Horse Racing Tips

A Horse Racing Tips System That Works – Lay These Horses!

A Statistical Approach to Selecting Horses

About a year ago, after trying every method known to mankind, and failing, I decided to investigate a statistical approach to betting on horses.  My initial feeling was that it certainly couldn’t be any worse than any of the other methods that I had tried.

I decided to begin my investigation by looking at betsafe.lv race favorites.  It seemed like a logical place at which to start. In addition, it was relatively easy to access lots of past data relating to favorites.  Therefore, testing out any theories that I developed was not going to present me with too many difficulties.

Between 1991 and the writing of this article, 123,261 races were held on the UK mainland.  Of these, 34% were won by the race favorite.  During this period, the highest strike rate occurred in 1991 and 1992 when 37% of race favorites won.  The lowest strike rate occurred in 2005 when only 31% won.  So, in 17 years, the strike rate has only varied by +/-3%.  I decided therefore that the strike rate of race favorites was a fairly consistent metric.

Now, from the above, it can be implied that, on a typical day’s racing at a typical racecourse, where there are six races on the card, approximately two races will be won by the race favorite.  This statement, although true, is not as helpful as it may first appear and is also somewhat mis-leading.

Why is the statement not as helpful as it first appears to b?  Well, although, statistically, two of the six races ought to be won by the race favorite, the statement provides us with no indication as to which two race favorites, out of the six, will actually win.

So, why is the statement mis-leading?  It implies that if there are six races at a particular racecourse on a particular day, then two of the races will be won by the race favorite.  However, the statement refers to a typical day’s racing at a typical racecourse on a typical day, not a specific day’s racing at a specific racecourse on a specific day.  Therefore, the statement refers to what will probably happen over the long term, not what will happen in the short term (i.e. on any one specific day).  As a result, on some days, no race favorites will win and, on other days, anywhere up to all six race favorites will win.

Why is it that there is this large variation in the short term (daily) strike rate?

The reason is statistical variation i.e. luck.  This variation causes results to become ‘clumpy’ in that there may be periods where all the favorites win and periods when few, if any, favorites win.

Why is this?

Because there is a certain amount of luck involved in winning a race.  If this were not the case, then the favorite would win every race.

I remember watching a flat race at Chester recently.  A horse accidentally clipped the hoof of the horse in front and three horses fell.  The horses, thankfully, received only minor injuries as did two of the jockeys.  The third, Seb Sanders, unfortunately suffered a broken leg.

In October 2007, George Washington (‘Gorgeous George’ as he became affectionately known as) ran in the Breeders’ Cup held at Monmouth Park, USA.  In the final straight, George Washington broke his cannon bone and, sadly, had to be destroyed.

Although these are, thankfully, relatively rare events in the world of flat racing, horses fall and jockeys are unseated frequently in chase and hurdle races.  Such is the nature of the sport.

Although these are ‘extreme’ examples in that, if they happen, it is almost certain that the horse will not win, it doesn’t always take an ‘extreme’ event to ensure that a horse will not win.

For example, some horses are notorious for ‘idling’ when leading in a race.  The reason is that they have a strong herding instinct which causes them to want to remain in close proximity to the other horses.  When a horse is in the lead and distanced from other horses, it begins to feel vulnerable.  When it feels vulnerable, it shows a tendency to slow down and allow the others to catch up in order to feel less vulnerable.  This is why jockeys ask for an effort from their mounts towards the end of a race.  In this way, the horse passes the finishing line when it has just taken the lead and is not presented with an opportunity to idle.  This requires a certain amount of skill, judgement and timing from the jockey, not to mention an understanding of the horse’s capabilities.  Jockeys, at the end of the day, are only human.  As such, they don’t always get things right.  Even when they do, Lady Luck may still intervene.  Sometimes, a jockey asks for an effort from his mount at just the right time but gets ‘baulked’ by another horse.  The timely run therefore stops before it has barely started.  Sometimes, a horse may ‘miss the kick’ at the start of a race, for whatever reason.  Horses have even been known to sit down in the starting stall or completely refused to start.  If any of these things happen in a short race, then normally there vendetta the horse’s chances of winning.

