridays. That rules out nearly a quarter of the meetings already. And the rationale behid this rule? God only knows! One day has to be the most profitable for this system, so it stands to reason that some days will be less profitable. But does it make sense to miss certain days of the week out? An annual race run on a Thursday this year will be run on a different day next year. Is that any reason for it not then to qualify. This is just nonsense in my opinion.
Not March, August or November - not the start of the season (March) or the end (November). I'm fine with that. Horses running early season may be run to get an idea of how the rest of the stable is shaping up. Or they may be running just to get a run into them. At the end of the season a horse may be run for one last shot at the prize money before it is retired or put away for the winter. But August? If anyone can explain the thinking behind not following a system in the middle of the season then I am all ears.
The notes for this system say that 2006 went against the profitable trend but the law of averages says this should average out and following these stats should produce winners, expecially in handicaps. I say hogwash!
What has happened here is someone has analysed the Redcar data and looked at how to produce the best profit from the data available. Horses have steadily been eliminated by applying filters as rules without thinking if the rules make any sense any more. This is something the human brain is prone to - fitting trends that don't exist to data it doesn't actually fit. This is only one step removed from having a system that says don't back any losers.
The systems for Thirsk, Ripon and Catterick are similar piles of poo.
I have no idea how many years have been analysed to build these systems (I suspect it is 2002 to 2006 based on other data in the article) but 76 qualifying races isn't many is it? When paper trading the general advice is to try and get a couple of hundred qualifiers under your belt before you decide whether or not it's working.
As I said earlier, I am offering this up as an example of how not to create a system. The trend is artificial, created by filtering the data to maximise the profit. And the authors even admit it made a loss last year.
There are hundreds of systems like this out there on the internet, especially on Ebay. You'd be wise to take each and every one of them with a pinch of salt. Before taking on a system look in detail at the rules (if available). Do they make sense or do they smack of having been drafted to fit the most profitable runners? Look at the stats over several years and pay attention to good and bad years. Is the system built on trends that still exist or has the BHB changed the rules such that the system is no longer profitable? Can you obtain a list of all historical qualifiers?
These are the sorts of things you should be thinking about when it comes to adding new systems to your betting portfolio.
I'm pretty confident I'm doing nothing naughty here as the systems are already in the public domain as they are printed in the latest edition of said publication. I have paraphrased as necessary to avoid copyright issues.
The magazine (who will remain nameless for now as there is no need to either promote or embarrass them) lists four systems, one for each of the courses they are covering in detail. I shall list the system rules for one of the systems and give the summary stats before offering my opinions on why this is the wrong way to go about things.
System 1
Redcar only
Fav with SP of evens to 11/4
Flat races exc 1m and 1m1f
2-15 runners
Soft, good or good-firm
Exclude selling stakes
Not Wednesday or Friday
Not March, August or November
33 winners from 76 races (43.4% SR)
124.8% ROI at SP
24.8% profit may seem like a decent return but take a look at this system in more detail. The key to system building is to find a profitable trend and to exploit it. These rules look like they have been fitted to the data and as such any profit is more likely to be coincidence than anything else.
Redcar only - I don't have a problem with this rule. Some trends show up better at certain tracks. However, it is worth noting that when building a system the performance at each track should be considered a filter rather than a major rule. That is you should be looking to identify a trend that makes sense for all racing and then see how the profit breaks down by track. It may be that for some reason some tracks don't fit the trend as well. This may be down to the course characteristics, e.g. galloping, undulating, left-handed etc.
Fav with SP of evens to 11/4 - concentrating on the market leaders is not an unusual ploy. But including SP in the rules is a rookie mistake. Starting price is not, and can not, be known until the race has started. And you can't generally back a selection when the race has started. If you need to use a indicator of the market either compile your own forecast odds or use those provided in a newspaper or online at Sporting Life or Racing Post. You can't include SP in a system rule.
Flat races exc 1m and 1m1f - these distances are odd ones, normally systems apply up to a certain distance or beyond a certain distance. To exclude the 1m and 1f races strikes me as baffling, unless there is something strange about the part of the course over which these races are run. The course is a left-handed oval of just over 1m with a chute providing a straight mile. So perhaps the problems arise with that straight mile and races run once round the oval. I dunno. But it doesn't sit comfortably with me.
2-15 runners - less than 2 runners and you don't have a race but you don't want fields to get too large. OK. Of all the rules this is probably the one I have least problem with. Horses can run into a lot of trouble in big fields so this one at least makes sense to me.
Soft, good or good-firm - using the going as a filter is not a problem, but the chosen goings are a little odd. Soft is OK, good is OK, as is good-firm. But not good-soft so why not? System rules should quote a range of goings because the official going is often slightly wrong. Soft ground may be riding good-soft, for example. But that throws this system out of line. Had the rules stated the going had to be soft through to good-firm I would be happy. Not too muddy, not too fast with plenty of scope in between if the official going is slightly out. But that's not what it says, and that makes me wonder how this rule came about.
Exclude selling stakes - some trends are not seen in some race types. Maybe it's the quality of the animal running in them. This seems a little too specific though. Claimers are OK, as are maidens. But not sellers. And a selling handicap would qualify. Does that not strike you as odd?
Not Wednesday or Friday - and now we are in to the plain ridiculous. Redcar hosts 17 flat race days this year. There are two Wednesday fixtures in that lot and two Fridays. That rules out nearly a quarter of the meetings already. And the rationale behid this rule? God only knows! One day has to be the most profitable for this system, so it stands to reason that some days will be less profitable. But does it make sense to miss certain days of the week out? An annual race run on a Thursday this year will be run on a different day next year. Is that any reason for it not then to qualify. This is just nonsense in my opinion.
Not March, August or November - not the start of the season (March) or the end (November). I'm fine with that. Horses running early season may be run to get an idea of how the rest of the stable is shaping up. Or they may be running just to get a run into them. At the end of the season a horse may be run for one last shot at the prize money before it is retired or put away for the winter. But August? If anyone can explain the thinking behind not following a system in the middle of the season then I am all ears.
The notes for this system say that 2006 went against the profitable trend but the law of averages says this should average out and following these stats should produce winners, expecially in handicaps. I say hogwash!
What has happened here is someone has analysed the Redcar data and looked at how to produce the best profit from the data available. Horses have steadily been eliminated by applying filters as rules without thinking if the rules make any sense any more. This is something the human brain is prone to - fitting trends that don't exist to data it doesn't actually fit. This is only one step removed from having a system that says don't back any losers.
The systems for Thirsk, Ripon and Catterick are similar piles of poo.
I have no idea how many years have been analysed to build these systems (I suspect it is 2002 to 2006 based on other data in the article) but 76 qualifying races isn't many is it? When paper trading the general advice is to try and get a couple of hundred qualifiers under your belt before you decide whether or not it's working.
As I said earlier, I am offering this up as an example of how not to create a system. The trend is artificial, created by filtering the data to maximise the profit. And the authors even admit it made a loss last year.
There are hundreds of systems like this out there on the internet, especially on Ebay. You'd be wise to take each and every one of them with a pinch of salt. Before taking on a system look in detail at the rules (if available). Do they make sense or do they smack of having been drafted to fit the most profitable runners? Look at the stats over several years and pay attention to good and bad years. Is the system built on trends that still exist or has the BHB changed the rules such that the system is no longer profitable? Can you obtain a list of all historical qualifiers?
These are the sorts of things you should be thinking about when it comes to adding new systems to your betting portfolio.