Staking while betting on value - Help

[danshan11]

I don't even know what the F**k you are talking about and I don't think you do either.

what part is confusing you? I can try and clarify. I said before and will say again Kelly is not for sports betting, we dont know our edge until the game starts and that makes kelly shaky. and if you use our edge as an average over the past 100000000 bets, that is flat betting. I am not sure what part you do not get, ask me and I will try and clarify it.

[danshan11]

to use kelly you need to use an edge
what edge do you use
if you model -120 and the current line is +100 is that your edge? if so what if it closes at -130 or +120?
if you take your last 100000000 bets and average your line value over those bets and get model avg -120 and close avg +100 is that your edge on this particular game and what if this particular game closes at +140 or -145?
the closest method we have to true implied is the closing line and we do not have that information before a game closes and that is too late.

now if you want to backtest Kelly you need to have 3 things 1, your #, bet at # and closing # and if you dont have those how can you backtest kelly over a significant sample size numerous times to come up with some sort of consensus?

[Okocha]

Jesus, from Ganchrow to this clown,please dont contaminate HTT with blunt ignorace,it is disrespectful and there are good forums for you out there where you can share your revolutionary ideas in this industry for a audience more suitable to your Bettiing IQ and bankroll

[danshan11]

Jesus, from Ganchrow to this clown,please dont contaminate HTT with blunt ignorace,it is disrespectful and there are good forums for you out there where you can share your revolutionary ideas in this industry for a audience more suitable to your Bettiing IQ and bankroll

typical answer, but you dont answer the questions do you, why?

[zorba74]

what part is confusing you? I can try and clarify. I said before and will say again Kelly is not for sports betting, we dont know our edge until the game starts and that makes kelly shaky. and if you use our edge as an average over the past 100000000 bets, that is flat betting. I am not sure what part you do not get, ask me and I will try and clarify it.

I was being a dick and I apologize. I just can't believe I'm in a Think Tank where a guy is actually saying the things you are saying (e.g. Kelly is useless, etc). Not sure why the guys here haven't put you in check already but not my problem. I tried to be helpful and I didn't care for the tone of your response, hence my dick response. When people help me, I'm grateful, I don't respond w/ silly BS.

[danshan11]

I was being a dick and I apologize. I just can't believe I'm in a Think Tank where a guy is actually saying the things you are saying (e.g. Kelly is useless, etc). Not sure why the guys here haven't put you in check already but not my problem. I tried to be helpful and I didn't care for the tone of your response, hence my dick response. When people help me, I'm grateful, I don't respond w/ silly BS.

no offense taken and no need to apologize
can you answer the questions I asked that would be really helpful for me for sure and maybe just maybe a few other members as well

[zorba74]

no offense taken and no need to apologize
can you answer the questions I asked that would be really helpful for me for sure and maybe just maybe a few other members as well

What question?

[danshan11]

the questions here

to use kelly you need to use an edge
what edge do you use?

#1if you model -120 and the current line is +100 is that your edge? if so what if it closes at -130 or +120?

#2if you take your last 100000000 bets and average your line value over those bets and get model avg -120 and close avg +100 is that your edge on this particular game and what if this particular game closes at +140 or -145?

#3now if you want to backtest Kelly you need to have 3 things 1, your #, bet at # and closing # and if you dont have those how can you backtest kelly over a significant sample size numerous times to come up with some sort of consensus?

[HeeeHAWWWW]

Closing price is completely irrelevant.

[Alfa1234]

we dont know our edge until the game starts and that makes kelly shaky.

100% agree, you can quantify your model's edge over a decent sample but will never know the edge on any particular game...which in itself makes Kelly almost unusable.

[zorba74]

the questions here

to use kelly you need to use an edge
what edge do you use?

#1if you model -120 and the current line is +100 is that your edge? if so what if it closes at -130 or +120?

#2if you take your last 100000000 bets and average your line value over those bets and get model avg -120 and close avg +100 is that your edge on this particular game and what if this particular game closes at +140 or -145?

#3now if you want to backtest Kelly you need to have 3 things 1, your #, bet at # and closing # and if you dont have those how can you backtest kelly over a significant sample size numerous times to come up with some sort of consensus?

