- This topic has 23 replies, 10 voices, and was last updated 13 years, 4 months ago by
.
- Posts
Yesterday on ATR in between races James Willoughby and Hugh Taylor discussed many facets of betting. The discussions have now been placed on their site for all to watch free. Anyone interested in betting should watch this. Although it is mainly directed at racing punters most of what was said relates to general betting and even poker.
To watch go to <!– w –>http://www.attheraces.com<!– w –> and then click on ‘watch live racing online’ when you get there go to the features tab on the right and you will see the segments listed. There are seven in all. You will have to register but this only takes a minute.
The discussions are.
1. Form Factor special 1. Recency and availability bias.
2. Form Factor Special 2. confirmation bias and anchoring.
3. Staking plans how much to bet and when.
4. Importance of jockeys
5. Times in racing. Race times and their use in punting.
6. Fooled by randomness.
7. Betting markets. How markets work and how to make use of themThe best series of advice for punters I’ve seen.
Well done ATR, Hugh, James and Enzo.
Any punter who takes his/her betting seriously should see this.
Agreed with almost all of what was said.Thanks Ken.
Cognitive/confirmation Bias
https://www.youtube.com/watch?v=VQ4J0o8-wT8Times
https://www.youtube.com/watch?v=wX5OZbj8bPMJockeys
https://www.youtube.com/watch?v=d8Mh8THIxKYRandomness
https://www.youtube.com/watch?v=f0_mGyiV1ngStaking
https://www.youtube.com/watch?v=5cHIIZCPudcBetting Markets
https://www.youtube.com/watch?v=K71d2JdCzjcValue Is EverythingIt’s good stuff. The psychology stuff is interesting, I posted a thread on heuristics a few weeks ago which I’ll add to in order to explore this aspect further with a bit more detail.
Good stuff indeed, nothing revelatory to this tired old punter but a more clear and concise guide to the essential thought processes required of and so so necessary to the wannabe profitable punter is difficult to imagine
Thanks to Cormack and KenH for flagging it up
Well done Messrs. Wiiloughby and Taylor
Oh! James you and your love of the Kelly Criterion. Fine when dealing with known true-chance events such as coin tossing (see Yeats’ thread on 2/1 about known 1/1) but flawed when dealing with estimated edge on estimated true-chance, as in horse racing
You have a disciple in Gingertipster, an agnostic in me. Howzabout Level Stakes??
It is not the Kelly Criterion that is flawed, as it is 100% mathematically correct – it is the usual punter flaw in not allowing for the error in estimation. Once you do that KC is the bees knees and always will be.
Very entertaining, I enjoy listening to both but as usual James’s theories on betting and odds are flawed and full of holes as confirmed yesterday by the organiser of the "Improve Your Betting" conference which James attended earlier this year.
Surely the proof of the pudding is in the eating, just how much money does James make from putting his theories into practice?
It is not the Kelly Criterion that is flawed, as it is 100% mathematically correct – it is the usual punter flaw in not allowing for the error in estimation. Once you do that KC is the bees knees and always will be.
When I came up with my staking plan I did not know what Kelly was.
When I first started punting I did it with level stakes, but profit/loss depended solely on how the outsiders did. Obviously you’d need more than sixteen 2/1 winners to make more than you would on one 33/1 winner.
Then changed my approach, betting to win the same amount, 1 point on a 33/1 shot and 17 on @ 2/1. But then everything depended on the short priced horses, because I lost far more when they’d lose than I would on the outsiders.
It makes sense that stake should be judged on two things, how much chance the horse has of winning and how much value is in the bet. One horse that I believe has a 25% chance of winning should have double the stake of something I believe has a 12.5% chance (provided there’s the same amount of value in the bet). And something I believe is "unbelievable" value should have more on it than something I consider only "good" value. Also, with less money at risk on outsiders, it makes sense to win more money on them.
