Forums

General Betting

There is currently 1 person viewing this thread.
zooot
28 Jun 16 14:25
Joined:
Date Joined: 27 Dec 04
| Topic/replies: 153 | Blogger: zooot's blog
data on 22000 horses, spreadsheet with 130,000 rows - lots of data.  Based on freely available data (which of course has lots of people looking at it) analysed in some quite sophisticate and creative ways.  Of course people with much more time consuming info / data collection methods would have info I was not accessing and so may find stuff I could not.  I personally chose the data analysis path rather than doing the tissue (form analysis) as my aim was to create a relatively automated approach with low time investment once the analysis was done - so the form experts out there are operating in a difference space to me and good luck to them.

This is what I found:
- most edges that seem strong with 6 or 12 months data disappear once you get to a coupe of years data cause they never really existed - they were temporary patterns
- stronger more sustainable edges tended to get ironed out with lots of data to be less than 3% and so make a loss after commission - what seemed like a 10% or even 20% edge slowly got ground down to not much when the long run of data was taken into account
- quite a few genuine edges remained but they only work with zero commission or at maximum 2%
- once you filter out the neutral and loss making situations from some remaining potentially/ apparently profitable edges (say 3-8% over commission) your huge volume of potential bets seems to shrink to just a handful a day or even one every few days  - so few it is hard to get volume needed to scale up and to test them or maintain confidence in them during bad runs that can last weeks or months
- lots of better edges sit in lower volume situations where your bets will move the market more, again limiting scaling up as people will copy you or the market will shift
- reversion to the mean is real and observable - edges can last for months, even a year and seem strong and sure but often suddenly plateau just as you get confident in them
- quite a few edges sit in higher odds zones where the ups and downs can be almost ridiculous, more like snakes and ladders  or a rollercoaster than steady gains.  The holy grail can appear before your eyes with strong steady wins over months and then all the gains can go in just a few days that make no sense but are really just normal streaks
-slightly over break even situations create long losing streaks that may be just a phase or could signal a market shift - which one ?  Can be costly to learn which ...
- other people seem to be either following in my footsteps or discovering interesting edges at the same time as me (probably simultaneously analyzing the same patterns I was and the same data) - A strong stable pattern of growth that you ride up and are just about to scale up on switches to break even in the space of a week or two and then straightlines like that thereafter
- I think betting markets are far more accurate and able to be studied than say the stock market which is much more emotional and driven by unmeasurable human perceptions and fake data so actually offers better opportunities than betting - much less perfect.
- by publicly releasing all the data it does BF helps ensure a more perfect market which is in its interests as the ultimate outcome in their mind is semi-breakeven punters grinding it out against each other with most of the money in the system being drained out as commission.
- stuff you would not believe influences betting patterns but often just slightly and not enough to easily exploit (I'll mention just one that has a small impact but never enough for me to make anything of and seems to vary in unpredictable ways so I don't think I am spoiling anyone's system - the number of letters in a horses name and the ratio between vowels to consonants in their name has a small but measurable impact on betting patterns at certain odds!

In short, over the longer run the markets are amazingly accurate and 5% commission is not just an annoyance, it makes most sophisticated betting approaches unprofitable.  The market adapts and shifts more and faster than you would expect.  Less liquidity, fewer mug punters, more crossmatching and betfair bots etc serve to flatten out edges over time and this is heading in one unfavourable direction.

So I leave after quite a few years, in significant profit still but with an average hourly earning rate over the years that would make a cleaning lady cry. I also have much improved Xcel skills that have helped me a lot at work.  Although I genuinely enjoyed the analysis I did, and trying all sorts of stuff, so am not bitter, ultimately it had to pay off in terms of profits as I am not a gambler and want a decent ROI for my time.

The point of this post?  Well mainly to spell out just how far you have to go to make money on BF long term and to highlight how what often seems like a system with simple analysis and short data runs  is probably not or will not last. 

So remember you have to go way beyond clever analysis of readily available data into detailed form analysis or highly sophisticated automated trading to make good money.  Or you have to get good at trading and spend a lot of tedious hours doing so.

Now with 2% commission there might be more scope but BF has gone to 6% and even 8% which is a joke in today's markets.  Lower liquidity, higher commission, more and more accurate markets is a tough equation to make BF appealing to the market.

Post your reply

Text Format: Table: Smilies:
Forum does not support HTML
Insert Photo
Cancel
Page 1 of 3  •  Previous 1 | 2 | 3 | Next
sort by:
Show
per page
Replies: 103
By:
LarpHager
When: 29 Jun 16 17:33
If you send me your historical data in an excel file, I can find something good in it.
By:
pxb
When: 30 Jun 16 02:48
zoot, that was very interesting, even though I never bet on horses.