Your historical Yield is basically your edge; the rest is just commentary.

The Sports Market is a Complex Adaptive System so you need to stay in perpetual beta in order to extend your edge into perpetuity but there are always edges to be had.

[danshan11]

Your historical Yield is basically your edge; the rest is just commentary.

The Sports Market is a Complex Adaptive System so you need to stay in perpetual beta in order to extend your edge into perpetuity but there are always edges to be had.

historical yield as your edge is AKA flat betting

[zorba74]

historical yield as your edge is AKA flat betting

Incorrect. If my historical Yield is 2%, that doesn't mean I have a 2% edge on every bet; it means I can ID value.

If you want to flat bet or "bet to win a fixed amount" etc, do as you see fit but stop disseminating these silly "Kelly is useless" statements. It may be useless for you; not others.

[danshan11]

how do you ID value bets? you mean if its baseball and on Tuesdays those are worth a 2% yield, I dont get it comparables what, please explain, I would love to get a good understanding

[zorba74]

how do you ID value bets? you mean if its baseball and on Tuesdays those are worth a 2% yield, I dont get it comparables what, please explain, I would love to get a good understanding

I think I've been very patient w/ your trolling, silly statements and questions. Your clown act doesn't belong in a Think Tank and I've humored you enough.

You can't ID value... I could have told you that after reading just one of your misguided posts.

[danshan11]

I think I've been very patient w/ your trolling, silly statements and questions. Your clown act doesn't belong in a Think Tank and I've humored you enough.

You can't ID value... I could have told you that after reading just one of your misguided posts.

you got a big mouth! you talk a lot of smoke but dont say anything, how do you ID value it is a simple question. how do you ID value?

[danshan11]

I think when you say historical yield, you are referring to your average return on sports betting over a past sample size, is that correct or wrong?

Cause I think that plays right into my argument against kelly, each bet is individual and even if you avg value is 2 or 3% that has absolutely nothing to do with this specific bet. you seem to have enough of an idea that you understand the efficiency of the closing line and that is the closest to the true probability of that particular game.
So if your yield historically is 2% and you base kelly on this particular game on 2% and the line closes 4 or 5% different doesnt that mess up the actual amount you should have bet on this particular game? and doesnt that essentially mean you are betting with a 2% edge on all bets you make using kelly and isnt that flat betting?

GAME A you think -120 it closes +100 using kelly you should have bet $0 on this right? but you still bet using your Historical Yield and assuming 2% edge
GAME B you think -120 and it closes -150 using kelly you should have bet a big number on this game right?
but you only bet using your Historical Yield and assuming 2% edge not the actual bigger amount you should have bet on this game right?
Game C you think -120 closes -120 you should have bet $0 on this game as well right? but you still bet using your Historical Yield and assuming 2% edge

[Alfa1234]

Cause I think that plays right into my argument against kelly, each bet is individual and even if you avg value is 2 or 3% that has absolutely nothing to do with this specific bet. you seem to have enough of an idea that you understand the efficiency of the closing line and that is the closest to the true probability of that particular game.
So if your yield historically is 2% and you base kelly on this particular game on 2% and the line closes 4 or 5% different doesnt that mess up the actual amount you should have bet on this particular game? and doesnt that essentially mean you are betting with a 2% edge on all bets you make using kelly and isnt that flat betting?

Agree with the above again, although the closing line part is incorrect as that also plays into the "average yield" bit. Closing line is only accurate over a large enough sample so even if you will actually know your edge against the closing line (by betting say seconds before the closing line on a stale odd), you could still be incorrectly estimating your true edge for that particular game as the closing line could be off for that game.