So, if I consider a horse a 50% chance of winning it will get 50 points staked if 6/5 or better is available (Evens = 50% + a margin for error). Something I consider a 25% chance gets 25 points if 7/2 or better is available (3/1 = 25% + margin for error), 12.5% 12.5 points @ 8/1 or better (7/1 = 12.5% + margin for error) etc etc…
But added to that: If something I believe has a 25% chance is available at 4/1 (fair 20%) then the differential between my assessment and price available is 5% (25 – 20 = 5)…
If I believe the the race is easy to work out the differential is multiplied by 5.Going back to the horse I believe has a 25% chance and available @ 4/1. It gets 5 (differential) x
If there are aspects that make the race a little tricky to figure out… I multiply the differential by only
Greater the chance of winning the greater the stake.
Greater the value the greater the stake.
But it still gives the advantage of the greater the price the greater the profit, despite less money risked on them:
As said: 25% chance available @ 4/1 gets 50 points @ 4/1 = 200 profit.
Something I consider a 50% (double the 25% chance) gets 50 points. If it is available @ 6/5 (45.5%) the differential is again around 5% (4.5%) 50 – 45.5 = 4.5 x 5 = 22.5 Then add the 50 making 72.5 rounded down 72 staked @ 6/5 = 86.4 profit compared to 200 profit for the 25% chance @ 4/1.
Something I consider a 12.5% chance available @ 12/1 (7.7%) has a differential again around 5% (12.5 – 7.7 = 4.8 ) 4.8% x 5 = 24 + 12.5 = 36.5 rounded down 36 staked @ 12/1 winning 432 points, compared to 200 profit of the 4/1 bet or only 86.4 profit @ 6/5. Although in practice the bigger prices are unlikely to have as big a differential.Value Is EverythingI would imagine Ginger you would find Kelly just as efficient to work out your stakes, the whole point it to maximise your edge and win the same amount of money depending on the edge. Like Robert99 says, it is mathematically correct, but most punters don’t even express their opinion of a race via probability so it would of course be very foreign to many.
If you have a 1.5 expectation then with a £1000 bank backing a 2/1 shot you would bet £250 to win £500. If you have a 1.5 expectation backing a 5/1 shot you would bet £100 to win £500, £50 to win £500 with a 10/1 shot with a 1.5 expectation etc.
I would always work on kelly to a divisor of the original run with a few stop losses and used progressively, betting to full kelly in the long term could be a bit of a roller coaster.
One thing I did disagree with James about was using seperate "banks". In my opinion it is best to climb in "steps", up the stairs instead of going up and down the mountain – staking a percentage of "bank".
In my betting two horses of equal chance and value racing only a month apart get the same stake – don’t see why it should have a stakes difference of hundreds or thousands of pounds… Depending solely on how well the punter has done in the meantime.
My bets come out of the same bank account as my bills. Of course if I get down to a minimum amount in there I’ll stop betting completely. The amount 1 point is worth in £s (stakes) remaining the same for a considerable amount of time before going up only when the money I have in the bank account has gone up quite a bit.
I am naturally cautious.
The trouble I have with betting a percentage of the bank is at some point every punter will go through a long losing run. Even the best punter will go through a bad run and this bad run is more likely than punters generally believe. Betting a percentage means a punter won’t lose everything, but one poor year can still wipe out years of steady hard earned growth.
I realise my way also means my profit does not go up in as steep a curve as by betting a percentage; but I’ll live with that.
Value Is EverythingIf you have a 1.5 expectation then with a £1000 bank backing a 2/1 shot you would bet £250 to win £500. If you have a 1.5 expectation backing a 5/1 shot you would bet £100 to win £500, £50 to win £500 with a 10/1 shot with a 1.5 expectation etc.
I would always work on kelly to a divisor of the original run with a few stop losses and used progressively, betting to full kelly in the long term could be a bit of a roller coaster.
I haven’t got to grips with everything about Kelly myself.
What do you mean by a "1.5 expectation" tb?Value Is EverythingThe term value comes from expected value or mathematical expectation, expectation etc.
If getting 5/1 about a 4/1 shot you can work out the expectation (or value as commonly referred) by odds on offer/true odds, using decimals, 6/5 = 1.2.
Thus meaning if accurate you can expect to make 0.20 units to a 1 unit stake long term, kelly will determine what proportion of your bank to make the right percentages basically.
They are not in a qualified position to be giving out psychological advice.
Learning & understanding a few theories doesn’t make you a psychologist.
I would imagine Ginger you would find Kelly just as efficient to work out your stakes, the whole point it to maximise your edge and win the same amount of money depending on the edge. Like Robert99 says, it is mathematically correct, but most punters don’t even express their opinion of a race via probability so it would of course be very foreign to many.
