Favs and losing runs

[thorns]

Can anyone run a database and maybe provide some stats?

On a long losing runs of favourites, do they become poorer and poorer value? Was thinking the other day that with the amount of people/systems that chase favs, after a loosing run of even 3-4 favs, would the odds on offer versus the true value of the favourite start to be affected?

What got me thinking about this was we were at the casino the other day, and there had been a run of 5 reds, so naturally, everybody started throwing all there money on the black because supposedly it was due to win. As it turned out, there were another 4 reds in a row, which quickly bankrupted a number of the players at the table who kept backing it.

Any thoughts on this, do you think this would apply to racing as well, and could we use this to our advantage in laying favs to lose?

[UselessBettor]

Think of it this way:

On 1000 spins of a wheel you will get 500 red and 500 black (assumign no green on roullete wheel).

So if you had 5 red in a row. Lets assume that we always get 500 reds every 1000 spins. So we have 495 more to go in the next 995 spins. Your advantage has now gone to 500/995 for black which is 50.2%

The above assumes that exactly 500/500 comes out for each colour and that there is no green on the wheel. If you include the green then your still at a disadvantage.

Now your talking favourites. They don't have a 50/50 chance. They are closer to 30%. So lets go 1000 races and you had 5 favs in a row lose. That means your expectation is now that 300 will win from 995 bets. Your increased chance has gone from 30% to 30.1%.

[Chrome Prince]

I've seen 15 reds in a row at Crown, certain cultures love going with patterns and walked out with plaques rather than chips.

[norisk]

I don't have any stats but I believe that on average the starting price of winning favourites increases slightly as your average race card progresses, which may suggest that they do not present poorer value on a losing run.

Just my observations.

[thorns]

I understand the probabilty stats, I'm more interested in wether the market chases favs on a losing run, hence making them under the odds, seeing as they are not fixed as per the casinos and fluctuate in accordance to the weight of money being bet on them, SR has nothing to do with what I am asking. So many punters target favs, and start doubling up, or betting larger as each one looses, as in their mind 'the next one is bound to win,' which those of us with a basic ground in stats and probabilitys is a comlete fallacy.

But, maybe i cant put across clearly exactly what I am asking??

[thorns]

I don't have any stats but I believe that on average the starting price of winning favourites increases slightly as your average race card progresses, which may suggest that they do not present poorer value on a losing run.

Just my observations.The SR usually drops though compensating, and generally the later races are more open handicap type races, instead of maidens full of no hopers at the start of teh card.

[norisk]

True, but the longer price may also suggest that it is not affected by those chasing the fav, doubling up or whatever.

[thorns]

Could be the case, but ultimately the ROI needs to be taken into account to see if that is the case or not. You may be getting $5 about a favourite in the later races, but if its really a $6 chance, your going to have a poor ROI backing them if you see what I am getting at.

[woof43]

Until you have classified how each Favourite became the public favourite (by a variable or combo of two) and then the strength of "gap" between the 2nd Fav. basically you will never know.

But to answer your question wagering bias does occur but not how your thinking.

Best tip
Start handicapping your fellow punter not the horses. In wagering there are only a few strong variables at play

Joe Public will play those types (per variable and gap strength) of favourites over and over the same way day in day out.

[norisk]

Well if we go back to your casino example, what did the red & black boxes look like on that 9th red spin? Red box choka full of chips & the black basically friendless I am guessing.

If similar scenarios were to play out at the track/tab, we should expect to see favourites well over bet in later races, & in my experience we often do, but there is enough 'smart money' around to keep said favourite at a decent price - make hay while the sun shines & all that;)

So what I am saying is that no, in general losing runs do not have an effect on the SP of a fav.

[UselessBettor]

Until you have classified how each Favourite became the public favourite (by a variable or combo of two) and then the strength of "gap" between the 2nd Fav. basically you will never know.

But to answer your question wagering bias does occur but not how your thinking.