I'll make one observation, With such a large dataset, any edges you can see are highly likely to be real, and the fact they dissapear after quite long time periods, means others are seeing them too. But I think you knew that already.
By:
zoom_top
When: 03 Jul 16 12:42
Hi zooot - very interesting post. It was pleasing to read a post which was not just whinging about this or that. I'm sure you have enough 'columns' to make an intelligent assessment. Your conclusion/s are in the minds of 90% of punters but that won't stop them. It's that pot of gold at the end of the rainbow that drives us all - the thought that there must be an undiscovered angle if only I could find it. I personally prefer the Greyhounds as the data is more consise. I come from a background of 30 years punting on Australian dogs. A friend of mine made over 3/4 million dollars some 10 years ago on Betfair over a period of about 5 years betting on UK Greyhounds. He told me that the turover was horrendous and the percentage profit he made was about 4-5%.(after Betfair takeout). He would not devulge his secret and just gave me sketchy details of his methods. Anyhow I decided 4 years ago to find out if it was possible. I am writing this as I feel like I have travelled some of the same tracks as you zooot. And so the 'adventure' began. I compiled one spreadsheet after another using data available at Racing Post. I ran parallel 'tests' using various systems. Some were very successful over say 3 months and when it came to putting the 'money' down they fell apart for seemingly no reason. I would scatch my head and say that's not possible. So I would re-write the system using the information I thought I had gained. Again greyhounds simply would not perform to my expectations. Where was I going wrong ? Was it grade, trap, midrace speed, beginning speed, overall speed plus a host of other factors all tried with various 'weighting' to determine which would work. I ran 'trials' over 3 months to 6 months. i still to this day try to find the elusive combination factors which overall will give me consistent results. I liken it to wresting an octopus ! My conclusion/s is/are that 'on the day' only some people know whether the greyhound is primed for victory and the off-course punter is not privy to this information.
By:
zooot
When: 04 Jul 16 14:56
Zoom_top
I had a good hard look at greyhounds (BAGS) and trialed a lot of stuff.  My conclusion was that their is inside knowledge for sure at work.  I logged data in the lead up to the off and found some interesting patterns that occur in the seconds before the off that pretty much spelled out the value bet in many cases.  But as you can't time the off exactly it is hard to exploit as it is split second moves with not much hint beforehand.  Like trying to catching a flash of lightning.  I also am sure that those with knowledge (both form based and inside track knowledge) push the odds around like crazy in the lead up to the off to create value for themselves through false favourites etc. and to lead everyone up the garden path then swoop in and snap up the value outsider.  But they also vary their patterns so that after a few days or weeks of a certain pattern, I think the do the opposite for a while and wrong foot everyone.  Once you have the knowledge to pin point true value or consistent track side info and a good bank it would be easy to do that sort of stuff either manually or most likely with a bot that has  3 to 5 strategies it mixes up randomly or in streaks to take the market this way or that sucking in the traders and the pattern hunters.  I am sure in greyhounds these factors lead to apparent systems / value being even more snakes and ladders than a non-manipulated market would be.

Years ago I read about dogs having things like two toes tied together with cotton or a bit of ginger up their bottom or sandpapering their paws to make them tender to create in the know advantage - who knows what goes on

On top of this is the huge natural randomness in all such events - apparent patterns can seem so real but be nothing but random streaks.  For example, if you set up a spreadsheet with 1000 hypothetical bets at say odds of 5 and set up to be roughly break even and then have a random number generator formula n a cell next to each bet.  Then cumulatively sum the results and graph the return patterns over time. Then calculate over and over and watch the graph create every shape you can imagine - long winning streaks followed by plunges to zero, steady returns for the whole set, zig zags up and down, a huge steep win streak followed by a horrific plunge - all due to just randomness.  This is very sobering to see and show what fools most people with short data / betting runs.

I am sure horse racing is well and truly having similar stuff go on. Some syndicates literally have teams of paid form analysts who specialist in certain tracks and compile odds.  They also have paid teams of high skilled programmers and analysts working out bots and odds strategies and have low commission due to their huge volumes that enable low margin strategies to work that lose money for average punters. With that sort of knowledge, why would you just stick money down as the market stands?  No, what you would do is push the odds around to favour you planned value bets for each race turning small value bets into huge value bets.  They would also be studying the same data I studied.  I was surprised in a way that I found as much value as I did in various angles but it seemed to me that once anything became too sustained over too long a period it was spotted by others (pros and people like me).  Some things remain that I found that seem to have missed them but they just sit in the 2-4% ROI before commission zone and so make a loss.

I know someone who used to breed greyhound who tapes every race on on two tracks and has his own rating system and wins consistently.  I tried my own system after a chat with him but found the value bets my crude efforts came up with to be way off the mark - I would lay something cause it seemed way too low odds and it would then steam in after I laid and go on and win.  I would have had to watch thousands of races over many years and get to know the trainers visit the tracks etc to move into his league - more investment than I have the time for.  He did not use the standard data available to all.  His knowledge of dogs came into it as well.

As for the pot of gold? Well it is more and more becoming the province of the few as time goes on.  The killer for me was realizing that I had lost confidence in any edge sustaining itself over time and being able to scale up to pay good money. I also realised how event such as horse racing, for all their randomness, are in a discrete little world that can be analysed very closely over time. Imagine the thousands of eyes staring down microscopes out there  at these little drops of pond water studying the movement of the creatures within.  Every horse fart and drop of sweat is logged smelled and analysed and built into someone's system. 

Some of the graphs my data produced of ROI mapped against certain factors produced almost perfect mathematical curves with barely a ripple in their lines.  I would look at them and just be in awe at how accurate the total market was.  Value exists in every race of course but can you grab that over and over?  You can out forecast the weather bureau on lots of days but their predictions over time of say daily maximums across the year will in general beat you by a long margin.

