Staking plan and 10/10 confidence bets

Have a maximum and a minimum bet, perhaps 1 to 10.

i roll a dice

i roll a dice As good as any clever staking plan.

I've never understood this staking plan argument.  Why would you bet on anything other than a 9/10?

I would imagine that for most people, there would be a direct correlation between stake/ confidence and odds.  Surely if you're getting value then you should be placing the same stake on all bets. 

This argument rears it's head on here all the time and I'd love someone to explain to me why they a) bet on something that they acknowledge they don't think will happen,
and
b) why they then reduce the stake for that.

If anything it makes sense to vary the stake in relation to the perceived gap between fair and actual odds and increase the stake when you think the gap is largest to take advantage of it?

As always the correct answer is "use Kelly"

As always Ken will refuse to acknowledge that maths is useful in gambling.

i find level stake betting is sure way to lose long term ,
esp as you find you have higher % winners at short odds ,
more confident large stakes ..hope your judgement is right

I personally don't believe in staking plans, if Kelly was so good
why do people not use it in casinos ? Besides that, I offered a staking at 20:04. To say that I don't believe maths useful in betting shows how out of touch lori is.

People do use it in casinos and most card counters have since at least 1956 when Thorp published his work on blackjack.

In fact, it's MORE useful in casinos than in other betting if anything because you know your exact edge at a given time, so you can bet right on the kelly limit.

*Thorp's book was actually published in 1962 sorry. Kelly's paper was 1956 I believe.

Given your 20:04 post, when do you decide whether to bet 1 or 10 if it's not based at least in part on kelly?

a) Do you bet more when you're more sure you're right?
b) Do you bet more when the prices are shorter?
c) Do you bet more when you think the price is more wrong?

If all those are yes, then you're basically using Kelly, albeit a learner's version of it. If they're no, then why are they no?

Also, the phrase "I don't believe in staking plans" is meaningless.

If you level bet you have a staking plan, albeit a fairly boring one. If you don't bet at all, then again, you have a very boring staking plan (Nothing wrong with boring)

So I don't really know what you mean by that phrase.

All Kelly does is optimises your win rate (assuming you're a winning gambler of course) to win as much money as quickly as possible.
There are many assumptions, as we all know, but you can allow for those via various means (Which is why I usually use the phrase "Kelly based staking" rather than "Use Kelly" unlike this thread).

It should be clear to most people that a level stake is a decent staking plan, and also that it's not optimal. Betting £2 on a 100-1 shot and £2 on an evens shot should set alarm bells ringing in terms of risk.

Also betting £2 when your bankroll is £20 is clearly not the same thing as 1000 years later when you've built up and you're now betting £2 out of a £2000 bank.

Betting £2 when you're 100% sure that something's going to happen (If for instance you're a Pakistani cricketer) is also obviously not the same as betting £2 on an 11/10 shot that you know should be evens (a toss of a coin let's say)

Kelly simply allows for these things and within the assumptions (which you can find either by googling threads on here, or by looking up the wiki entry) will grow your bank at the fastest possible rate, it's not some crazy hocus pocus, it's mathematically proven.

"It should be clear to most people that a level stake is a decent staking plan, and also that it's not optimal. Betting £2 on a 100-1 shot and £2 on an evens shot should set alarm bells ringing in terms of risk"


why? if your betting level stakes and you see an even money shot that you think should be 4/5 or a 100 shot that you think is say an 80's shot, why should that set alarm bells ringing iyo?

Because your expected growth is massively different, for the same reason that if there's a lottery rollover that pays more than £14million (and technically giving you an edge), you'll still have a negative expected growth from betting it unless you're very rich - because you win it so infrequently.

lori, I bet only when the price is wrong, and my stake
is determined by the price, the shorter the price, the
bigger the stake. You can decide which of your three
questions that answers. Am I betting 'Kelly' ?

One question, does 'Kelly' win only if your method is a level
stake winner ?

Well of course it does, because it doesn't bet in situations where it doesn't have an edge, as well you know.