Best tip
Start handicapping your fellow punter not the horses. In wagering there are only a few strong variables at play

Joe Public will play those types (per variable and gap strength) of favourites over and over the same way day in day out.
What a very insightful post. Woof43 could you elaborate a bit. This is something I never ever thought of (I guess in a round about way maybe).
Can you give an example of a race and the variables one should use? How do you rate your fellow punters (overbets, underbets) ? Any insight would be very interesting.

[norisk]

seconded

[Bhagwan]

The stats show that the average price of the Fav in the 8th race is higher than

the 1st race & corresponding races there after, on the day.

One would approx break even targeting the Fav on the last race betting level stakes.
The SR is approx 24%, yet its av price is around $4.00 mark.

This 24% SR on the 8th race is a lot lower than say race No.1 onwards

The average price increases on a sliding scale from races 1-8.
and its SR drops from races 1-8 correspondingly.

[The Ocho]

I've tried to look all over the place (and I know I should know this :rolleyes: ) but what is the generally accepted statistic/strike rate of faves for both the win and the place?

I thought it was 33% and 66% but then some have said 30% win. I still don't know the place percentage though.

Any one know?

Thanks.

[Bhagwan]

Approx 64% for the place where there are not dual favs

[The Ocho]

Thanks Bhagwan :)

[Chrome Prince]

Here's a breakdown win and place % for favourites by race and Win POT %.

Race 1 35.95% 66.71% -13.73%
Race 2 33.95% 65.12% -13.86%
Race 3 33.56% 64.40% -11.86%
Race 4 31.94% 62.97% -13.72%
Race 5 30.96% 61.89% -13.39%
Race 6 30.15% 60.57% -12.23%
Race 7 28.20% 58.86% -14.95%
Race 8 27.73% 57.83% -14.10%
Race 9 26.73% 56.67% -15.33%
Race 10 26.36% 56.33% -12.85%

All Races 31.54% 62.27% -13.50%

*12 years Metropolitan, Country and Provincial data.

[The Ocho]

Thanks CP. The faves sure do head downhill strike rate wise in the later races however, as said in the thread, the prices may then rise to compensate?

[Mark]

All been done before and is very useful to know, but will the majority put this knowledge to use?

[darkydog2002]

I,ll hazard a guess and say NO.

[partypooper]

Strange, an ancient systematic approach from the UK was to target the Favs in the 1st 3rd and last races. Looking at Chromes' list maybe there's some thought behind it after all.

[Chrome Prince]

To address Mark's observations,
I don't think many will take advantage or for that matter should take advantage as they are only raw stats. The picture changes somewhat when adjusting for certain other reasons.

For example, field sizes, and jumps races are often the first race in winter.
Class races are usually later as well, so more often open events price wise.

Note the win %, place % and POT all decrease accordingly, not denoting overbet horses, but denoting average field size.
In general, increase in average field size relates to increase in average price, so naturally the loss percentage will increase.
Due to the favourite longshot bias, the longer the favourite price, the more percentage you lose. The shorter the price of the favourite, the less percentage you lose.
These are general comments only, obviously there are specific differences, but I certainly wouldn't be rushing out to back the favourite in a maiden in field size of 4, when I can bet on a good horse with good form in a Group One race in a field of 4, that just happens to be race 7.

Just my thoughts.

[Mark]

Agree 100% CP.
I take note of those stats and I also use short-term stats. eg yesterday on Syd, Melb, Gold Coast races up to race 6 there had been 7 favs & 10 in-the-market winners, this told me that some "bolters" were overdue and so I laid accordingly. Races 7 & 8 at those meetings were won by 1 fav and 5 good priced winners....thank you. Of course it's not alwasy that simple but it helps to know roughly where you're at.

[woof43]

What a very insightful post. Woof43 could you elaborate a bit. This is something I never ever thought of (I guess in a round about way maybe).
Can you give an example of a race and the variables one should use? How do you rate your fellow punters (overbets, underbets) ? Any insight would be very interesting.

Just seems like yesterday (2003 or around that time when I raised this same issue, fast forward a decade and here we are again).