One system I walked away from indicates the challenge.  Over two year it was quite profitable.  But the number of bets was so few - around 15 per week.  Then in December 2016 after steady growth for 18 months it went flatline for 6 months.  Perhaps it will shift back into profit but can I be bothered doing all the analysis and placement for 6 months without earning a cent when for all I know it may be the end of the system or maybe a temporary blimp?  Stuff that.  Even it it did kick off again, I bet if I scaled up the market would shift to adjust to my bets and the value would go.

All very interesting but it reaches a point where challenge becomes a toil and other hobbies must be found.
By:
pablo-fanque
When: 04 Jul 16 15:22
hi zooot

I sent you a PM

thanks
By:
aye robot
When: 05 Jul 16 19:42
" the number of letters in a horses name and the ratio between vowels to consonants in their name has a small but measurable impact on betting patterns at certain odds!"

When you saw this - did alarm bells not ring????

What you're seeing when you see things like this is random clustering, it exists in all data sets and it is not the same as causal correlation. If your methodology shows up this kind of relationship then your methodology is flawed. If you do a statistical test and it backs up this kind of finding that tells you that the test is wrong - or that you're applying it wrongly.

How would you know the difference? Well if there's a causal effect then you should usually be able to say what it is - or at least posit some plausible options. Then when you've found that mechanism you should usually be able to make further predictions and test them against your data to prove it. In this case there is no plausible causation that I can see.

It's possible to observe an effect but not know the cause, but in that case you'd want to have a very strong signal before you believed in the effect. This is obviously a weak signal or you wouldn't be using the caveat "at certain odds." If there was some causal relationship that I/you/we might not have thought of then you would certainly expect it to be apparent across the full range of prices.

The maths of this is very straight forward and requires no statistical training, just apply the following:

small effect + no causation = b0ll0cks.
By:
pxb
When: 05 Jul 16 23:36
the number of letters in a horses name and the ratio between vowels to consonants in their name has a small but measurable impact on betting patterns at certain odds!

I can think of several reasons to do with human behaviour that would cause this.

And there is no necessary relationship between size of an effect and statistical significance. With very large datasets as we have here, small effects would be statistically significant.

Although I do agree that you should always look for a cause for any effect you find.
By:
aye robot
When: 06 Jul 16 10:31
Go on then pxb - explain how the ratio of consonants to vowels in a horses name effects its price, then we'll try to work out a way to make a further prediction to test the hypothesis.

If you're going to say something about certain types of names sounding more appealing and therefore getting over bet then I'm calling it in advance. In the last few years I did a research project botting up well documented psychological/behavioral phenomena in the conditions most likely for them to occur - for example I did a big study into prospect theory in in-running markets. I even called the robots Amos in tribute to Amos Tversky. There was no discernible signal. The description of prospect theory won a Nobel prize (along with other discoveries) but betting markets are sufficiently efficient to overwhelm it's effects in real time. That's why I don't believe that calling your horse "Toto" rather than "Tophat" can move its price.

More broadly - the approach is simply wrong. If you fish for "patterns" in data then you will find them, then you'll invent some spurious explanation for why they occur and an even more spurious one for why they "disappear". All of this is evident in the OP. A much better approach is to make a prediction based on a hypothesis then try to prove it. Genuine edges are perpetual and do not fade.
By:
bettinghelp
When: 06 Jul 16 12:17
I have a hunch that horses with Arabic names have a higher ratio of consonants to vowels than non-Arab owned ones, and since they're predominantly owned by wealthy Arabs, and are better bred horses with greater ability, they thus have shorter prices in general.

But if the hunch is correct, it's just a linguistic pattern visible within a set of results. Do I win a Nobel prize if I'm right? Or should we consider horses whose names contain Boy, Girl, Lass or Lad, who are invariably donkeys?  'Twasn't always thus, mind you.
By:
aye robot
When: 06 Jul 16 13:40
If that were the case it would be very easy to prove with numbers - just test all Arab owned horses regardless of their names to see if they offer extra value. No hunches needed.
By:
bettinghelp
When: 06 Jul 16 14:15
I didn't say anything about value. My hunch is about the possible relationship between the letters in horses' names and their ownership, which in turn might lead to one possible explanation (i.e.their generally shorter prices) for what appears to be betting-related patterns connected with the vowel/consonant ratio.

Since I don't have the data, I'll just have to stick with hunches for now.
By:
aye robot
When: 06 Jul 16 14:38
It's true that you didn't say anything about value - I was thinking more about the OP.

Even so I think what we're seeing here is a good illustration of narrative/causation type bias. The OP makes a suggestion that is almost certainly spurious but rather than see it for what it is there is a tendency to believe a story - even a not very plausible story rather than the numbers. Is it even true that Arab owned horses have consonant heavy names (or whatever)? I doubt it, and I certainly don't have evidence for it. Is it true that Arab owned horses generally go off shorter? I don't see why it should be, even if they are generally better horses that would usually mean they run in better races with other good horses so that should wipe the slate on that one. People will tend to believe a causal story - or worse yet an anecdote over evidence because of thinking biases, that's a weakness that we have to understand in ourselves if we are to overcome it.

For wider world examples think of the MMR controversy, "zero tolerance" policing or homeopathy.
By:
screaming from beneaththewaves
When: 06 Jul 16 23:26
My "anecdote" involves an independent betting shop owner telling me twenty years ago that races won by horses with long names were always welcome in his shop, since so many of his punters couldn't be bothered to write out their names. If true, then extending that over the thousands of betting shops in the country would cause such horses to drift to bigger prices.

