Serious question....

[Richkas]

I've heard many different awnsers to this question so I thought maybe Ganch or Illusion could give me the correct awnser.

Say a guy had 20k and wanted to bet a set % of his bankroll on each game and try to win long term. Should it be 2%

[durito]

27.35353526% per

[smitch124]

8 decimal places? Wow you need pennies, a hacksaw and a microscope...

[Ganchrow]

This is where the Kelly Criterion comes in to play.

Optimal bet sizing will be a function of payout odds, edge, current bankroll, and other bets outstanding. In general, when placing a single win/lose-only bet, a full-Kelly bettor should wager to win a portion of bankroll equal to the edge on the given bet.

So for example, with a $20K bankroll, a bettor expecting to win a given bet at odds of -110 54% of the time (edge = 54% * 21 11 - 1 ≈ 3.0909%), should wager to win 3.0909% of bankroll. At -110, this corresponds to a wager size of 3.0909% * 11 10 = 3.4% of bankroll, or $680.

You can play around with the Kelly calculator to get an idea of how the numbers work out for different odds/edge combinations.

[Data]

So for example, with a $20K bankroll

This example adds to many posters' confusion and strenghtens their wrong beliefs that if they have $20K to bet with means that their Kelly bankroll is $20K.

[Ganchrow]

This example adds to many posters' confusion and strenghtens their wrong beliefs that if they have $20K to bet with means that their Kelly bankroll is $20K.

That's probably a topic best not covered on the first day of Kelly 101.

[Justin7]

If a guy is asking this, he has no clue what his edge is. In that case, I'd tell him to bet 0%, and put the 20k in a 1-year CD.

[SlickFazzer]

2% per bet has been a number that has been written about over the years by many people and seems like a safe number for most people.

[Matt Rain]

This example adds to many posters' confusion and strenghtens their wrong beliefs that if they have $20K to bet with means that their Kelly bankroll is $20K.

Any thread/further info on that topic? I think I understand what that means, but I'd like to know for sure.

Put simply, if you've got no edge, you've got no Kelly bankroll?

[Data]

Any thread/further info on that topic? I think I understand what that means, but I'd like to know for sure.

Put simply, if you've got no edge, you've got no Kelly bankroll?

No, not that. Here is a good starting point:

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

[dwaechte]

If you don't want to use Kelly or cannot estimate your edge, 1 or 2% is generally pretty safe, but it all depends on your risk tolerance.

[Ganchrow]

If you don't want to use Kelly or cannot estimate your edge, 1 or 2% is generally pretty safe, but it all depends on your risk tolerance.

I'd tend to disagree with this characterization.

Just because a player might either be unwilling to use Kelly or unable to estimate his edge doesn't suddenly render the straightforward mathematical principles of exponential growth/decay suddenly lose their moot.

Betting 1% or 2% of bankroll (or 10% or 0.1%) will only be "pretty safe" on a standalone bet provided that the payout odds and edge on said bet are (all else equal) sufficiently low/high, respectively. So if a player is betting sufficiently large ML underdogs and/or at sufficiently small edge he may very well find himself facing with expected bankroll decay irrespective of his personal prejudices towards Kelly/his skill at estimating his own edge.

Now admittedly, the impact of this on a bettor's expected utility will be even more pronounced for fractional Kelly (i.e., more risk averse) bettors for whom expected growth maximization is already too risky an endeavor, but this in no way should suggest that if a player chooses not to bet n-Kelly its recommendations lose their importance.

So does this make sense, or am I just overly tired and rambling?

[Data]

Now admittedly, the impact of this on a bettor's expected utility will be even more pronounced for fractional Kelly (i.e., more risk averse) bettors for whom expected growth maximization is already too risky an endeavor, but this in no way should suggest that if a player chooses not to bet n-Kelly its recommendations lose their importance.

So does this make sense, or am I just overly tired and rambling?

Now admittedly, I am drunk and tired too but this did not make any sense to me. Now, stop fooling around and give me my god damn utility function. People want utility function!

[kevk87]

New here, but would you measure utility using Stochastic Utility Assessment?

[tacomax]

New here, but would you measure utility using Stochastic Utility Assessment?

Nice first post.

[reno cool]

2% seems legit start for a few reasons.

1. You're not likely to be overbetting a winning strategy.
2. You're not likely to have a big edge and not be aware of it.
3. You give yourself a fair chance to be ahead over hundreds of trials even if your strategy is a slight looser.
4. You'll get in a bunch of plays before loosing a large chunk, which will further help you evaluate your strategy.

[gm2022]

If a guy is asking this, he has no clue what his edge is. In that case, I'd tell him to bet 0%, and put the 20k in a 1-year CD.

WISH I DID THAT!

[eidolon]

what are CDs in the market?

[JerseyShop101]

what are CDs in the market?

Music sold down at the flea market.

[Ganchrow]

2% seems legit start for a few reasons.