You are effectively betting a version of Kelly according to your answers above yes, although you're not adjusting for "how wrong" the price is, so you're using a primitive version of it.

A question for you:

Does level stakes betting win only if your method is a level stakes winner?

You can't magically make winners happen, you can only optimize what you win from your edge. Kelly does this mathematically as long as you play within the assumptions.

In the real world, you have to adjust for the assumptions of course depending on how accurately you can define your edge (In blackjack you can know exactly, hence the better usage in casinos), however there's no reason to not understand what kelly is and why it tells you the optimal stake for your bet.

If you're not a winning player, then the optimal stake is no bet, that's a very good starting point for many of course.

I bet to return 100 points depending on price of course, as in an evens chance
would be 50 (to return 100) a 3/1 chance 25 points and so on, backing or laying.
I don't work from a 'bank' my bank being my wealth, and increase the 'return' as
my Betfair balance increases. I don't try to calculate a degree of 'edge'for
staking purposes, but only bet when I have minimum 'edge', whether it be perhaps 20%
or 40%. I'm very comfortable with what I do. Is this a version of 'Kelly' ?

I've posted it before, but how Expected Growth works is explained very clearly here

http://forum.sbrforum.com/handicapper-think-tank/29009-expected-value-vs-expected-growth-kelly-criterion-part-i.html

Once that's understood then the fact that something like Kelly exists should be very easy to grasp. The maths behind Kelly a little tougher but I think that understanding the article is probably more important than understanding Kelly, and the article isn't particularly hard (although for some it'll probably take some work and concentration to get through as it does contain new concepts)

"My bank being my wealth" is the same for everyone surely?

Yes, you're basically using something related to fractional Kelly. The thing you would change if you were using Kelly is that you'd bet to win more points if the price were "very wrong" as opposed to just being "wrong".

In Kelly you would bet the second you have any edge at all, however in the real world you build in a buffer as it's difficult to know your exact edge in something like a football match. (Although I'll argue it's easier to get quite close than most people think)

The very simple version of how expected growth works can be seen as

a) If you bet when you're betting below the correct odds, it's clear you'll go broke in the end no matter how you stake

b) If you bet everything you own on every bet, it's clear you'll go broke in the end no matter how big your edge (unless the event is predetermined)

When you strip Kelly of it's mysticism, all it actually does is works out how cloes to b) you can go without overstaking.

"My bank being my wealth" is the same for everyone surely?

Not necessarily some people put x amount into betfair say £500 and hope to turn it into double or more. They consequently make too many bets, chase and eventually lose it all.

When you realise you should only place what you can afford (often more than the 'bank' you put into betfair to bet on one individual event but several times into betfair and make profit withdrawal, is when you start to bet properly and make money!

On kelly; it's brilliant if you have a calculable edge on thousands of past data to use in the future but far too risky and a waste of time otherwise.

Level stakes are ok if you bet on similar odds every time. I now largely prefer to bet level stakes because the odds are similar each time but I sometimes half my stake if I'm not 100% sure. Some would say surely this is a no bet then but I'm up with the 50% stake bets and see this as a bonus! If I lose then I only lost half of what I normally would :)

b) is a fairly obvious point. Overstaking is a killer for any gambler.

a) is also fairly obvious also. you want to be getting the best odds you can get. as an aside, i disagree with people who bet on outcomes that they don't think will happen just because they see extreme value in the price. surely if the bet is not going to come in, or has a very slim chance of coming in, its not advisable to take it. But thats another topic.

the most i will stake on a bet from here on out is going to be 10% of my bank and that will be for what I see as really, extremely strong bets. that will be for boxing bets that i am very confident on.

my base bet percentage will be 1-2% of bank depending on how strong I see it. that will be for horse racing lays. i lay favourites in the win market and i've seen that i can make it work short term at least. question marks about how it will go long term but im not using a massive bank to try it out. I think this is the right way for me to do it.

Amazing how often people's instincts converge to the maths.