First up, most this isn't really going to be helpful for the "backfitters" here as most continually mix Strike rate and ROI together. Remember Strike Rate only applies to Handicapping/Rating and ROI applies to Wagering.

It's not too hard to develop a list of variables or a combination as all have been listed here previously in database threads and just putting on your "joe public thinking cap" instead of the computer handicapper cap, you should be able to develop your list.
To know how the crowd looks at all handicapping variables, you have to test each one and then see which one or combination they use or the ones they don't use. To do that you need to build a generic , single factor simulator that can give you probabilities studies for each factor. All you will do once you have run a test is to determine the relationship between each factor and the crowd odds (not strike rate, not ROI just plain simple ODDS) Using a simple application of multivariate analysis will do this easy.
A simple test is to use the above to determine a set of "win" probabilities and then plot them against the actual post time odds in a scatter diagram, all you do then is to use Correlation Analysis and Correlation Coefficient and you will be on the road to success.

Just remember. for crowd "measures" accurate means close to actual odds , not observed finish percentages

[Bhagwan]

Everyone does that already .
It works good
They are called Bookies.
They make a living out of it.

But can you pick a little vinner.

[UselessBettor]

Just seems like yesterday (2003 or around that time when I raised this same issue, fast forward a decade and here we are again).

First up, most this isn't really going to be helpful for the "backfitters" here as most continually mix Strike rate and ROI together. Remember Strike Rate only applies to Handicapping/Rating and ROI applies to Wagering.

It's not too hard to develop a list of variables or a combination as all have been listed here previously in database threads and just putting on your "joe public thinking cap" instead of the computer handicapper cap, you should be able to develop your list.
To know how the crowd looks at all handicapping variables, you have to test each one and then see which one or combination they use or the ones they don't use. To do that you need to build a generic , single factor simulator that can give you probabilities studies for each factor. All you will do once you have run a test is to determine the relationship between each factor and the crowd odds (not strike rate, not ROI just plain simple ODDS) Using a simple application of multivariate analysis will do this easy.
A simple test is to use the above to determine a set of "win" probabilities and then plot them against the actual post time odds in a scatter diagram, all you do then is to use Correlation Analysis and Correlation Coefficient and you will be on the road to success.

Just remember. for crowd "measures" accurate means close to actual odds , not observed finish percentages
Woof,

I understand everything your saying above except for one point and thats my fault because I just can't get my head out of the sand. Happy to discuss in email if you don't want to post it here. You can email me at

betfinder@exemail.com.au

The part I am missing is how does this help us profit. Lets say we take barrier position and we find that it has some correlation to the win odds. (I haven't tested just picked a variable at random). How does this help us when analysing the race ? Lets assume we find the correctation is X*1.222222 (made up) which means barrier 1 has avg odds of 1.22, barrier 2 has avg odds of 2.44, etc.

How do we use this information to profit ? Do we now disregard this variable because we know the public uses it ? Or do we somehow put this information to our own use for laying/betting.

[woof43]

Once you have tested and found all the variables that work to some degree, that is the odds step down smoothly, you start to then categorize the favourites. This where you earn your money, there are infinite ways of doing this, i'll leave this up to you, but do a lot of thinking, success depends on how well you can learn to categorize races.

Your next step is to then go back thru your historical database for races that match each categorization. Then from your actual observations (statistically speaking), you then develop a probabilty matrix for every runner in each race or down as far as you can.

Should'nt be too hard to work out where it is all going now.

Obviously the categorizing is the key. pick a bad category you botch the race, but fortunately there is a fair amount of "forgiveness", but once you study this a little and eliminate the obvious bad way of doing things, it's not too hard.

I remember in the late 90's this type of categorizing was being developed where they had limited set "probability" tables but now with super computers set ups such as in HK, can develop far more complex catergory models and then "on the fly" search thru your database spaces searching for exact matches.

Something to think about

[Chrome Prince]

I'm at a bit of a loss also.
There are so many variations, and subsets that surely you'd need an infinite and changing matrix.