For what it's worth, when I bought a yearling, I made sure her name boasted the maximum number of characters allowed for a thoroughbred (18 including spaces). She raced 10 times, won four races and I backed her for three of those wins. She drifted in price for nine of her 10 races, including all her wins, for which I backed her on-course with bookmakers who knew I was the owner, and every time her SP ended up bigger than the price at which I'd backed her. The only time she shortened was when the stable backed her (I didn't - she looked overtrained) and she had to be pulled up.

Purely anecdotal, proves nothing, but does make plausible zooot's finding that "the number of letters in a horses name ... has a small but measurable impact on betting patterns at certain odds."
By:
Capt__F
When: 06 Jul 16 23:33
or even Jesus saves
By:
DFCIRONMAN
When: 06 Jul 16 23:40
zooot You say :-

In short, over the longer run the markets are amazingly accurate and 5% commission is not just an annoyance, it makes most sophisticated betting approaches unprofitable.  The market adapts and shifts more and faster than you would expect.  Less liquidity, fewer mug punters, more crossmatching and betfair bots etc serve to flatten out edges over time and this is heading in one unfavourable direction.
=================================================================

Markets are "accurate" in that they are priced to a "book" %........but almost every race have "value" in the odds......especially BF odds.

There are "patterns".........numerous ones......but, in a highly competitive type race ( say a handicap hurdle race where say 16 runners)runners),it is unlikely that patterns alone would not clearly indicate a horse with a strong chance of winning . Numerous factors usually have to be considered and taken into account. Those are the factors that are not so "easy" to spot.

"Patterns" in certain types of races are more consistent.......say listed races where all horses are same age. However, the odds generally will be fairly low due to such "patterns" being known by bookmakers etc......and some "patterns" can easily be spotted by an average punter!

Probably at least 100 factors can be used to rate horses in a race from basic data (race card)..........What makes it more difficult for someone with little knowledge of horse racing in UK are the effects that GOING CHANGES have on race results.

It is highly complex looking for edges that can produce high ROI ....and may take decades before you find one. As life can interrupt such a search.......births, marriage, illness and of course your deathDevilCry.........so can take years before you find 1.

At least you had a go......Cool


BTW climate change may make many edges disappear.....but also some other edges to be found. Tongue Out
By:
pxb
When: 07 Jul 16 11:13
I have a hunch that horses with Arabic names have a higher ratio of consonants to vowels than non-Arab owned ones,

That's part of it. Different languages have different ratios of consonants to vowels, and quite a lot of horses have foreign names.

Rather than the horses being better or worse, I'd say it's simply that a significant number of punters aren't the most literate people and avoid horses with non-english names, because they don't know what it means, don't like the 'foreign' sound of it, can't spell or pronounce it.

Similarly, with longer names, they are more likely not to know what the words mean or be arsed to remember and write the name down as Screaming says.

Obviously, without knowing the exact correlations I can't be more specific.

I wouldn't look for an effect in running, which would be dominated by more sophisticated punters, less swayed by the horse's name.

In the unlikely event I ever own a horse, I'd chose an obscure and hard to spell word like 'eschatology gyre'.
By:
pxb
When: 07 Jul 16 12:03
or worse yet an anecdote over evidence because of thinking biases

I'd say you are the person dismissing evidence because of 'thinking biases'.

And anecdote is merely a small dataset. And note the anecdote presented here is consistent with the evidence in the OP. In scientific circles we would call that independent verification of results, albeit with a small sample.
By:
zooot
When: 07 Jul 16 13:54
On the topic of word length and vowel ratios, there is heaps of research that says these factors impact on readability in english. Do a google search on readability indexes.   Punters read, scan, memorise and get fixated on thousdands of horse names as part of the betting process.  Of course it would have an impact of some sort over the long haul in terms of tendencies to favour or not favour certain names and readability of names.

Words with lots of vowels are typically harder to read than those with a higher ratio or consanonts.

Betfair's UK listing pf horses by odds order changes the way it works compared to other markets.  It changes from country to country in impact due to cultural factors and the ways markets are presented.  Over time I also think it has shifted as fewer average punters use betfair.  The relationships were more clear a few years ago.

My findings were that the effects are small and lessen with longer odds.  They only impact pre.-off.  The effects are also more fickle than other form based factors.  It is an interesting thing to note but in my experience it will not make you any money.  I also think it would work differnetly on betfair compared to in bookmaker odds.

It is / was real but I would not recommend it as a factor or filter.  As it is so much left field to form factors I found it never supported other filters you may use and so does not necessarily help improve an existing system or data set.

