- This topic has 132 replies, 20 voices, and was last updated 16 years ago by
.
-
Posts
-
Any and every edge is ephemeral. Here today-gone tomorrow. Rely instead on confidence in your own ability. It may be that correctly interpretating form is your forte. Perhaps it’s the use of the correct stats and trends. Maybe you have developed a successful system. You may be good at sifting the wheat from the chaff in the many postings you see. Whatever it is, if it’s working for you and are making a profit, that is your so called edge.
Avoid under 2/1 shots though.
I have just finished reading this book Fooled_by_Randomness and I must say, it was food for thought .. has anyone else read this?
Dave Jay, as a matter of interest, does it change your view on any of what was "discussed" on the infamous VDW thread?
Having embroiled myself in that particular waste of time, I was too busy trying to disembroil myself from it to be sure whether you were agreeing with my remarks about randomness, Monte Carlo simulations and Black Swan events, or not.
I did not feel that there was anything particularly remarkable or profound about Taleb’s book, but it was heartening to read someone giving it with both barrels on the subject.
This is my first, and possibly my last, posting on the "Systems" Forum.
Second-last post, that should have been.
How many times will a coin have to be tossed for the split to be 50:50, or even 55:45?
The more tosses you make, the nearer to 50:50 you are likely to get in % terms but the further from a neat split in absolute terms.
100 tosses might be 48:52
1000 tosses might be 490:510
10000 tosses might be 4965:5035The absolute difference is respectively 4; 20 and 70, but the % difference is getting smaller.
Yet if you were to get 11/10 about the coin toss, you would consider that an edge as such, but even those odds aren’t enough to compensate for the randomness factor.
Randomness can work for you as well as against you.
Read all about it!
First Sea Lord Prufrock about face, 7 minutes after first and last visit to the Systems ForumHaving embroiled myself in that particular waste of time, I was too busy trying to disembroil myself from it to be sure whether you were agreeing with my remarks about randomness, Monte Carlo simulations and Black Swan events, or not.
Prufrock, I wrote my own montecarlo simulation about 3 years ago and I’ve been fiddling around with it ever since. I do agree with whole-heartedly with your comments about randomness, but I am more interested in what effect it really plays, in real time. I coming around to think that you can’t really know that you have an edge with absolute mathematical certainty and that luck plays a far bigger part in things than we would care to imagine.
the problem lies in our lack of ability to estimate our error rate, except in a circular manner. To compute an error rate one needs a probability distribution; to be confident about the probability distribution one needs an error rate.
Regarding VDW .. I think you should bear in mind that everyone needs to start somewhere and assigning points for various factors effecting a horses performance and then creating an odds-line isn’t a bad way to start. A lot of punters regard themselves as recreational gamblers but enjoy the challenge, I put myself in this group. Everyone needs to start somewhere, a systematic approach to solving the puzzle has got to be better than a chaotic one? So in answer to your question, No the Taleb book hasn’t changed my views .. I think we were both arguing from different view points. I wasn’t putting forward VDW as an ultimate system but more like a good starting method.
‘Fooled by Randomness’ sounds interesting; it’s gone on my Christmas list.
I think the difficulty deciding whether one has an actual edge and just what statistical tests and sample sizes are needed to prove unequivocally that you’re not just on a ‘lucky run’ thanks to the vagaries of chance and random sequence, is what keeps the demons of the losing run and bank wipe-out tapping on the shoulder of all punters regardless of how ‘long term’ their profit is.
If one is indeed in possession of an actual edge then IMO it cannot be quantified in terms of the analysis/methods/filters/betting strategy you employ to solve any individual race or group of races (each is a unique event and challenge), but rather can only be qualified by an indeterminate statement such as ‘I have an edge because I show a long term profit, but what that edge is – other than in bottom-line profit-on-turnover terms – I don’t know’
I quite agree with Robert’s questioning of the validity of edge-to-odds Kelly-type staking as this, by definition, depends on you knowing with certainty what your edge is on an individual bet. It’s been said many times: Kelly and allies work brilliantly when back-fitted.
Furthermore as one can’t know precisely what edge you have on a single bet other than ‘x is a value bet, ‘y is a value lay’ then one should stick to Level Stakes.
I wasn’t putting forward VDW as an ultimate system but more like a good starting method.
If you deleted the word "good", I could buy that for a dollar.
luck plays a far bigger part in things than we would care to imagine.
I actually think one of the biggest edges I (imagine I) have is not just that I long since realised that luck plays a bigger role than I had once thought but that I now embrace that fact to my advantage.
Uncertainty does not have to be a negative experience.
I quite agree with Robert’s questioning of the validity of edge-to-odds Kelly-type staking as this, by definition, depends on you knowing with certainty what your edge is on an individual bet. It’s been said many times: Kelly and allies work brilliantly when back-fitted.
Furthermore as one can’t know precisely what edge you have on a single bet other than ‘x is a value bet, ‘y is a value lay’ then one should stick to Level Stakes.
I thought they were good points as well.
Nonetheless, I think that not knowing the precise edge you have does not invalidate the principle behind Kelly, namely that you should stake proportionately more or less of your bank according to the trade off between the likelihood of something happening and the edge you have.
Surely this still holds true even if your assessment of the likelihood of something happening and of the edge you have is subjective and unprovable. What you are testing, as with most betting, is the accuracy of your judgement.
If your judgement is good then you should be staking more when Kelly tells you to stake more. And if it isn’t then you arguably shouldn’t be staking at all.
I actually think one of the biggest edges I (imagine I) have is not just that I long since realised that luck plays a bigger role than I had once thought but that I now embrace that fact to my advantage
Prufrock .. if you were to put a figure on luck what would it be? I think pure luck decides around 20% of race outcomes, at least.