[Bhagwan]

Seeing that the 1st & 2nd fav win 50% of all races .

My dont you just separate them using all that stuff .

Bingo
50% winners.
One will be a millionaire by lunch time next week.

[The Ocho]

And Leon's getting LARRRGGGGER

I can't remember what happened in The Matrix but I remember Flying High.

[UselessBettor]

Once you have tested and found all the variables that work to some degree, that is the odds step down smoothly, you start to then categorize the favourites. This where you earn your money, there are infinite ways of doing this, i'll leave this up to you, but do a lot of thinking, success depends on how well you can learn to categorize races.

Your next step is to then go back thru your historical database for races that match each categorization. Then from your actual observations (statistically speaking), you then develop a probabilty matrix for every runner in each race or down as far as you can.

Should'nt be too hard to work out where it is all going now.

Obviously the categorizing is the key. pick a bad category you botch the race, but fortunately there is a fair amount of "forgiveness", but once you study this a little and eliminate the obvious bad way of doing things, it's not too hard.

I remember in the late 90's this type of categorizing was being developed where they had limited set "probability" tables but now with super computers set ups such as in HK, can develop far more complex catergory models and then "on the fly" search thru your database spaces searching for exact matches.

Something to think about

I am going to assume there is only a small subset of variables that step down semi smoothly with the odds. Barrier is going to be one I am sure.

When you say categorise favourites/races I am a little confused. Im not sure what you mean by categories. Are we categorising favourites based on the variables we found ? or based on something else ? Im confused.

hmmm ... I jsut reread it and I am still confused. Im trying to work out how this is different to just a set of mechanical rules ?

We have categorised things based on odds and not profit .... But how does this help. sorry I know it muyst be frustrating for you to try and point this out to me but its frustrating the ******** out of me that I can't understand it.

[TheSchmile]

Hi Uselessbettor,

If you don't mind I'm going to send you a quick email with a question?

The Schmile

[Chrome Prince]

As I understand it, categorising is probably the confusing term here.
I think an actual to expected ratio could be what woof's implying (I could be wrong).

This information is calculated by working out the actual / expected ratio for each parameter. The actual / expected ratio is like an enhanced Impact Value which takes into the account the price of each horse.

You can work out the actual / expected ratio (A/E) by dividing the actual number of winners by the expected number of winners. The expected number of winners is worked out with the following formula:

Expected Winners=sum of all 'odds' where odds=1 / (SP + 1)

If you exclude any bookies over-round for now, then the odds of each runner is 1 / (SP + 1). An Evens shot is therefore expected to win 0.50 of the time, whilst a 3/1 shot is expected to win 0.25 of the time. If we had 100 3/1 shots, then we would expect 25 of them to win (100 * 0.25).

The A/E ratio can be used to show how efficient a parameter is. A figure of 1.00 means 'as expected'. A figure less than 1.00 means not as expected.

In situations where the figure is above 1.00 this means that the parameter is being underbet by the public.
Figures below 1.00 indicate that the parameter is being overbet by the public and is not winning as often as it should.

There are flaws contained within, you cannot account for unknown differences, you can only identify underbet or overbet by joe public.

I abandoned this, because the unknown factors are more than three times the influence on price, than the known factors.
Using the above can identify what market order should be and even price, but does not account for the most important factor of all.
The fluctuations in betting.

The differences are quite apparent when you take all favourites at Top Fluctuation, drifters at TF, and Firmers at TF.
This is exactly why, bookmakers won't let you take TF on firmers once the betting has commenced.
Joe Public has been betting on this horse, but Joe Private has lumped ten times the amount on the same horse for reasons that aren't in any database, formguide or video.

[UselessBettor]

Are you sure that is what he is talking about ?

If that was the case you can do it with any combination of variables.

Im not so sure thats what woof was trying to describe.

[woof43]

Are you sure that is what he is talking about ?

If that was the case you can do it with any combination of variables.

Im not so sure thats what woof was trying to describe.

Exactly..