I have not looked much lately as I lost interest in it but years ago horse names that were 13 letters long used to show a persistent kink in the data over multiple years and races types.  Up to 2014 this was quite pronounced.  I think it has faded as the markets got more and more driven by super sharp forma analysts.  It would make an interesting Phd topic for a statistics student or linguistic student.
By:
pxb
When: 07 Jul 16 21:34
Behavioural aspects of betting are of interest to me. Probably the first 'edge' I found on bf was purely behavioural, and nothing to do with the merits of the teams. Still works today, although perhaps not so well.
By:
zoom_top
When: 08 Jul 16 17:48
Hi Zooot-
My internet has been down for 3 days due to my choice of living in a regional area. Anyhow, I just thought I would let you know that I read your return post. I have no idea how this thread got on to horse name lengths acting as a barometer in any way at all - very curious ! Anyhow it is clear that you have delved very deeply into the punting business - far more deeply than I would ever aspire to. I have never really worried about bots, market manipulation or trading. Some would say - well if you're not a trader on Betfair you have Buckleys. My theory has always been that if there an angle or angles even within one race let alone over a period to which the general public has not given enough credit and I'm able to use my wealth of general knowledge about the nature/foibles of greyhound racing then I can isolate one dog which may be to my advantage. Well that theory has been tested to within an inch by the guile of the racing manager. I had never encountered such fiendish tight grading until I embarked on my 'journey'. 3/1 the field or better is not uncommon. In a field of 6 or 8 as it is here in Australia, I had never seen so many races where the percentage difference between the top 4 favourites was so miniscule or prone to fluctuate to such an extent that, should you be using that as a yardstick for the 'actual' chances you would be all at sea. These movements confirmed my theory that I could beat the dogs if I ignored these movements and approached it using my well established mores. So .... Just find the dog, wait for the best price that is being offered and click your mouse. That's the plan. The inside knowledge that I mentioned earlier was always going to work against me but I wasn't aware in the early days to what extent it was going to cause a problem. I am currently 'on a break' from punting while I think about my next move. I won't be backing any NSW dogs soon as the Greyhound Industry here has been banned as of 1st July 2017. Too much 'wastage' which is probably true.
By:
henryluca
When: 08 Jul 16 23:41
ZOOT

A very very good topic and a thread I hope is contributed to by many.

Initially some outstanding matters I would be pleased to understand.

1. What is considered to be an edge. Perhaps a definition would be helpful.

2.What is the freely available data you used.

3. What do you mean by sophisticated and creative ways

4. What was the [i]analysis
technique-for the data
[/i]


An understanding of these matters will help understand your findings and comments.

PS
By:
zooot
When: 09 Jul 16 08:34
Henry
- by edge I mean a consistent pattern of value where ROI is positive over long runs of races where signals / factors which can be measured repetitively from a data source that can be obtained consistently enough to analyse and to use to then predict future value.

- betfair publishes SP result everyday with a range of criteria listed such as morning odds, sp odds, volume etc. and this data goes back years (quite tedious to download to excel over long runs - it has interesting stuff but is a bit like a bunch of gold miners panning in the same riverbed over and over.  The gold thins out as the years go on.

- better edges would come from data that individuals or syndicates source themselves from race videos or a host of sources (very time consuming to develop and maintain for a lone punter). For example, one pro punter began years ago with a one page checklist of factors that were measured for every horse from video footage.  By now they would be doing that times 10 to maintain their edges in terms of data (that is beyond the resources of all but syndicates or fanatical full timers).
Or else you can use excel to log data live pre-off and analyse that.  That tends to produce huges volumes of data and leaves you chasing the mirages of often meaningless and manipulated odds movements.

- sophisticated and creative analysis - well that is for people to find themselves by analysing the hell out of data looking for patterns that mean something real not for me to give away.  Obvious logical things will generally have been found long ago and the value squeezed out of them.  There are still plenty of things left but as I have said they seem to typically sit below about 3% ROI before commssion in the longer run or, if profitable, leave so few bets in such special situations that it is hard to get volume.  Others, face long no return periods or sustained losing streaks (often dramatic downs and ups) that take the fun and confidence out of them (picture sticking with a theoretical "profitable" system that was losing money for say 6-8 months of the year)  Or else they sit where $$$ volume is lower and your bets in that zone start to shift the odds ironing out patterns. For example, a few things I found only worked if your bet was the first one matched after you placed it.  If someone jumped in front you missed out on getting matched at value, if you chased higher to get matched, the value was gone - small edges are so sensitive and fragile. Anything like that would not allow scaling up at all. A minimum stake bet might get matched, go bigger and bang the traders jump in front.
- analysis was using xcel forumulae to build up indexes for chosen criteria that signalled when to bet next time these were met.

A conclusion is that the market is:
- incredibly accurate long term as it is so heavily studied by so many people that most measureable factors are measured and factored in to hundreds / thousands or punting strategies.  As the average punter market declines this force that moves market into value situations is declining.
- the readily avaliable data has been mined so much that year on year it is declining in value (I think I observed live the death of some of the significant edges that remained over the past 12 months).
- remarkably easily moved by fairly small scale bets and / or persistent betting patterns  in value situations to move back to no value
- a plaything for some deep pocketed, low commission paying big players who have evolved to use computerised systems to shift markets in their favour to exaggerate value for them in ways that only they (and perhaps a few very sophisticated others)can know live and that are set up to confuse and mislead the average trader and punter.

Sadly, I think BF are well aware of this and see the sportbook as the answer at a corporate level and thus have shifted their strategy in a way that is speeding up what was probably a natural decline of the exchange.

Seriously I think the stock market has way more value than BF long term as a betting venue as there are fewer true measures of value - it is mostly about hype and opinion without the reference point of wins and losses and odds to create genuine value measures as is possible in sports betting and commission is lower (but is in general very over valued right now so does not mean value exists day to day or year to year - jumping in now in most cases may be a terrible idea but next year who knows).
By:
zoom_top
When: 09 Jul 16 11:54
Hi Zooot
A conclusion is that the market is:
- incredibly accurate long term as it is so heavily studied by so many people that most measureable factors are measured and factored in to hundreds / thousands or punting strategies...