Dave,
When it comes to the supposing of true prices it can be seen that betting a true prices, each and every bet, can produce an overall profit or loss by chance (randomness or luck). The greater the number of races, the higher the profit or loss, but there is an almost zero chance of the imagined zero profit. None of that profit or loss has been derived by edge or ability, or the lack of both. Some (but not so many) will even make a profit taking 3/1 prices for 4/1 horses, purely by chance.
Being in profit is not necessarily a measure of edge nor even competency.
Doing consistently better than that which chance allows is the only true measure. If you don’t measure that, then you are not certain of edge / ability over the long term. Humans are wired short term so when the losing run or Black Swan occurs they may make irrational decisions to change things. They also make irrational decisions to change when things are going "too" well. For them it is a matter of the short term followed by a disaster, then a new direction, followed by disaster (much like politics). So the thread you have introduced could be very helpful if it sticks more to quantitative ways of determining the amount of randomness and the increased odds you need to overcome that.As far as luck deciding race outcomes you could get a better idea by the number of favourites that lose ie some 67%. The number of first and second favourites that both lose ie some 55% and so on. So the best the market can do is to reduce your average chance of a loss from a sticking in a pin method (90%) to something like a half of that. The market being the best sum total we are told of all the available relevant information before each race. So your 20% figure on race outcomes is probably far too low.
Robert .. that’s the general thrust of my argument and the dangers of a 100% book versus a 120% book. You’re far more likely (20% more) to be randomly lucky, so measuring your edge is harder. I tend to think of runs of luck lasting around a year in racing, I don’t know if anyone else would agree with that?
First of all for anyone who doesn’t know what a Black Swan event is, this is the definition,
A Black Swan Event should satisfy three criteria;
1 – it carries upon its occurrence a disproportionately large impact. The impact being extremely large, no matter how low the associated probability, the expected effect (the impact times its probability), if quantified, would be significant.
2 – its incidence has a small but incomputable probability based on information
available prior to its incidence
3 – a vicious property of a Black Swan is its surprise effect: at a given time of observation
there is no convincing element pointing to an increased likelihood of the event.Examples of a Black Swan event would be the effect of the WTC ( 9/11 ) attacks, Tony McCoy falling off a horse, remounting and going on to win the race. Anything that unexpectedly happens and then when it does you are surprised why you are surprised.
To measure luck, you can’t use actual versus expected techniques as in favourites and second favourites, in my opinion, because you don’t expect longer price horses to win and when they do their chances are reflected by their odds.
I have been looking at the problem in this way, by grouping analysis into outsider / not-outsider.
If you took a sample of 1200 flat handicap races with 8 to 11 runners and said everything that won at a price bigger than 10/1 was down to luck. Around 14.5% .. so, 14.5% of all winners come from the outsider range. I may be barking up the wrong tree, or your idea might be better Robert, I’m not sure .. starting off with the horses that don’t win might be the way to go.
To measure luck, you can’t use actual versus expected techniques as in favourites and second favourites, in my opinion, because you don’t expect longer price horses to win and when they do their chances are reflected by their odds.
I have been looking at the problem in this way, by grouping analysis into outsider / not-outsider.
If you took a sample of 1200 flat handicap races with 8 to 11 runners and said everything that won at a price bigger than 10/1 was down to luck. Around 14.5% .. so, 14.5% of all winners come from the outsider range. I may be barking up the wrong tree, or your idea might be better Robert, I’m not sure .. starting off with the horses that don’t win might be the way to go.
Dave,
It is not then clear what you mean by luck if it is not the unexpected or unlikely happening versus expected. I would define an outsider as a horse ranked below horses whose prices totalled 80%+. My suggestion was in answer to to your question and for an outsider to win, it means the first favourite must fail AND the second favourite AND so on. So the outsider is "lucky" to win, as the win has only come about by the unlikely serial failure of all others more favoured. "Luck", or the unexpected, starts much earlier in the rankings in my view and increases as you go down the list. That happens 55% of the time (1st and 2nd fav fail) so it is more expected than unexpected. Outsiders will have longer odds by definition, but in a 10 runner race there are more of them than "favourites". An outsider of the front two is actually more likely, but which one is the problem.
I know what you mean and this isn’t easy, I suppose because there are so many ways to look at things. It’s finding something that’s workable and correct.
I am thinking this could be a co-efficient rather than a static factor based on the likelihood of the favourite failing, so it would be different for every race based in the front ranking odds. I suppose you could do something like compare the probability density function of the front two, with the density function of the rest of the field?
I don’t think there is any definitive answer to this problem because of the nature of subjective probabilities.
When we bet on a horse at evens, we have no idea what the true odds are because they cannot be known. All we have is the historical data that every horse in history that was sent off at evens won X% of races.
We might deduce that this is typical of all races and that we can estimate probabilities from historical data. This is probably the most practical approach, although it might not satisfy the purists or statisticians.
In the long run, based on past results, we can expect horses to win in proportion to their starting prices, although this is not a linear relationship. In the short run, anything can happen.
I know this is back to where we started, but I honestly believe there is not much to gain by taking the statistical analysis deeper than this.
I think you might be right Artemis. You know you are going to be lucky and unlucky, in equal measure, but can’t know when you will be.
It won’t stop me writing a model for it though ..
Some people don’t half talk some DRIVEL on here.
"When we bet on a horse at evens, we have no idea what the true odds are because they cannot be known. All we have is the historical data that every horse in history that was sent off at evens won X% of races."
No wonder that with the information that you suppose "is all thats available" you end up debating the relevence and importance of lucK And how to calculate and make an allowance for it.
My GOD
- You must be logged in to reply to this topic.