[Okocha]

Its hilarious to see people who cant quantify their edge and as a result leaving money under the table in each bet creating a rationale that Kelly is flawed and that their staking methodology is the optimal one whilst the epitome of optimality in what maximizing E.G concerns is Kelly


Alfa you are stating thats impossible to determine your edge with a reasonable margin of error every wager,and thats probably right for 99 percent of the betting population but that critic to kelly in sportsbetting can be mitigated with the use of fractional kelly



in that scenario just be highly conservative in staking to Kelly, because you need longer periods to reach your blended edge than if u know your exact edge per wager (by this, i mean you can often have a model that gives EVs of say, +7%, 2,5%, +3%, -1% that over time pay +2%, but frequently turns out the odd negative EV wager)

Assuming your methodology is on point its not easy to overbet 0,25 0,33 kelly etc over time even 0,5 kelly if you have an edge and its fine to turn out betting a small percentage of negative ev wagers and i assume thats where you are failing in your rationale,its not the end of the world assuming you re not allocating constantly a big roll% on negative ev wagers,and then again if you do that your model is flawed not kelly,and you should strive to improve your model not bashing kelly

Danshan if you are not beating the market in markets where liquidity is relevant you should review if you have any edge at all,and staking is the least of your concerns specially if you have routinely big drifts against you dont need a big sample to jump the boat and reasses your methodology to acquire an edge

And danshan yeah its impossible to know ur edge in SB like in blackjack hence fractional kelly,and the less confident you are the more prudent you should be on your fraction preference

[Alfa1234]

Its hilarious to see people who cant quantify their edge and as a result leaving money under the table in each bet creating a rationale that Kelly is flawed and that their staking methodology is the optimal one whilst the epitome of optimality in what maximizing E.G concerns is Kelly

Alfa you are stating thats impossible to determine your edge with a reasonable margin of error every wager,and thats probably right for 99 percent of the betting population but that critic to kelly in sportsbetting can be mitigated with the use of fractional kelly



in that scenario just be highly conservative in staking to Kelly, because you need longer periods to reach your blended edge than if u know your exact edge per wager (by this, i mean you can often have a model that gives EVs of say, +7%, 2,5%, +3%, -1% that over time pay +2%, but frequently turns out the odd negative EV wager)

Assuming your methodology is on point its not easy to overbet 0,25 0,33 kelly etc over time even 0,5 kelly if you have an edge and its fine to turn out betting a small percentage of negative ev wagers and i assume thats where you are failing in your rationale,its not the end of the world assuming you re not allocating constantly a big roll% on negative ev wagers,and then again if you do your model is flawed not kelly


Danshan if you are not beating the market in markets where liquidity is relevant you should review if you have any edge at all,and staking is the least of your concerns specially if you have routinely big drifts against you dont need a big sample to jump the boat and reasses your methodology to acquire an edge

Your method of using Kelly is not really Kelly, now is it? Accurate use of Kelly (or fractional Kelly) requires adjusting each wager because you know the exact edge of that wager. You are describing a way of adjusting bet size on average edge and have in a way adapted Kelly to adjust for the risk of not knowing your real edge per wager. This is not much different than the method I use and described above.

[zorba74]

Your method of using Kelly is not really Kelly, now is it? Accurate use of Kelly (or fractional Kelly) requires adjusting each wager because you know the exact edge of that wager.

This is incorrect. You don't know your exact edge which is why you can't use Full Kelly; you can most certainly estimate your edge and use a modified version.

You are describing a way of adjusting bet size on average edge and have in a way adapted Kelly to adjust for the risk of not knowing your real edge per wager. This is not much different than the method I use and described above.

Kelly is a geometric mean maximization technique and you have not discovered a staking system that trumps it. I don’t care how much your forum buddies gas you up; sorry.

[Alfa1234]

Fractional Kelly is the exact same thing as full Kelly, it merely uses a smaller part of the roll for each bet. The entire point of Kelly is using an optimal stake compared to the edge you have on each particular bet. No matter how much you try to get around it, it's a fact. If you do not know your exact edge for a bet, you cannot use Kelly correctly. It's a fact and has nothing to do with optimising or being able to "estimate" your edge. It can never, ever work the way it's intended to work if you are guessing edge sizes. I never claimed I found a system that trumps it, I merely claimed the system I use works fine.

[u21c3f6]

It is about using the concept of Kelly that is important and not whether or not you count the coins in your change jar to make sure you are making the optimal Kelly wager.

Joe.