Just looking at that statement for a minute ......

If we look just at the favourite in a high liquidity race we can plot the percentages of the field and assume then that we don't need to create our own mechanism to establish a market on the race or any similar races. We cam say that task has been done for us by the punting public? So unlike days gone by when bookmakers would employ people to set the prices, that task has now disappeared? The problem is horses being animals don't necessarily know where they are in the market and run according to their percentage chance. I have run tests on this - not the accuracy of what you are saying but rather around the expectancy aspect. So over a number of races the stats tell me say that the favourite, when 10% higher in the odds than the 2nd favourite will win X number of races I can establish a betting system based on this information. I think that what I'm trying to say is that if you statement is true ( and I'm not necessarily disagreeing with it ) then why is it that most of the betting public ( traders excluded ) are still backing horses which statistically have a very low chance of winning ?
By:
romfordiron
When: 09 Jul 16 11:54
Really interesting read Zoot. An elegantly summarised analysis of the difficulties in beating commission when punting. To me, I've never been able to see how anyone can make decent money from punting on an exchange, unless of course, they have inside information or courtside etc. There are still a multitude of ways to make money on an exchange, but the commission structure has always penalised outright punting.
By:
zooot
When: 10 Jul 16 04:45
Romfordiron

Yep commission levels is my main beef with the exchange - 5% is outrageously high relative to the profit potential for outright punting.  It may be necessary to pay for BF high cost structure.  Perhaps a super low cost exchange model could exist one day and offer 1% commission - that would kick life back into exchange betting.

I don't think most punters on BF know just how much a 5% leak is draining them. It is sort of hidden if you look at your statements. I always plotted my analysis of my history and data analysis to show cumulative gains / losses in graphical form and had a no commission graphs and one with commission factored in.  So often I would see a really promising pattern and then see it turn b/e or loss making with commission applied.  Commission is like a great big hot iron onto wrinkled fabric that smooths out all the value in the system.  Graphs without commission have so much more life in them (ups, downs, zig and zags etc)  Then you add in commission and they so often shift into a steady sloped decline.  If you get up to 7 to 8% commission as in some markets almost nothing survives in terms of profit.  Now some super high volume punters pay much lower commission than me and this gives them a huge advantage.  Combine that with a bit of manipulation and the board becomes tilted to their favour in a major way.  Commission discounts are a BF controlled discouragement to smaller punters.  They may not always be aware of it consciously but the steady commssion leak in their bucket of funds and need to top up funds tells them something is not right.  I know the cry of we need the big players for liquidity pop up but I would prefer thinner markets with more value and a level playing field and lower commission for all that a big subsidy for the syndicates.

You can reduce your commission by having multiple bets in the one race but that supposes you can find lots of value bets per race. My analysis kept ending up with too few bets to make this strategy work.

Zoom top
I have found all sorts of quite dramatic value patterns at low odds (and other odds) such as you may find or are investigating. For example there may be a sustained pattern where horses with odds of say 1.8 to 2.3 in certain situations produce a marked spike in my graphs that screams out that you have hit the pot of gold.  It can be like a hot zone where a sharp spike shoots up at 1.8 and then at 2.3 shifts back into loss.  You study your data set and it seems consistent across many situation and so must be real. So you target that spike and odds zone and wait for the profits to roll in.  Well what happens often is that it goes okay for a while and then suddenly hits a huge losing streak - almost as though someone is trying to play with you head or was watching you ready to suck you in and then pull the rug out from under you.  Well over time I have concluded that a lot of odds zones exist with hot and cold zones - the ROI from say 1.5 to 10 does not follow a steady return level on any given time period.  But what happens is that these zones shift over time in (for me) unpredictable ways.  Your hot zone can shift to cold zone in a space of a few days and the cold zones can be like black holes you fall into.  Such zones exist all over the place in betting in short runs of data.
They dissappear over the long run in any data analysis.  They tend to exaggerate both winning streaks and losing streaks compared to say a computer generated set of events.

They are a big factor in leading us all into apparent winning systems / approaches and then dashing our hopes with a sudden deep losing streak.  Where do they come from?  Well partly just from randomness - random numbers plotted will form all sorts of weird patterns that appear like trends.  But also human behaviour, systemised betting, weather etc. will tend to vary over time shifting zones around.  On top of this is manipulation, which if done well by big players, would purposefully try to create and destroy patterns to maximise the pain for those being manipulated.  A common stock market feature is that it tends to shift in ways that create maximum pain for the most people.  If you could control markets to some extent, you would try to wrong foot people as much as possible or periodically shift your whole approach just as people start to get familiar with it.

One comment I read once from a bookmake insider was that they periodically changed the whole methodology behind how they calcuated their SP (which is a key reference point for all of us)  so that it stuffed up all the punters analysis.  Even if it cost them short term, wrong footing the market long term is very important.  Well the same philosophy would be built into syndicates approaches.  Don't use the same strategy too much purposefully let an approach run for a while then do a big switch around.  For those who doubt such syndicates exist, there is plenty up material in the public domain including tax investigations to prove they do.  If you made up say 25% of the market in a race on the exchange and you had quite accurate form analysis enabling you to create a book price for each horse what would you do?  Place lays down exactly as per your book? Or would you try to exaggerate a possible favourites low odds by clever use of betting patterns and push out the odds of a genuine long shot then cumulaye lays of the favourite and backs on the outsider?  Course you would try to shift it in your favour.  Which creates value if you too can analyse form and spot the value.  But for data analysis approach like mine this creates false patterns and unfavourable losing streaks in profitable systems that damage confidence.  In the stock market one person summed it up like this "there is a big club and you are not in it".