Pulling numbers straight out of my wazzoo, I'd say that you're probably betting far closer to Kelly than you realise (As has been the case when I've looked at "instinct" gamblers - of which I'm partially one myself)

1-2% normal and 10% tops fits nicely into that wheelhouse imo.

In football I find some times 'a little bit wrong', a little bit more often 'wrong'
but almost never 'a lot wrong'.In fact with regard to the last 'a lot wrong' I go
back and have another look !!
The 20%+ I seek satisfies me, as does my staking. I don't believe these things need to
be slide rule accuracy as in 20%, 22% 24% etc. I'm more than happy with what I do.

I agree about not needing to be slide rule accuracy fwiw. While I can work out the Kelly for everything I want to do, and can probably quote a few as well, it's more about the "knowing what it should be" than doing it exactly (for me at least) as a lot of it is only there for a moment anyway.

Factor in further elements like trading and not wanting to overbet at bookies etc etc and it's only a rough guide

I'd certainly not disagree with your staking system like I do with most that are posted on here - however if you ever offer me 2.1 about a coin toss where I get to provide the coin and you call (A true 50-50 in other words), I'll be betting the max allowed by Kelly to the penny.

It is important to understand that no one game will decide the overall outcome of a long term betting strategy.
Thus essentially your staking plan must conform to the basic need to keep your losses containable till your wins come in to put you ahead hopefully.
Therefore you must be able to work out ( from past results)  approximately what is the " worst case" percentage drawdown in your capital you can expect to experience in your inevitable losing streaks.
You then calculate your stake working back from that percentage and the current amount of risk capital you are playing with.
AS they say on Wall St, the most important thing in investing is not to run at the first sign of any losses.
The same in gambling.
If you don't stick around and weather out your losing runs, you will not profit in the long run.

FAFH - do you have an example or can you explain in more detail how to calculate the stake?  I'm ahead at the moment but I think I may be overstaking.  I have about 15 months of daily P/L figures with the "expected" ups and downs, and my bank has survived so far.

Any advice greatly appreciated.  Thanks in advance.

Ok....

How do you know you have value? Are you betting out of line odds on other books/comparing price to bf? I assume BF to be pretty spot on 99.9% of the time on the true probability of an event happening.
If you are betting out of line odds at other books, why have a staking plan? Why not bet as much as your bank/limit allows lay off, or preferably dutch against another book? I'm no mathematician, but I'd say that the constant guarantee of a profit would out weigh a potentially inefficient staking plan, despite in the long term you should have value - esp if you have a small bank and playing against variance?

I'm probably missing the point completely as I'm no gambler, but thats how I see it.

How do you know you have value? Are you betting out of line odds on other books/comparing price to bf? I assume BF to be pretty spot on 99.9% of the time on the true probability of an event happening.

Disagree. An experienced gambler knows when the true probability of an event happening is higher than the odds available so he can bet on the outcome knowing that he is getting better odds. This happens all the time. I would not say betfair is spot on 99.9% of the time with the probability.

I understand your second point but it doesn't apply to me. Its a decent point though.

But betfair has the advantage of not one persons opinion, but hundreds, if not thousands on one event, so I'd guess that it is probably the most realistic model (but I'm no experienced gambler), so if a bookie has a price above that its a value bet, below, its a -ev bet. When I see people on this forum where £500k has put chelsea at 1.30 to win (and lets imagine that becomes SP), and the individual reckons (with confidence) that 'in his book' chelsea should be 1.25, and therefore getting a value bet at 1.30, is incorrect imo. If sid james offered chelsea at 1.35, that would be the only value bet.
Unless that person happens to be a pakistani bowler I suppose...

Unless the gambler is creating a balanced book, or only placing bets where a bookmaker is higher than price at exchanges (where overround as near to 100% as poss), I can't see long term how they can profit only betting one side on here. But again, thats me assuming BF is near as poss a true reflection of the odds.