[Okocha]

Here s the guy betting Openers for 200 euros and with probably a monthly turnover of 500k- 1m at best tryng to reinvent the wheel in sportsbetting going against millionaires like thorp,posters like ganchrow poker joe and negating that virtually everyone at the top use or used kelly at some point until the market couldnt bare their wagers, because like a bacterie they expanded fast and were corrosive up until the point they couldnt find more space to expand
But i guess if you are betting 3-4 digits at best why spend your time tryng to quantify your edge?Just go after that Estonian Basket 20 ev bet

[zorba74]

Fractional Kelly is the exact same thing as full Kelly, it merely uses a smaller part of the roll for each bet.

No, it most certainly isn’t; if it was, they wouldn’t have two different names.

The entire point of Kelly is using an optimal stake compared to the edge you have on each particular bet.

In Sports betting, you don’t want to optimize your stake; if you did, one would use Full Kelly.

There are Pros that supposedly do use FK (I obviously disagree). I think they are nuts but it’s not my place to tell another man what his edge is or isn’t. All I can say about FK staking is that even if it is empirically right, it’s philosophically wrong (w/ apologies to Einstein who was dead wrong when he said something similar).

No matter how much you try to get around it, it's a fact. If you do not know your exact edge for a bet, you cannot use Kelly correctly.

Not trying to get around anything, you’re just not listening. I have said that one cannot know their exact edge, which makes FK not, “Fit for purpose,” when it comes to sports betting. You can certainly estimate your edge and use Fractional Kelly.

#Facts

It's a fact and has nothing to do with optimising or being able to "estimate" your edge. It can never, ever work the way it's intended to work if you are guessing edge sizes.

Sure it can’t.

I never claimed I found a system that trumps it, I merely claimed the system I use works fine.

I agree, you didn’t and I am sure your system is fine; I don’t know anything about it.

[HeeeHAWWWW]

There are Pros that supposedly do use FK (I obviously disagree).

I've heard of between 1/5 and 1/2 Kelly (and to throw my own hat in, I use about 2/5). I'm taking it as given anyone doing this has been through extensive calibration testing, uses a correct simultaneous modifier, and is aware of the dangers of correlated errors.

Full Kelly really doesn't make an awful lot of sense in the real world. You gain 33% more median EG compared to half-Kelly, but the quartiles and 5%/95%iles are insanely far away. Your gains are also seriously compromised by market limits - full Kelly will hit those slightly quicker on average, but it'll also near wipe you out far more often.

[zorba74]

It is about using the concept of Kelly that is important and not whether or not you count the coins in your change jar to make sure you are making the optimal Kelly wager.

Joe.

PokerJoe? It's GameBred.

Have you reconsidered my offer to join my Think Tank on how to overturn statistical laws?

We need you. Harabalob of whatever his name is never got back to me but now I Know it's because he was mulling NBA offers

[zorba74]

I've heard of between 1/5 and 1/2 Kelly (and to throw my own hat in, I use about 2/5). I'm taking it as given anyone doing this has been through extensive calibration testing, uses a correct simultaneous modifier, and is aware of the dangers of correlated errors.

Full Kelly really doesn't make an awful lot of sense in the real world. You gain 33% more median EG compared to half-Kelly, but the quartiles and 5%/95%iles are insanely far away. Your gains are also seriously compromised by market limits - full Kelly will hit those slightly quicker on average, but it'll also near wipe you out far more often.

Completely agree but supposedly these dudes are the real deal; at least one of them. I have no clue how you can achieve the granular accuracy necessary to pull FK off. I mean, how can you have an estimate w/out error bars and w/ FK there is no room for error bars. They may be talking out their ass, you never know and I don't know these guys personally but I have seen them on Twitter (and I'm hardly ever on) clowning people who don't use FK and like I said, supposedly one of the dudes is the real deal.

I was a 1/8 to 1/4 Kelly guy; I'm an Exchange Trader these days but I built my trading roll on the back of Fractional Kelly. I know for an unmitigated fact it works just fine which is why I take umbrage at these, "Kelly is useless," statements.

[danshan11]

if you use an average edge you are not using kelly as it is intended. 1/4 kelly half kelly one quarter leg of kelly, is all the exact same method with the only adjustment being the percentage of the full kelly wager you are using.