In greyhounds you see it all the time in the fast moving thin markets. So often the early drifter suddenly steams in.  Close to the off the value bets start to become apparent as dfriters steam in.  But the by the time the evidence is there the value is long gone.  Or it happens in the last 15 seconds before the off in a massive swoop of value by those in the club.

What I am saying is that when you analyse things and decide to lay all horses between odds of x and y who have done xyz in their last two races and who farted twice in 2015 in the final straight before losing narrowly, you may have something, or you may have nothing.  If you have something, it may be that in two weeks time for factors you may never understand, the odds zone x and y may have shifted to be x+.6 and y+.6 and so all you bets below x+.6 are now in a black hole zone that kills your profits and only become apparent weeks later when you re look at your system with more data.  Then you shift up to the new zone and it may or may not hold.

Over time all these zones melt into a long run data set and become little kinks in it that hint at value or may provide a few % one way or the other but not enough to overcome commssion. In my experience value exists more often between 10 to 30 odds for example than witn favourites.

Now if you did true quality form analysis and understood racing like some forumites appear to, all this discussion is immaterial as you are playing a different game and can take advantage of value spikes when they occur on a horse by horse basis and just not bet when things are manipulated or zoned out of your value zone.  I have considered that path but don't think I have any special abililties in that area and my aim was to produce something that was not too time consuming in the long run.  Form analysis or self collected data is very time consuming. So I tried to make the most of available data to avoid that commitment and faced the limitations of that approach.  Trading also works but I've done that before and seem to have reached a point where I can no longer face the tedium and frustrations of trading (was fun years ago when it was new for me and easier.
By:
zooot
When: 10 Jul 16 07:24
Another thing you must genuinely take into consideration is your tolerance for losing streaks.  If you read up on betting you know they will happen but reality is quite different.  For example

System A laying

From September 2014 to April 2016 made ₤2500 after commssion with odds 6.6 to 15 with 3800 bets risking on average about ₤100 per lay so ROI over that time 0.6%.  Thin pickings after commssion but just profitable and ₤2500 is nothing to sneeze at.  Note you have risked ₤380,000 pounds to win that money.  The worst losing streak took ₤1000 off the earning peak.  You just don't get steady returns.

Now between April and July 2015 it made nothing with a flatline pattern
From September 2015 to April 2016 it made nothing with two deep losing streaks and one recovery streak

Since April I don't know - it is another pause? Or the end of the system?

Looking at the overall graph of cumulative return for the whole time period you would say it looks okay. But those periods of no return are very long. Months and months plugging away earning nothing. That is the reality of low profit systems.  Sitting through them is not a thrill a minute.

Now if you KNOW they are temporary it might be okay but given so many trends fade back to the mean the chances are that long pauses or declines in return are the end.

Quite often people kick off their systems or scale it at the peak of a trend once they have enough of a track record to justify them.  Then in the natural cycle of trends, it is more than likely that this will be when a pause or decline sets it.  That is fun and has probably stopped many good approaches in their tracks.Betfair was never easy but in the goldrush early days miners could stumble upon viens and nuggets all over the place.  There was more room for the excited explorer to find something easily or pioneer a new approach.  Now it is for big boys and clever niche operators all digging away in the same well panned river bed.

All just stuff to keep in mind as you chase the pot of gold so you don't waste too much family time, money or fun in the real world learning lessons that are summarised above.
By:
lewisham ranger
When: 10 Jul 16 11:27
how depressing is this thread Laugh

have a punt, win or don't win, it's a game, just like life's a game Cool
By:
zoom_top
When: 10 Jul 16 13:37
Hi Zooot-
Your approach to statistical analysis, punting and emperical realisations/implications is quite different to the way I approach the endeavour. I have enjoyed reading a different point of view and have a more sombre/sober attitude as a result of it. I can visualise you giving a lecture on the subject to would be investors/punters. I wouldn't think to compete with you intellectually as I feel I would be out of my league. I wonder though, if punting is not now possible for you, if you were to totally embrace your findings that is, where to now ? All avenues for investment, particularly here in Australia, are very very competitive. The Stock Market is not the gold mine it once was and you certainly won't make much on bank deposits - term deposits !!  Do you wipe the blackboard and take a fresh approach to Betfair betting ? Do you still believe that you will uncover something you might have missed or is all hope gone ?
By:
starship
When: 10 Jul 16 13:56
i make a living via punting.
i do one greyhound track
i get the dvds
study the form
look at a race, cross of the dogs i dont fancy
i bet at BETFAIR ON THE PRICES of the other dogs.
i could bet a dog that i make 3rd fav because of the price on here
been doing it for the last 20 years.
was a bit easier before betfair, when u could get a decent bet on in the shops.
them days are gone.
so far this year i have had 6 BET365 Accounts closed.
the ricks they make on the early prices are frequent.
will post on tuesdays the ricks,
if u have a BET365 accounts, bet the ricks and if u like, arb back on betfair
By:
zoom_top
When: 10 Jul 16 17:26
ricks ????
By:
starship
When: 10 Jul 16 18:17
mistakes.
ie yahoo victor was 7/1 on the early prices.....i made it 2/1
sp was 9/4
By:
zooot
When: 10 Jul 16 23:12
Zoom top

Starship's approach is the way I would go if I got into it again. I know someone who makes good money doing something similar.