Luck does play a part in horse races.  More in some, less in others.  However, luck doesn’t come ‘evenly distributed’.  Good luck comes when it comes.  Bad luck comes when it comes too.  To a large extent, it is the uneven distribution of luck that cause the ‘lumpiness’ in results.

Although we cannot determine exactly which type of luck we will get and when, we can calculate how this might affect results over the long term.

We can use the following formula to calculate what the maximum number of consecutive races that there will be without a winning favorite:

Ln (No. races)/ – Ln (1 – Strike Rate)

Where:
Ln – is the natural logarithm
Strike Rate – is 34% = 0.34%
Number of races – is 123,261
Applying the formula:

Max. number of consecutive races = Ln(123,261)/-Ln(1 – 0.34)
= Ln(123,261)/-Ln(0.66)
= 11.72/0.42
= 28.21
= 28 (rounded)

Therefore, there will be a maximum of 28 consecutive races without a winning favorite.

We can use a similar argument to calculate how many consecutive races there will be with a winning favorite.

Again, we use the formula: Ln (No. races)/ – Ln (1 – Strike Rate)

Where:
Ln – is the natural logarithm
Strike Rate – is 100% – 34% = 66% = 0.34%
Number of races – is 123,261

Max. number of consecutive races = Ln(123,261)/-Ln(1 – 0.66)
= Ln(123,261)/-Ln(0.34)
= 11.72/1.08
= 10.87
= 11 (rounded)

Therefore, there will be a maximum of 11 consecutive races with a winning favorite.

So, if favorites win 34% of their races, it is to be expected that there will be a maximum of 11 consecutive winning favorites and 28 consecutive losing favorites in 123,261 races.

As well as there being maximum winning and losing runs, there will be a host of lesser winning and losing runs.

Being aware of this information is one thing.  Taking advantage of it such that the appropriate bet (back or lay) may be placed at the appropriate time in order to make a profit is an entirely different proposition.  The problem with winning and losing runs is that it is impossible to determine when one will begin and when it will end.  Yet, in order to make a long-term profit, we need to know this information.

Although we are able to state that we will have winning and losing runs and are able to determine what their maximum lengths will be, we are unable to determine when such runs will start and when they will end.  As such, it will not be possible to determine which type of bet to place and when to place it.  Therefore, making a profit is impossible using this method.

Does this mean then we should abandon all idea of using statistics to assist us in our attempts to relieve the bookies of their cash?

No, not just yet.

Let’s explore this topic a little more.

I have analysed a large number of selection systems.  Some of these systems select horses which are to be backed to win, some select horses which are to be layed to lose.  Some of the systems only select favorites, some systems only select non- favorites and some select both favorites and non-favorites.  Although the systems have vastly different selection processes, remarkably, they have an important  property in common.

They are all cyclic.

What do I mean by this?

•    None of the systems that I have ever investigated win consistently over the long term.  They all go through winning and losing periods.

•    My investigation also shows that if a system goes through an exceptionally profitable period, an exceptionally un-profitable period will soon follow.

•    During an un-profitable period, systems will lose a sum of money which is at least equal to the profit made during its profitable period.

•    The longer that a profitable period last, the longer the un-profitable period will last for.

•    The cycles are neither period-dependent nor going-dependent (state of the ground).  Some tipsters often blame the state of the ground for their poor results.  My investigation shows that poor results are not dependent on the going.

So, what creates the cycles?

Luck.

Whether some people care to admit it or not, good and bad luck, over the long-term, comes in equal amounts.  In the short-term, however, it may not.

Much the same may be said of systems.  Some days, there will be lots of good luck and some days there will be an equal amount of bad luck.  Over the long term, things will balance out.

By way of an example: One day last year, I could do no wrong.  I layed a number of horses that day.  Most started favorite.  They all either fell early on, unseated their riders, or refused to jump – all except my last selection.  It was involved in a photo finish which took an age to declare.  The Betfair punters had already decided that my selection had won and that my bet was as good as lost, given the odds of the two horses involved in the photo.  My selection was about 1.3 on the photo on Betfair and the other was about 4.0.  When the result was announced, my selection was placed second and I had won my bet.  The previous week had been an unmitigated disaster for me.  I couldn’t pick a loser if my life depended on it.  Even if my selection turned out to have three legs and a wheel, it would still have won.
So, where does this get us, if anywhere?