Alfa you say we cant use the closing line as the true edge and in theory this is correct but in practice it is absolutely the closest thing we have to the true implied that we currently know of and again without having access to the closing line when we bet we are doing one of two things if we are trying to use kelly.
We are betting the wrong amount on every bet where are estimate does not match the closing line
we are not using kelly at all if we are using an estimated edge on all bets this is closer to flat betting

[Alfa1234]

I didn't say Kelly is useless, I said it cannot work the way it's intended to work. You can certainly use a fractional Kelly if you are estimating edges and it can certainly grow your roll in a reasonably efficient manner, certainly more efficient than simply flat unit betting. It just can't do it 100% the way it's supposed to work if you have to guess the edge for a bet. That doesn't mean it won't work. I think we are arguïng about semantics here. In the end an adjusted "to win" amount for every bet is some form of Kelly if you are estimating your overall edge and don't change your estimated edge for every bet.

I have long ago reached my max betsizes for the things and time I'm betting and know my edge as it hasn't changed much over the years, so you could be right and I don't know what I'm talking about.

[danshan11]

that is why I said I did not want to get into a kelly discussion because for sports betting it seems nearly impossible to use it in its natural form. I use a kelly type method because I bet X to win 1 unit and that in a sense is a 3rd cousin of kelly.

and for people that truly have a long term edge and used kelly they are now betting max amount they can get down and there bets have literally nothing to do with kelly anymore. If your bankroll is 25,000,000 and you can max bet only 50,000, kelly is useless for them again because they are betting 50,000 every game period

[HeeeHAWWWW]

I didn't say Kelly is useless, I said it cannot work the way it's intended to work ..... I think we are arguïng about semantics here.

Indeed.

However, one of the curious practical aspects of Kelly is that errors don't matter, as long as your mean error is zero. Demonstrating via 10k simulations of 1000 bets, starting roll $10k, full Kelly, estimated edges of 1% to 6%.....

No errors: median $17.5k, quartiles $8.7k & $36.4k
5% errors: median $17.8k, quartiles $8.7k & $36.7k
10% errors: median $17.5k, quartiles $8.7k & $36.9k

They balance out perfectly! Not exactly intuitive, I'll admit.

[Okocha]

that is why I said I did not want to get into a kelly discussion because for sports betting it seems nearly impossible to use it in its natural form. I use a kelly type method because I bet X to win 1 unit and that in a sense is a 3rd cousin of kelly.

and for people that truly have a long term edge and used kelly they are now betting max amount they can get down and there bets have literally nothing to do with kelly anymore. If your bankroll is 25,000,000 and you can max bet only 50,000, kelly is useless for them again because they are betting 50,000 every game period

Danshan you are severely understimating the utility of Kelly for most of the bettors out there.in a previous post you ve mentioned market cap could be reached in weeks,which is indicative again of zero experience in betting but a lot of experience in BS propaganda

Assuming a bettor sphere of action is mid to top tier makets it could take years to reach market limits,something that wouldnt even occur without the aid of kelly and using the 1% with domino method or chair method you are advocating in this forum,which is basicallly the i dont know my edge method lets leave money on the table and mitigate the compounding effect method

[danshan11]

you are confusing me with someone else I did not say anything about market cap and nothing about a few weeks and you are confusing the stock market with sports betting. you are also confusing pro bettors and syndicates with think tank SBR members, MOST not all are just bettors by themselves staking a few hundred bucks and dont have a clue about edge. You are out here telling everyone use kelly and that is reckless. I am not a bettor advocate so go ahead and scream kelly from the rooftops.
facts
you cant use kelly correctly in sports betting because you NEVER EVER know the true edge of a particular game
Kelly for someone who has an edge and bankroll keeps growing will be done with kelly in a short time and again kelly would be useless

example
if you have a bankroll of 1,000,000,000 ( similar to your bankroll) and you can max bet through all outs 1,000,000 a game, it is impossible to use kelly in this situation so again kelly is useless

[HeeeHAWWWW]

facts
you cant use kelly correctly in sports betting because you NEVER EVER know the true edge of a particular game

See my last post, this is a myth.