vpFREE2 Forums

Bob Dancer's LV Advisor Column - 26 MAR 2013 (skill vs luck?)

I found these figures to be surprising, but I assume the math behind them is solid. I am interpreting the figures as saying "luck is xx% and skill is 100-xx% in contributing to your likelihood of experiencing real-life returns consistent with the theoretical return of the game after nn hands" - is that correct? If so, it's fascinating to me that the skill factor doesn't begin to exceed the luck factor until 800,000 hands have been played!

My peak play rate (which I know includes about one error per hour) is about 800 hands per hour, and so that's probably well over my lifetime play range, as I think I only play 4 hours a day, for two-four days a month, and even if it's several times that, it's still many lifetimes.

Even for a pro playing 40 hours a week, 50 weeks a year, and 1,000 hands per hour, which is 2,000,000 hands a year, getting the luck factor under 10% would take a full lifetime of play! Seems a little too tough to me, considering that there are reportedly many successful pro's, and that no one seems to know someone who plays perfectly and yet loses year after year after year, which according to these figures, should be happening to a significant percentage of such players.

As I'm curious what the luck/skill factor is for other games (e.g., blackjack with card-counting producing a 0.5% edge for the player or a 1.0% edge for the player), can you (or anyone) provide a short math tutorial on how to do the calculations that generate such a chart?

THANKS!

--BG

···

======================

5b. Re: Bob Dancer's LV Advisor Column - 26 MAR 2013

Edge can be positive or negative, one side or the other has
the edge, for example here's perfect 9/6 Jacks (a negative
edge game):

at 1 hand, luck = 99.9%
at 100 hands, luck = 98.9%
at 1,000 hands, luck = 96.5%
at 2,000 hands, luck = 95.2%
at 10,000 hands, luck = 89.8%
at 100,000 hands, luck = 73.6%
at 200,000 hands, luck = 66.4%
at 400,000 hands, luck = 58.3%
at 800,000 hands, luck = 49.7%
at 2,000,000 hands, luck = 38.4%
at 5,000,000 hands, luck = 28.3%
at 10,000,000 hands, luck = 21.8%
at 100,000,000 hands, luck = 8.1%

I guess you're postulating a player who plays Jacks so badly
that they have a -3.5% edge? That would look like this
(close to Bob's numbers):

at 1 hand, luck = 99.2%
at 100 hands, luck = 92.7%
at 1,000 hands, luck = 80%
at 2,000 hands, luck = 73.8%
at 10,000 hands, luck = 55.8%
at 100,000 hands, luck = 28.5%
at 200,000 hands, luck = 22%
at 400,000 hands, luck = 16.6%
at 800,000 hands, luck = 12.4%
at 2,000,000 hands, luck = 8.2%
at 5,000,000 hands, luck = 5.3%
at 10,000,000 hands, luck = 3.8%
at 100,000,000 hands, luck = 1.2%

Luck(hands) = SD = sqrt(variance x hands)
Edge(hands) = edge x hands
%Luck(hands) = 100 x Luck/(Luck + Edge)
Luck=Edge at N0 hands, N0 = variance/edge/edge hands
For FPDW+0.25%, variance=26, edge=.01, N0 = 260,000 hands

other N0's:

west-point.org/users/usma1955/20228/V/Bank_NO1.htm

···

--- In vpFREE@yahoogroups.com, Barry Glazer <b.glazer@...> wrote:

I found these figures to be surprising, but I assume the math behind them is solid. I am interpreting the figures as saying "luck is xx% and skill is 100-xx% in contributing to your likelihood of experiencing real-life returns consistent with the theoretical return of the game after nn hands" - is that correct? If so, it's fascinating to me that the skill factor doesn't begin to exceed the luck factor until 800,000 hands have been played!

My peak play rate (which I know includes about one error per hour) is about 800 hands per hour, and so that's probably well over my lifetime play range, as I think I only play 4 hours a day, for two-four days a month, and even if it's several times that, it's still many lifetimes.

Even for a pro playing 40 hours a week, 50 weeks a year, and 1,000 hands per hour, which is 2,000,000 hands a year, getting the luck factor under 10% would take a full lifetime of play! Seems a little too tough to me, considering that there are reportedly many successful pro's, and that no one seems to know someone who plays perfectly and yet loses year after year after year, which according to these figures, should be happening to a significant percentage of such players.

As I'm curious what the luck/skill factor is for other games (e.g., blackjack with card-counting producing a 0.5% edge for the player or a 1.0% edge for the player), can you (or anyone) provide a short math tutorial on how to do the calculations that generate such a chart?

THANKS!