All your first steps are, is to categorize the Fav and how strong or the gap over the 2nd fav. and to find correlations in the odds to finished odds nothing more or nothing less.

You will be testing as many variables (in formguides or tipsheets) as possible to find ones with a correlation and know which ones don't.

I didn't tell you, but once you find a variable or combination, you would then breakdown this down into subsets of races based on the fav odds <3.00 etc.

Of course there is holes in what I have been explaining, the above approach is to provide very good first order approximation to those precious crowd odds.

Just with the above information I'm able to provide trifecta combinations that provide positive expectations for each race catergory.

Once you have completed the above your then ready to move onto actually using your own handicapping in "cluster analysis".

[The Ocho]

Is this some sort of Twilight Zone or something?

No wonder I'm no good at picking winners OR losers. :rolleyes:

If that's the sort of detail one has to go into I am obviously doomed to failure and should just stick to trading.

[darkydog2002]

Hey Ocho,
Its called Bamboozling one with ****.
Are you sure that one of these posters is not Mistermac?

[The Ocho]

Hey Ocho,
Its called Bamboozling one with ****.
Are you sure that one of these posters is not Mistermac?
Mistermac? I feel like a Big Mac now.

But I actually feel like Mister Magoo who can't see where he's going with all that bamboozling that's going on. :D

[rails run]

Mmmmm pickles!

Woof has a nice system that has evolved from 1000's of hours of work going back many years. To simplify it matches race types against previous fav results and strongly factors observations less utilized such as 'turn times'. I suggest it also weighes heavily on determining 'false favs' as the races to bet up on with tri's/quinella's/exacta's, etc. He will continue to do nicely from it and he'll prosper from his effort.

I understand his study but prefer to keep rules to a minimum and keep the system much simpler than many of those posted here. A bit like Darky. It makes the data easier to interpret and adjustments become obvious.

To try to give an example of keeping things simple, does anyone share the belief the 6 box is automatically disadvantaged in a dog race for whatever reasons? There's a pure gut observation shared by most and it's surprisingly not hammered out yet by punters.

[The Ocho]

I thought the same thing about the 5 box on the UK dogs (6 runners) where the stats for most tracks also agreed. Suffice to say that when it came time for ME to lay them it didn't pan out too well. :(

[rails run]

Hi Ocho
Yes, UK greys are a different scenario all together. 6 RUNNERS and I've read all boxes even out for the win. But AUS is a squeeze in the middle.

p.s- Keep up the trading as you've found your gift!

[Chrome Prince]

To try to give an example of keeping things simple, does anyone share the belief the 6 box is automatically disadvantaged in a dog race for whatever reasons? There's a pure gut observation shared by most and it's surprisingly not hammered out yet by punters.

No it's not automatically disadvantaged.
Depends on the track and distance, but most importantly the racing style of the other runners versus the runner in question.

But the stats may show different because they incorporate those not disadvantaged and those disadvantaged.

It's important to take note of how many races were not disadvantaged and the strike rate, versus those disadvantage and the strike rate.

It's the same with most scenarios, only punters are not good at splitting data in general, only the pros.

[rails run]

Hi Chrome
I agree the market will factor most things into the odds. I just love laying 6 at odds below $6 and watching it get smacked frequently in the first 50 metres.
By the way, your work on IAS ratings is a gem and showing 9.4% POT. Thnx

[Chrome Prince]

Looking at 90 dog races today:

box 6 won 8 of them or 8.89%
box 5 won 15 of them or 16.67%
box 4 won 6 of them or 6.67%

Illustrates box 5 is not twice as good as boxes 4 and 6, no way.
Listen to the audio or watch the video and it's not hard to see that it's not the box, it's the makeup of the field.
That's why champion dogs can start in any box and win, average dogs need a good draw and luck.
Poor dogs, doesn't matter what box, they need pure luck.

[rails run]

I'll add I lay 6 between $3-$6 to avoid the really good one's you're referring too. Out of the measly 8 winners today how many fell within this price range?