Wayne
A simple example.
If your expected max downturn is ( say) 25 points and you wish only to lose a max 25% of your capital on any major losing sequence, then you should only bet 1 point per 100 points of risk capital you have allocated to your system.
The amount of risk capital you have allocated or accumulated should be that figure when your system last reached its highest profit level.

Jimbo
I think the mistake people make in staking according to the " value " of any one particular bet is that any bet " value ", good or bad,  is only ever an expectation of an outcome.
Anything can and will happen on any one event.
Over a longer period of time the performance of " good value " bets should outweigh the performance of " bad value " bets, that's for sure, but in the short term, that's not for sure.
So I maintain that your stake should be almost totally related to trying to maintain your capital base reasonably intact in bad times, which will inevitably occur in the best of betting systems.
As far as comparing BF odds to bookie odds to evaluate whether you have value or not, well that assumes that the bookie price is somehow related to value in an absolute sense.
I'm not sure about that assumption. The prices the bookies show are those that will effectively give them a balanced book, which is not the same thing.

Understanding Kelly Criterion

A hypothetical sequence of 100 bets at even money

For this example we win 54% of our bets

Kelly says we win 54 bets and lose 46 bets the difference is 8% So we bet 8% of our bank

If we had a bank of 1000.00 we should bet 80.00 first bet, then 8% of re-calculated next bank amount.

It is important to note that Kelly formula does not take into account Commission.

Real life scenario. We have a 1000.00 bank and a commission rate of 5%. 54% winners. Now some ifs ~
If we placed 1.00 on all our 100 selection we win 8.00 then deduct comm. = 4.30 profit (our effective edge = 4.3%)
To utilize our bank of 1000 We now look at what we would make in profit at different percentages of our bank
taking into account 5% commission.

% of bank	Profit
0.10%	4.3
0.20%	9.3
0.50%	24.99
1.00%	49.35
2.00%	90.78
3.00%	122.85
4.00%	145.75
5.00%	157.61
6.00%	157.69
7.00%	148.44
8.00%	127.72
9.00%	96.15
10.00%	55.29
11.00%	6.21
12.00%	-49.09
13.00%	-111.07
14.00%	-176.79
15.00%	-246.16
20.00%	-582.02

From this we see that Kelly at 8% would not be the most profitable purely because of commission.
The ideal % of the bank to bet would be in the range 5 to 6%. If you bet more than 11% of your bank
you would actually lose.
As a general rule the best % of your bank to use is your edge %, that gives you a little lee way.
In the above example I would bet 43.00 From a float of 1000 on an even money chance that I calculated had a 54% chance of winning.

Calculation error should have been a 5.3% edge at level stakes
The 4.3% was adjusting the bank after every bet.

So my last line should be a 53.00 Bet on a float of 1000.00

I assume BF to be pretty spot on 99.9% of the time on the true probability of an event happening.

I disagree.

kenilworth
If you take every even money shot BF's probability is 99.9% right but we don't
have to take every even money shot, we can pick and choose. With those choices
it is quite easy to get a 54% strike rate on the even money prices we choose.
Cheers

On further study.
you may say what if I can get 2.2 price instead of 2.0 (even money)
It seems that whatever Kelly suggests as a percent of your bank to bet
at the odds you can get, the optimal for 5% commission is 2.7% less than
Kelly, Kelly says 15.67% of bank at decimal price 2.20 if 54% strike rate. If paying 5% commission you would
make that 15.67-2.7 = 12.97%

Kelly agrees with your first example as far as I can see.

54% at 2.00 with 5% commission is 54% at 1.95, which comes out at 5.58% of bank to bet.

*That is to say you have to calculate it based on the odds you're actually getting, not the odds on the screen.