--BG

> 5b. Re: Bob Dancer's LV Advisor Column - 26 MAR 2013
>
> Edge can be positive or negative, one side or the other has
> the edge, for example here's perfect 9/6 Jacks (a negative
> edge game):
>
> at 1 hand, luck = 99.9%
> at 100 hands, luck = 98.9%
> at 1,000 hands, luck = 96.5%
> at 2,000 hands, luck = 95.2%
> at 10,000 hands, luck = 89.8%
> at 100,000 hands, luck = 73.6%
> at 200,000 hands, luck = 66.4%
> at 400,000 hands, luck = 58.3%
> at 800,000 hands, luck = 49.7%
> at 2,000,000 hands, luck = 38.4%
> at 5,000,000 hands, luck = 28.3%
> at 10,000,000 hands, luck = 21.8%
> at 100,000,000 hands, luck = 8.1%
>
> I guess you're postulating a player who plays Jacks so badly
> that they have a -3.5% edge? That would look like this
> (close to Bob's numbers):
>
> at 1 hand, luck = 99.2%
> at 100 hands, luck = 92.7%
> at 1,000 hands, luck = 80%
> at 2,000 hands, luck = 73.8%
> at 10,000 hands, luck = 55.8%
> at 100,000 hands, luck = 28.5%
> at 200,000 hands, luck = 22%
> at 400,000 hands, luck = 16.6%
> at 800,000 hands, luck = 12.4%
> at 2,000,000 hands, luck = 8.2%
> at 5,000,000 hands, luck = 5.3%
> at 10,000,000 hands, luck = 3.8%
> at 100,000,000 hands, luck = 1.2%
>

It's not a good way of thinking about luck versus skill as opposed to calculating a confidence interval for your wins and losses. It could be you expect to win a ten milliom dollars with a sd of one million and that would be called 10% luck. However the lowest number in the confidence interval is above 8 million.

To be more nitpicky, calling sd risk is not appropriate. It's a measure of risk for sure, but there are many possible normalizations that you should consider to make the two things comparable. There isn't much logic to blindly dividing edge by standard deviation.

···

--- In vpFREE@yahoogroups.com, Barry Glazer <b.glazer@...> wrote:

I found these figures to be surprising, but I assume the math behind them is solid. I am interpreting the figures as saying "luck is xx% and skill is 100-xx% in contributing to your likelihood of experiencing real-life returns consistent with the theoretical return of the game after nn hands" - is that correct? If so, it's fascinating to me that the skill factor doesn't begin to exceed the luck factor until 800,000 hands have been played!

My peak play rate (which I know includes about one error per hour) is about 800 hands per hour, and so that's probably well over my lifetime play range, as I think I only play 4 hours a day, for two-four days a month, and even if it's several times that, it's still many lifetimes.

Even for a pro playing 40 hours a week, 50 weeks a year, and 1,000 hands per hour, which is 2,000,000 hands a year, getting the luck factor under 10% would take a full lifetime of play! Seems a little too tough to me, considering that there are reportedly many successful pro's, and that no one seems to know someone who plays perfectly and yet loses year after year after year, which according to these figures, should be happening to a significant percentage of such players.

As I'm curious what the luck/skill factor is for other games (e.g., blackjack with card-counting producing a 0.5% edge for the player or a 1.0% edge for the player), can you (or anyone) provide a short math tutorial on how to do the calculations that generate such a chart?

THANKS!

--BG

> 5b. Re: Bob Dancer's LV Advisor Column - 26 MAR 2013
>
> Edge can be positive or negative, one side or the other has
> the edge, for example here's perfect 9/6 Jacks (a negative
> edge game):
>
> at 1 hand, luck = 99.9%
> at 100 hands, luck = 98.9%
> at 1,000 hands, luck = 96.5%
> at 2,000 hands, luck = 95.2%
> at 10,000 hands, luck = 89.8%
> at 100,000 hands, luck = 73.6%
> at 200,000 hands, luck = 66.4%
> at 400,000 hands, luck = 58.3%
> at 800,000 hands, luck = 49.7%
> at 2,000,000 hands, luck = 38.4%
> at 5,000,000 hands, luck = 28.3%
> at 10,000,000 hands, luck = 21.8%
> at 100,000,000 hands, luck = 8.1%
>
> I guess you're postulating a player who plays Jacks so badly
> that they have a -3.5% edge? That would look like this
> (close to Bob's numbers):
>
> at 1 hand, luck = 99.2%
> at 100 hands, luck = 92.7%
> at 1,000 hands, luck = 80%
> at 2,000 hands, luck = 73.8%
> at 10,000 hands, luck = 55.8%
> at 100,000 hands, luck = 28.5%
> at 200,000 hands, luck = 22%
> at 400,000 hands, luck = 16.6%
> at 800,000 hands, luck = 12.4%
> at 2,000,000 hands, luck = 8.2%
> at 5,000,000 hands, luck = 5.3%
> at 10,000,000 hands, luck = 3.8%
> at 100,000,000 hands, luck = 1.2%
>

Well, as much as I'd like to take credit, I really can't. It's called the "Sharpe Ratio", proposed by a Nobel laureate:

google.com/search?q=sharpe+ratio

···

--- In vpFREE@yahoogroups.com, "vpplayer88" <vpplayer88@...> wrote:

There isn't much logic to blindly dividing edge by standard deviation.