Specialise in tracks or race types and track factors that are hard to measure and on information that is not generally available to the mass market.  That is my point  in these posts - its hard and so you have to go way beyond the obvious or the easily available info to make money. 

A flaw in my approach of lots of data analysis is that I have not had the extra dimension someone like Starship would have where he can spot value on the day and pick the eyes out of a days racing.  Data analysis approaches done my way gives you a generalised value set on any given day but I had no ability to say anything was above or below its true odds for the day or spot a pushed out odds situation by the manipulators. 

The guy I know who studies greyhound form will say " I spotted a dog at odds of 20 that should have been around 6" and he jumps on and leaves the rest alone.  My approach on the same day would have probably had a whole mix of bets from poor value to good value all mixed up with a light overall value edge.  Whereas he would usually be getting well above the value point most of the time on much fewer bets.

You can make money trading as well and with that at least you know very quickly if you are making money.  When trading most days should be profitable.  If not you have to improve you technique - so you won't waste lots of time on dead ends.  Just don't chase and have the patience of a saint.

Want to make money:
- trade
- go the form approach and become an expert and learn how to price up a race
- create some amazing data set of your own info but remember the points above still
By:
zoom_top
When: 11 Jul 16 13:58
Thanks zooot I won't shoot myself after-all Cool
By:
pxb
When: 11 Jul 16 22:10
Specialise in tracks or race types and track factors that are hard to measure

That applies in other sports as well. I am well aware that people are now using predictive numerical models. Even bf now publishes the results of theirs. I also see indications that there are momentum bots similar to those on stock markets. And who knows what else.

There are still people making money on here without collecting and analysing data. I know because I'm one. But you need to specialize.

Thx again zooot. Very interesting.
By:
siwaadupa
When: 12 Jul 16 12:59
Friend of mine received 3800(this includes holiday pay) Euro for 35,5h work per week assembling BMW 7 series(ok, not easy job which requires physical abilty- mainly walking)- which gives you around 160h per month and you have guaranteed holiday pay(28 days) bonuses paid plus you put money to pension pot. When I see people here, chasing for peanuts or they are happiest in the world because they have bots which gives them 5 pound per day it just amazes me.
Peasant assembly workers, very often without school have more in BMW than best shark traders here, who call themselves sports traders with salary lower than cleaners, even uber pays more now! I got caught paying surging charge 2.2-27km ride with uber gave him 80Euro gross for half an hour(70Euro net).
Keep wasting your life for profitable system here ;)
By:
screaming from beneaththewaves
When: 13 Jul 16 00:12
For years I've been posting how the ridiculously high commission rate is killing the exchange.

Imagine you're a decent judge and a decent punter who has a few £100 bets every day. Not trading, just straight backs or lays, taking a view. Exactly the sort of punter the exchange needs. Your commission discount is an apparently reasonable 20%, making your effective commission rate 4%.

Being a decent judge, at the end of the year you've come out on top, to the extent that your gross profit before commission is £3500. You're plainly a shrewdie, in fact one of the very shrewdest punters in the land if you can consistently beat the overround to that extent over a long period. So what does you Betfair account say your net profit is after commission?

I'll tell you: MINUS £3500.

How the hell can you, a winning punter carry on? Never mind the losing ones, who'll have paid very nearly the same amount of commission as you have, but for the privilege of losing on top of it.

All the whining about the Premium Charge on here is a red herring. Betfair love the Premium Charge whiners, because a tiny, winning minority have succeeded in drowning out and concealing the real Betfair rip-off: of the huge, losing majority.
By:
pxb
When: 13 Jul 16 09:04
For many people I suspect, including me, bf is just a hobby, with the bonus of making some money and giving sport some extra interest. It was intellectual challenge for me. I don't need the money.

I doubt there are a hundred true professionals left on here. And as I said earlier I'm tired of doing stuff just to keep away from the PC. It's a chore (the novelty of figuring out how to do it has worn off) and I'm likely to leave soon because of it.
By:
romfordiron
When: 13 Jul 16 10:48
I agree that commission is an issue in attracting punters, but it has been there since the start, and was there during the boom times. For many price sensitive punters, it's a question of weighing up the cost of commission against the slightly worse prices and ability (or lack of) to get a bet on with the bookies.

PC, however, is also forcing punters away indirectly by pushing liquidity providers away. Once main markets are illiquid, there's nothing to attract big punters. The dead side markets are meaning the small £10-£20 punters are looking elsewhere.

The lack of advertising for the exchange is obviously a further issue.

None of these issues are likely to be resolved soon - it'd cut short term revenues and the business has 2 sportsbook brands to promote.

I can foresee there will always be betting exchanges - individuals and bookies will always want to hedge at the true market rate, and it'll be there for the big bettors who can't get on with the bookies. However, I think it will be a much more streamlined operation.
By:
lewisham ranger
When: 13 Jul 16 12:00
most premium charge complainers are surely just bot users.
Page 1 of 3  •  Previous 1 | 2 | 3 | Next
sort by:
Show
per page

Post your reply

Text Format: Table: Smilies:
Forum does not support HTML
Insert Photo
Cancel
‹ back to topics
www.betfair.com