Betfair try to claim it's the losers that pay commission yet the price shown on the screen factors in the winners paying it. You have to allow for this before betting of course :)

So let's assume that you might, just might, lose 10 evenchance bets in a row,and then win 11 in a row. So you win 11 out of 21, being 52.38 %.
So, using your example , you bet ( say) 5 % on your declining bank balance
Bet  Amount Bank
            100.00
1    5.00    95.00
2    4.75    90.25
3    4.51    85.74
4    4.29    81.45
5    4.07    77.38
6    3.87    73.51
7    3.68    69.83
8    3.49    66.34
9    3.32    63.02
10   3.15    59.87
11   2.99    62.71 * 5 % comm. now payable
12   3.14    65.69
13   3.28    68.81
14   3.44    72.08
15   3.60    75.50
16   3.77    79.08
17   3.95    83.36
18   4.17    87.32
19   4.37    91.47
20   4.57    95.81
21   4.79    100.36

If on the other hand you just flat bet 5 units, you would end up with 102.25, that is lose 50 ( 10 * 5)and win 52.25 ( 11* 4.95).

That is, Kelly makes you worse off, not better off.

As explained by Lori with those figures you have given ~
52.38% and your odds even money less commission would be decimal odd of 1.95
Put that into Kelly and this is what you get.

Bank	Bets	P/Loss
100.00	2.20	-2.2
97.80	2.15	-2.15
95.65	2.10	-2.1
93.55	2.05	-2.05
91.50	2.01	-2.01
89.49	1.96	-1.96
87.53	1.92	-1.92
85.61	1.88	-1.88
83.73	1.84	-1.84
81.89	1.80	-1.8
80.09	1.76	1.67
81.76	1.79	1.7
83.46	1.83	1.73
85.19	1.87	1.77
86.96	1.91	1.81
88.77	1.95	1.85
90.62	1.99	1.89
92.51	2.03	1.92
94.43	2.07	1.96
96.39	2.12	2.01
98.40	2.16	2.05
100.45

2 things to compare here is that 1. your gaining in this case .02% per bet.
So because you are compounding up You will eventually make a lot more than level stakes.

2. Is after 10 bets at level stakes of 2.20 you would be down 22.00 where Kelly is only losing 19.91

I know who would last the longest losing sequence and if the win rate held true which one would make the
most profit in the long run.

Of course all your calculations assume that you know your edge before you bet.
You might think you know it but you don't really.
You will only actually know if you have an edge at the end after a large number of bets.
Everything else is conjecture and what you might call a " perceived" edge.
As anybody on here will know ad nauseam, I believe that the concept of " value" betting is highly overblown, because one person's good value calculation is another person's ( equally qualified to judge) bad value.
It's all in the eye of the beholder.
Bookies don't make absolute value judgements, they just set prices to attract customers on the side of the book they need to balance their positions.
Sure the original opening prices are relevant to historical stats, but they don't necessarily mean anything in terms of what the bookies think or don't think is value..

You might think you know it but you don't really.

You probably know it to within 0.5% though, which is fine. It's really not that hard to have a very good idea of your edge (As I understand it , it's harder in horse racing than normal sports though)

Jimbo747 Joined: 15 Feb 08
Replies: 1850 30 Sep 10 20:34 
Ok....

How do you know you have value? Are you betting out of line odds on other books/comparing price to bf? I assume BF to be pretty spot on 99.9% of the time on the true probability of an event happening.


For me this is where so many people go wrong.  They assume that the betfair market ( or any other market is correct ).  Why?  Because its obviously popular opinion?  Any bet that is struck pre match should be evaluated before you look at the market.  I go through the weekend fixtures and price them up myself and if they are spot on with the prices then no bet - if they are out of line then you take advantage.

For me when you look at the market first, it then sort of distorts your thinking, rather than doing your thinking first and then looking at the market

Lori
I do admire your confidence that your value assessments are correct to such a degree in a field of betting where there is no mathematical certainty of anything ever happening on average, unlike in a casino.
More particularly if you are then prepared to back up such value assessments with such a theoretical approach to staking, such as Kelly.
You're a braver ( though some might say a stupider) person than me.
But the bottom line is whether you win long term.
You claim you do, and I have no reason to disbelieve you, and so that's all that matters essentially.
However, I very much doubt that many others who try to follow your strict value/ Kelly approach will fare as well.
Again I might be wrong.
After all it's all just personal opinions and beliefs we are expressing on here.