Fix link:
http://www.west-point.org/users/usma1955/20228/V/Bank_NO1.htm

···

--- In vpFREE@yahoogroups.com, "nightoftheiguana2000" <nightoftheiguana2000@...> wrote:

Luck(hands) = SD = sqrt(variance x hands)
Edge(hands) = edge x hands
%Luck(hands) = 100 x Luck/(Luck + Edge)
Luck=Edge at N0 hands, N0 = variance/edge/edge hands
For FPDW+0.25%, variance=26, edge=.01, N0 = 260,000 hands

other N0's:

west-point.org/users/usma1955/20228/V/Bank_NO1.htm

--- In vpFREE@yahoogroups.com, Barry Glazer <b.glazer@> wrote:
>
> I found these figures to be surprising, but I assume the math behind them is solid. I am interpreting the figures as saying "luck is xx% and skill is 100-xx% in contributing to your likelihood of experiencing real-life returns consistent with the theoretical return of the game after nn hands" - is that correct? If so, it's fascinating to me that the skill factor doesn't begin to exceed the luck factor until 800,000 hands have been played!
>
> My peak play rate (which I know includes about one error per hour) is about 800 hands per hour, and so that's probably well over my lifetime play range, as I think I only play 4 hours a day, for two-four days a month, and even if it's several times that, it's still many lifetimes.
>
> Even for a pro playing 40 hours a week, 50 weeks a year, and 1,000 hands per hour, which is 2,000,000 hands a year, getting the luck factor under 10% would take a full lifetime of play! Seems a little too tough to me, considering that there are reportedly many successful pro's, and that no one seems to know someone who plays perfectly and yet loses year after year after year, which according to these figures, should be happening to a significant percentage of such players.
>
> As I'm curious what the luck/skill factor is for other games (e.g., blackjack with card-counting producing a 0.5% edge for the player or a 1.0% edge for the player), can you (or anyone) provide a short math tutorial on how to do the calculations that generate such a chart?
>
> THANKS!
>
> --BG
> ======================
>
> > 5b. Re: Bob Dancer's LV Advisor Column - 26 MAR 2013
> >
> > Edge can be positive or negative, one side or the other has
> > the edge, for example here's perfect 9/6 Jacks (a negative
> > edge game):
> >
> > at 1 hand, luck = 99.9%
> > at 100 hands, luck = 98.9%
> > at 1,000 hands, luck = 96.5%
> > at 2,000 hands, luck = 95.2%
> > at 10,000 hands, luck = 89.8%
> > at 100,000 hands, luck = 73.6%
> > at 200,000 hands, luck = 66.4%
> > at 400,000 hands, luck = 58.3%
> > at 800,000 hands, luck = 49.7%
> > at 2,000,000 hands, luck = 38.4%
> > at 5,000,000 hands, luck = 28.3%
> > at 10,000,000 hands, luck = 21.8%
> > at 100,000,000 hands, luck = 8.1%
> >
> > I guess you're postulating a player who plays Jacks so badly
> > that they have a -3.5% edge? That would look like this
> > (close to Bob's numbers):
> >
> > at 1 hand, luck = 99.2%
> > at 100 hands, luck = 92.7%
> > at 1,000 hands, luck = 80%
> > at 2,000 hands, luck = 73.8%
> > at 10,000 hands, luck = 55.8%
> > at 100,000 hands, luck = 28.5%
> > at 200,000 hands, luck = 22%
> > at 400,000 hands, luck = 16.6%
> > at 800,000 hands, luck = 12.4%
> > at 2,000,000 hands, luck = 8.2%
> > at 5,000,000 hands, luck = 5.3%
> > at 10,000,000 hands, luck = 3.8%
> > at 100,000,000 hands, luck = 1.2%
> >
>

I know what a Sharpe ratio is. It doesn't have anything to do with dividing wins and losses into luck versus skill. It's a measure of return per unit of risk, for one possible normalization of risk.

For companies that can freely borrow and lend at the risk free rate, they want to maximize a sharpe ratio. It has applications in portfolio theory for determining your optimal allocation to risky assets.

This has nothing to do with the decomposition you are trying to give us. 95% or 99% confidence intervals are very useful for that purpose. There is some actual interpretable content in them.

···

--- In vpFREE@yahoogroups.com, "nightoftheiguana2000" <nightoftheiguana2000@...> wrote:

--- In vpFREE@yahoogroups.com, "vpplayer88" <vpplayer88@> wrote:
>There isn't much logic to blindly dividing edge by standard deviation.

Well, as much as I'd like to take credit, I really can't. It's called the "Sharpe Ratio", proposed by a Nobel laureate:

google.com/search?q=sharpe+ratio