1. You're not likely to be overbetting a winning strategy.
2. You're not likely to have a big edge and not be aware of it.
3. You give yourself a fair chance to be ahead over hundreds of trials even if your strategy is a slight looser.
4. You'll get in a bunch of plays before loosing a large chunk, which will further help you evaluate your strategy.

1. Then why not bet 0.45%? What if you're betting big dogs?
2. The bettor, by assumption, is already ignorant of this.
3. Then why not bet 0.27%? And anyway, is this really the appropriate metric for an advantage bettor (namely, one's probability of profitability even if one's strategy were a loser)? If so why not just bet Martingale? And don't forget, you're likely to give yourself a better than fair chance of being down "after hundreds of trials even if your strategy is a slight" loser. Ought that not factor in as well?
4. Then why not bet 0.11?

Look, the point is that the selection of 2% per bet without considering odds or edge is essentially arbitrary and in many instances flat-out wrong and detrimental to bankroll growth. That said, 2% per bet is probably a fair enough stake size for the typical recreational bettor and by extension probably the neophyte advantage bettor as well.

But let's not imbue 1% to 2% per bet with magical qualities (which seems to be done frequently on many message boards) as if they somehow represent the Platonic ideal of bet size for underdog, favorite, and teaser bettors alike. They doesn't. They're just a made up numbers shat out the proverbial anus of Groupthink, Minister of Conventional Wisdom.

Now if you're placing bets near even at some as-yet-unknown but clear and palpable advantage, is 1% or 2% probably a fair number? Yeah, it probably is. But the degree to which is it is (or is not not) is based on a whole lot more than just "risk prerefences and such a bettor's time is most likely better spent understanding the basics of Kelly and coming up with a reasonable estimation of his per-bet edge than of building twin altars to the numbers 1 and 2 and allowing the gods to make all his bet sizing decisions for him.

[Ganchrow]

New here, but would you measure utility using Stochastic Utility Assessment?

You're going to have rephrase that in a more meaningful and less jargonistic manner.

EDIT: Insofar as this corresponds to a technique derived solely for the purposes of technical stock analysis then (speaking only for myself), the answer would be a fim no.

[Data]

New here, but would you measure utility using Stochastic Utility Assessment?

No we would not as this would mean work. We are here to let Ganchrow do all the work and give us the answers in a form of nice and ready to use formulas.

[reno cool]

1. Then why not bet 0.45%? What if you're betting big dogs?
2. The bettor, by assumption, is already ignorant of this.
3. Then why not bet 0.27%? And anyway, is this really the appropriate metric for an advantage bettor (namely, one's probability of profitability even if one's strategy were a loser)? If so why not just bet Martingale? And don't forget, you're likely to give yourself a better than fair chance of being down "after hundreds of trials even if your strategy is a slight" loser. Ought that not factor in as well?
4. Then why not bet 0.11?

Look, the point is that the selection of 2% per bet without considering odds or edge is essentially arbitrary and in many instances flat-out wrong and detrimental to bankroll growth. That said, 2% per bet is probably a fair enough stake size for the typical recreational bettor and by extension probably the neophyte advantage bettor as well.

But let's not imbue 1% to 2% per bet with magical qualities (which seems to be done frequently on many message boards) as if they somehow represent the Platonic ideal of bet size for underdog, favorite, and teaser bettors alike. They doesn't. They're just a made up numbers shat out the proverbial anus of Groupthink, Minister of Conventional Wisdom.

Now if you're placing bets near even at some as-yet-unknown but clear and palpable advantage, is 1% or 2% probably a fair number? Yeah, it probably is. But the degree to which is it is (or is not not) is based on a whole lot more than just "risk prerefences and such a bettor's time is most likely better spent understanding the basics of Kelly and coming up with a reasonable estimation of his per-bet edge than of building twin altars to the numbers 1 and 2 and allowing the gods to make all his bet sizing decisions for him.

Many players may think they have handicapping skills enough to overcome the vig inherent in sports betting. If such a player wants to bet a % of bankroll 2% is a better # than most to try.
If for no other reason than that his edge is more likely to be around 2% than >5%.
I assume that he is placing about even money bets, or to win 2%.
Now, if his edge is less than 1% but bigger than 0, he would need to make close to a million bets to be confident of a profit.(betting .5% or flatbetting) Plus it has to be close to impossible to estimate ones edge to within .1% when handicapping. So if he's costing himself profit by overbetting at 2% it's because his edge isn't high enough anyway.

Personally, I feel 2% would work for a fairly common bettor.

But this is still a generic approach,

[Bullajami]

Now if you're placing bets near even at some as-yet-unknown but clear and palpable advantage, is 1% or 2% probably a fair number? Yeah, it probably is.

I assume that he is placing about even money bets, or to win 2%.

Personally, I feel 2% would work for a fairly common bettor.

But this is still a generic approach,

If Ganch and reno can essentially agree on a bet-sizing strategy, how far off can peace in the Middle East really be?