Well, as I stated earlier, all systems are cyclic in that they go through profitable periods and un-profitable periods.  This is the key to success.  This is how we can make a profit and make our gambling pay.  Here’s how:

Firstly, we select and monitor a system.  It doesn’t actually matter which system we use.

Why?

Because all systems are cyclic.

We monitor the results of the system over a relatively long period of time.  Three months or 1,000 bets is usually sufficient – but the longer the better in my opinion.

At the end of this period, we then calculate the system’s long-term strike rate.  To do this, we divide the number of successful bets that we would have had by the total number of bets and then multiply the result by 100.  For example, if the system has been monitored for a period of three months and there were 1,029 selections during this period of which 983 of them won, then the long-term strike rate of the system is:
983 x 100/1029 = 95.53%.

Secondly, all systems are cyclic.  For this to be true, the short-term strike rate of the system must rise and fall.  Therefore, we need to calculate the short-term strike rate of the system.

How do we do this?

This is a two-step process.

The first step is to define what short-term is.  The second step is to calculate the short-term strike rate.

Let’s first deal with the definition of short-term.

If the short-term strike rate of a system is monitored and the short-term definition used is too short, we will find that the strike rate will increase and decrease like a yoyo and it will be difficult to determine what the short-term trend of the system is.  If the short-term definition that is used is too long, we will find that the strike rate will barely change with time and it will be equally difficult to determine what the short-term trend of the system is.  Therefore, the optimum short-term definition is found through trial and error.  Ideally, a period of time or a number of bets should be chosen which causes the strike rate to increase and decrease relatively smoothly.

I have also found that using a ‘rolling’ short-term strike rate is more effective.  For example, if a period of 3 months is chosen to determine the short term strike rate, then as the latest results are added, those that are more than 3 months old are discarded.  Likewise, if the latest 50 results are used to determine the short term strike rate, then as each new result is added, the oldest result is discarded to leave the latest 50 results.

At the end of each racing day, the result for each selection must be determined and added to the previous short-term results.

Also, at the end of each racing day, a new point must be added to a graph which depicts how the short-term strike rate varies with time.  From this graph, we can determine if the short-term strike rate is increasing, decreasing or has become erratic.

If the short-term strike rate is increasing and its value lies below that of the long-term strike rate, then the system is going through a winning phase and its used should be continued.

If the short-term strike rate is increasing and its value lies above that of the long-term strike rate, then, although the system is going through a winning phase, a losing phase is close.  It is therefore recommended that use of the system should cease.  Of course, results should still be collected and plotted during this period as though its use had continued.

If the short-term strike rate is decreasing, and its value is above that of the long-term strike rate, then the system is going through a losing phase and there are two options available to us.

The first is to cease using the system until the short-term strike rate begins to increase again.  Of course, results should still be collected and plotted during this period as though its use had continued.

The second option available to us is to reverse the system’s usage.  If the system is being used to back horses to win, it could be used to lay horses to lose.  If the system is being used to lay horses to lose, it could be used to back horses to win.  Of course, results should still be collected, plotted and treated, during this period, as though its usage had not been reversed.

If the short-term strike rate is decreasing, and its value is below that of the long-term strike rate, then the system is going through a losing phase but the system will soon begin increasing again.

If the short-term strike rate has become erratic, then the system is going through a phase where it is neither profitable as a backing nor as a laying system.  It is therefore advised that its use should be discontinued until such time as it becomes stable again.  Of course, results should still be collected and plotted during this period as though its use had continued.

Essentially, the above is the basis of a statistical approach to betting.  It involves calculating both the short-term and long-term strike rates and only using the system at the appropriate times i.e. when the short-term strike rate is increasing and below that of the long-term strike rate.

This system is so general that it can be used in conjunction with any other system whatsoever.  Not only can this method be used in conjunction with horse racing systems, the principles embodied in this article can be applied to other sports too e.g. Greyhound racing, football, etc.

Although this method may be used to increase the strike rate, and thus the profitability, of a system, it cannot completely eliminate losing bets.  No system can ever achieve this.

 

“GPWA Verification”