Just to be entirely clear, while I preach that everyone learn how Kelly works, there's no massive need to follow it strictly. (Although again, if you get odds on a coin you know is fair, you really should get very close)

It serves the huge purpose of setting you a maximum (Which can be recommended to the experienced and new alike) and also give a feel for how to alter bet sizes based on the various factors.

In the real world there are several factors that keep you away from the maximum... in the order they swirl into my mind..

1) Personal utility - There comes a point where you're happy with a certain size win and uncomfortable with a certain size loss.
In theory I can lose 98% of my bank and rebound, but in practise I have to pay rent at some point and feed my insatiable urge for curries at least once a week.

2) Time. You don't want to be hammering away working out your bet size to three decimal places when in a hurry.

3) Disguise.  While thinking you're an idiot/crazy, a bookie will at least take passing interest in someone who puts in strange size bets. I don't like people taking an interest in me (Except on forums, where obviously I'm a bit of an attention wh**e or I wouldn't be posting)

4) Accuarcy. The one people always dig up. You can't be sure of your exact level of winning, so you have to allow for this.

Again though I suggest everyone at least have an idea what's going on, treat the number as a maximum, and build something they're comfortable with within the Kelly framework. It should certainly allow most experienced gamblers to be a little more aggressive on occasion (I say that because most beginners will be shocked how low the maximum is most times)

Sound like very good, commonsense rules to follow.
Just drop out the Kelly and all should be fine.
Just joking.

For me this is where so many people go wrong.  They assume that the betfair market ( or any other market is correct ).  Why?  Because its obviously popular opinion?  Any bet that is struck pre match should be evaluated before you look at the market.  I go through the weekend fixtures and price them up myself and if they are spot on with the prices then no bet - if they are out of line then you take advantage.

For me when you look at the market first, it then sort of distorts your thinking, rather than doing your thinking first and then looking at the market


But what makes your opinion more 'correct' than that of the market, and opinions of 100's+ other players? Thats the part I don't yet grasp. As in my earlier example, if chelsea are SP 1.30 to win on here, and 1.35 at ladbrokes, imo the 1.35 represents value. If you put Man Utd at your own odds of winning as 1.5, yet is 1.6 to back at BF, in your opinion this makes it value, whereas to me that would be true probability in majority of cases and would not be value long term. How is your figure more accurate than one formed with thousands of opinions? I just want to understand why popular opinion is wrong more than 0.1% of the time?

How is your figure more accurate than one formed with thousands of opinions?

Hard work for the most part.

For instance in the days of that daft system (forgive me if I get this backwards, I wasn't doing much soccer at that time) where people were told to lay the draw and cover it with 0-0, laying 0-0 was a profitable proposition (I seem to remember buzzer was the one who posted about this (after the event of course, he's not daft))

The sheer weight of complete idiots was too much for the market to handle.

There are less obvious examples of this all the time, international events can pick up heavy bias for instance. I remember when Bruno fought Tyson, you could bet 1/3 Tyson in England and 9/1 Bruno in the USA. The internet wasn't really around much then to level off the arb like it would today, but that doesn't mean that people have stopped betting with their heart rather than their brain cells, it just means the differences are more subtle and around for less time.

Popular opinion would be wrong ALL the time if it was one man one vote, but unfortunately the clever guys get more than one vote. (IE they have more money in general, especially when betfair take the bets off the super-rich fish) which means the market gets pretty close to efficient.

I suspect you're underestimating the sheer time and effort (often by teams of people) that goes into unearthing an edge. It's a rare gambler in this day and age who can do it simply by feel (but they do exist) alone. A combination of feel, maths and never ending thirst for information (team news, weather news, likely strategy, whatever) goes into keeping things as accurate as possible.

Some people for example get up at 3:33am to see what the weather is going to be at a godforsaken golf course in Wales in 4 hours time, simply because they might get an extra few quid on the event being delayed or going ahead..............

feed my insatiable urge for curries at least once a week.

Me too. I just gamble so I can afford more curry.