[Chrome Prince]

Only one of them was between $3.00 and $6.00 so good going, the rest were two odds on dogs and the rest were outsiders.

[rails run]

Thanks Chrome. Would have been another good day if BF was up. Under-thought but has caveman logic.

[woof43]

The backfitters out there should ask themselves the question, how can Joe Public get it right time in time out in there probs / actual results, yet when they run their back fitted systems, they fall into a hole at some later stage and they have wild fluctuations/variations.

My response was a method in how to have very little variation thru the data spaces, plain and simple.
Case closed.

Re Box 6

I can post some stats re the #6 box by justing adding one more variable you can identify plenty of dogs to lay.
As I keep a national database for greyhound racing i can provide an example.

One wouldn't use a blanket statement and say #6 is the worst box at Cranbourne 520m
These are the avg finish times and the avg split times fro the above distance.

1 2 3 4 5 6 7 8
31.168 31.075 31.115 31.134 31.107 30.938 31.025 31.109

1 2 3 4 5 6 7 8
5.624 5.620 5.626 5.641 5.604 5.588 5.607 5.625

One would ask why? This is where the physical artifact of the track takes over.

[UselessBettor]

Exactly..

All your first steps are, is to categorize the Fav and how strong or the gap over the 2nd fav. and to find correlations in the odds to finished odds nothing more or nothing less.

You will be testing as many variables (in formguides or tipsheets) as possible to find ones with a correlation and know which ones don't.

I didn't tell you, but once you find a variable or combination, you would then breakdown this down into subsets of races based on the fav odds <3.00 etc.

Of course there is holes in what I have been explaining, the above approach is to provide very good first order approximation to those precious crowd odds.

Just with the above information I'm able to provide trifecta combinations that provide positive expectations for each race catergory.

Once you have completed the above your then ready to move onto actually using your own handicapping in "cluster analysis".

Woof,

I think Im starting to understand. Let me paraphrase this to check I have this right.

The first thing I do is find variables which give me an understnading of the crowd odds. Once I find variables (such as barrier) which give an approximation of the crwod odds based on that variable I think start to categorise the actual results of any type of category. This could be last start winners, favs under $3 , top rated horse by tatts, top tipster selection, etc. It might even be more specific categories such as track/barrier combinations.

Once I have these two things I plot them on a scatter plot. By doing this with two different colour sets (say red and blue) I will visually be able to see clusters of where certain categories do deviate from the expected odd distribution.

Based on this clustering I should be able to find profitable situations that long term will be profitable ?

Is this right ?

If it is this still feels like backfitting but I guess your generalising the backfitted rules rather then being specific.

Of course I might have taken this the wrong way completely. Can you tell me if I am on the right path or if I have gone walking off on my own path again.

[woof43]

The variables you will be trying to find are ones that are "crowd behaviour" variables not handicapping variables such as Favs under $3.00 etc. don't think like a computer handicapper, this is where people go wrong they try and do analysis by what works with my handicapping, instead you need to be finding races statistically alike. You need to categorize by underlying statistically similarities You then have knowledge in how to apply effective handicapping based on these categorizes.

[woof43]

Woof,


Once I have these two things I plot them on a scatter plot. By doing this with two different colour sets (say red and blue) I will visually be able to see clusters of where certain categories do deviate from the expected odd distribution.

Based on this clustering I should be able to find profitable situations that long term will be profitable ?

Is this right ?


Yep that's right.
But
Your results from this first step, will show how people tend to make wagering mistakes within the clusters boundaries.

You need to take real advantage of your wagering knowledge, by applying effective handicapping techniques to the clusters and then learn how to handicap each cluster effectively.
You will then know why races are different and you will have designed workable methods from first principles and not by "try it and see".

[UselessBettor]

woof43,

I am going to go ahead and do this analysis. do you have somewhere I can contact you so that I can run things by you as I progress. I suspect this is going to take me a while but I think I am starting to understand.

[rails run]

Please do not provide links to commercial tipping sites. Thank you. Moderator.