If you have a 1.5 expectation then with a £1000 bank backing a 2/1 shot you would bet £250 to win £500. If you have a 1.5 expectation backing a 5/1 shot you would bet £100 to win £500, £50 to win £500 with a 10/1 shot with a 1.5 expectation etc.
I would always work on kelly to a divisor of the original run with a few stop losses and used progressively, betting to full kelly in the long term could be a bit of a roller coaster.
A QUARTER of the betting bank on a 2/1 shot!
TENTH of the betting bank on a 5/1 shot!
TWENTIETH of the betting bank on a 10/1 shot!"Roller coaster" tb, it’d be like continuously ascending and descending Everest.
With bank/stakes going up and down. No wonder they talk about a "fraction" kelly.Value Is EverythingIt is not the Kelly Criterion that is flawed, as it is 100% mathematically correct – it is the usual punter flaw in not allowing for the error in estimation. Once you do that KC is the bees knees and always will be.
Indeed, Kelly is 100% mathematically correct because it’s an equation with probabalistic coefficients, which is all well and good when the true probabilities are known
I take your point about factoring in error when dealing with estimated probabilities, which is essentially why the wise punter prices up overbroke or, if pricing up round, is cautious and uses ‘half kelly’ or variants…if taken by Kelly
No problem with Kelly’s formula: an impressive and seductively, conveniently simple equation that could only ever have been devised by one in an ivory tower who spouted theory but didn’t part with folding
Big sense of deja vu Robert. Didn’t we and others do this to death what now seems a lifetime ago on Smartsig?
I asked then, and am still waiting for a reply: who has actually bet Kelly rigourously on horse racing and found it outperforms level stakes?
I bet level stakes and monitor my bets to Kelly, amongst other staking plans. Nothing has bettered levels, long term; which of course doesn’t necessarily mean levels is the bees-knees, just that I’m incapable of identifying true-edge in the somewhat chaotic event that is the horse race rather than the coin toss: no, just a woolly, uncertain, undefinable edge that can’t be represented by integers or real numbers that can be plugged into Kelly’s equation, but is nonetheless there…if you make a profit, long term
Terrific explanation by tbracing over on Yeats’ thread regarding the superiority – or safety – of Kelly over Levels when the true chance is known
Mostly good stuff that everyone should take in.
The Kelly Criterion sounds fine, but if I read it correctly it very much depends on how much of an edge you have.
In horse racing this is very subjective.
I work out my own speed handicap ratings that I think give me an edge- but does it ? – and if so how much of an edge ?
If I think a horse has a 50% chance of winning and the odds are 3-1 I will bet accordingly.
Similarly if I think a horse has a 20% chance of winning and the odds are 3-1 I won’t bet.
Who is to say my analysis is correct ?
If you analyse form in the traditional method of collaterel form do you have an edge ? especially as most other punters work on the same principles ?
I agree that your bets should reflect your thoughts on the chances of it winning but these are your thoughts and not mathematical certainties.
I personally believe in keeping things simple and bet no more than 5% of my bank on any one horse and I only have two stake levels, 5% to win and 2.5% for e/w bets.
My approach to staking is to have more on the more of a % edge I think I have. Of course that is only effective if your estimate of winning probability is better than the estimate of the market you’re betting into. I’m still working on that part.
Great point Molehorse. I have great respect for both James and Hugh but it did come across as a sort of dip into some convenient pop psychology theories.
The biases they highlighted are intuitively appealing as having the potential to impact a person’s betting and/or the accuracy of the market but whether they actually do so would require evidence/testing. I know Nottingham Uni do a lot of work on gambling psychology so there may be some studies but I’d like to see them before jumping to the conclusion that these biases do have an actual impact.
All that said, I think James tweeted to the effect that the limitations of programme-making had an impact on the content they would have liked to put over so I am sure he has a fuller, more rounded view than was glimpsed.
It’s a fascinating topic.
Another interesting topic is the subject regarding the impact self-image might have on a person’s potential to be a successful bettor.
With bank/stakes going up and down. No wonder they talk about a "fraction" kelly.
- You must be logged in to reply to this topic.