- Joined
- 7 December 2014
- Posts
- 357
- Reactions
- 0
At some stage medicare funding is likely to be reduced IMO re:IVF.
Thought Bubbles from the Deep
- Thread starter DeepState
- Start date
DeepState
Multi-Strategy, Quant and Fundamental
- Joined
- 30 March 2014
- Posts
- 1,615
- Reactions
- 81
HNY and greetings from Tokyo.
A man walks into a hotel lobby in Minato-ku and finds himself seated next to a long term investor.
Man: Konbanwa, Gaijin. I have infinite funds. I see you are a long term investor, impervious to drawdowns.
LTI: Indeed. I set off metal detectors at airports.
Man: Truly impressive.
LTI: So, why are we talking?
Man: Analysis of your trades has shown that you get 60% of your trades right and you have positive expectancy. That's truly incredible!
LTI: Indeed. I set off metal detectors at airports. It's a real hassle I tell you. Much easier to fly private.
Man: Ok then. I offer you a trade where I will take whatever is in your portfolio. Don't worry about it, I am a cousin of Kuroda, and we can print as much money is needed to support our pledge to you.
LTI: What?
Man: It gets better. I will match your 60% hit rate. I will create a coin which flips AWESOME 60% of the time and DRAWDOWN 40% of the time. Each time AWESOME comes up, I will increase your funds at the time by 90%.
LTI: Awesome.
Man: But, given this is investment, if DRAWDOWN comes up, I am going to take 90% of whatever is in your funds at the time.
LTI: Of course.
Man: So each time you do this, you have your majestic positive expectancy. You expect to make 0.9 x 0.6 - 0.9 x 0.4 = 18%. Almost as good as Buffett.
LTI: Yep, I'm about as good as Buffett, so that sounds right.
Man: But, what about those drawdowns, can you take them?
LTI: I invest for the long term. I set of metal detectors at airports. Drawdowns don't concern me.
Man: So, if we think of one coin flip as being a month's outcomes, we can do this maybe a couple of hundred times? Let's say 240? Twenty years equivalent? Compounding the entire portfolio each period, just like you would as a long term investor who doesn't need to worry about such things as living off cashflow from the portfolio.
LTI: You're going to need Kuroda to make good on all the money you will owe me in a few minutes. You're the dumbest schmuck I have ever come across and, as a rampant capitalist, am duty bound to lead Japan into another round of quantitative easing. OK. My funds have been transferred. Let's roll. Positive expectancy. Long term investment horizon. Impervious to drawdowns. You are screwed.
Man: (Pours Sake sitting in a hotel lobby at 1:00am Tokyo time). Let's roll.
What happens next?
---
PPP basket (Sake, Espresso, Sushi) suggests that YEN is close to parity with AUD. I have no signal.
A man walks into a hotel lobby in Minato-ku and finds himself seated next to a long term investor.
Man: Konbanwa, Gaijin. I have infinite funds. I see you are a long term investor, impervious to drawdowns.
LTI: Indeed. I set off metal detectors at airports.
Man: Truly impressive.
LTI: So, why are we talking?
Man: Analysis of your trades has shown that you get 60% of your trades right and you have positive expectancy. That's truly incredible!
LTI: Indeed. I set off metal detectors at airports. It's a real hassle I tell you. Much easier to fly private.
Man: Ok then. I offer you a trade where I will take whatever is in your portfolio. Don't worry about it, I am a cousin of Kuroda, and we can print as much money is needed to support our pledge to you.
LTI: What?
Man: It gets better. I will match your 60% hit rate. I will create a coin which flips AWESOME 60% of the time and DRAWDOWN 40% of the time. Each time AWESOME comes up, I will increase your funds at the time by 90%.
LTI: Awesome.
Man: But, given this is investment, if DRAWDOWN comes up, I am going to take 90% of whatever is in your funds at the time.
LTI: Of course.
Man: So each time you do this, you have your majestic positive expectancy. You expect to make 0.9 x 0.6 - 0.9 x 0.4 = 18%. Almost as good as Buffett.
LTI: Yep, I'm about as good as Buffett, so that sounds right.
Man: But, what about those drawdowns, can you take them?
LTI: I invest for the long term. I set of metal detectors at airports. Drawdowns don't concern me.
Man: So, if we think of one coin flip as being a month's outcomes, we can do this maybe a couple of hundred times? Let's say 240? Twenty years equivalent? Compounding the entire portfolio each period, just like you would as a long term investor who doesn't need to worry about such things as living off cashflow from the portfolio.
LTI: You're going to need Kuroda to make good on all the money you will owe me in a few minutes. You're the dumbest schmuck I have ever come across and, as a rampant capitalist, am duty bound to lead Japan into another round of quantitative easing. OK. My funds have been transferred. Let's roll. Positive expectancy. Long term investment horizon. Impervious to drawdowns. You are screwed.
Man: (Pours Sake sitting in a hotel lobby at 1:00am Tokyo time). Let's roll.
What happens next?
---
PPP basket (Sake, Espresso, Sushi) suggests that YEN is close to parity with AUD. I have no signal.
LTI gets too close to really powerful magnet - but unfortunately the rest of his body is not close enough.
.........
risk of ruin is just about guaranteed
At any point if two trades in a row are wrong he has lost 99% of the equity figure at the time. Correct?
- Joined
- 6 January 2009
- Posts
- 2,300
- Reactions
- 1,130
My first thought was sounds like a good wager for LTI, however how wrong I was.
LTI can never win and will eventually have less than 1 cents left of capital no matter if he gets 6 straight win followed by 4 straight losses or vise versa.
For LTI to win he needs the payout to be no greater than %38 with 20% being the sweet spot.
Going to have a play around with some numbers on win loss ratios, % lost on each trade and % gained on winning trades.
Thanks for brain game TFBD.
Cheers
LTI can never win and will eventually have less than 1 cents left of capital no matter if he gets 6 straight win followed by 4 straight losses or vise versa.
For LTI to win he needs the payout to be no greater than %38 with 20% being the sweet spot.
Going to have a play around with some numbers on win loss ratios, % lost on each trade and % gained on winning trades.
Thanks for brain game TFBD.
Cheers
- Joined
- 25 February 2011
- Posts
- 5,688
- Reactions
- 1,231
Mathematically the sequence of losses should prove irrelevant to the final outcome as it is the result of a series of multiplications. Provided 60% of the outcomes in the sample are wins then 18% profit will be the net result. This is may be an example of yet another sound theory that is prone to failure when when applied in the real world..
- Joined
- 1 May 2007
- Posts
- 2,904
- Reactions
- 52
Think the problem here is win amount and loss amount is dynamic rather than fixed.
- Joined
- 25 July 2010
- Posts
- 1,120
- Reactions
- 284
Actually, the theory is even wrong in this case. If you open Excel and actually play out the whole thing, using the probabilities mentioned, the LTI always loses, and by a big margin.
If they were playing for a fixed amount, rather than an amount relative to total capital, this would be different (but not necessarily reflective of an investment strategy).
To be completely honest, I'm not exactly sure how to cleanly represent the actual theory applied, only that the calculation provided is misleading because it uses relative amounts of capital (i.e. 0.9 * 0.6 - 0.9 * 0.4 = 18% is actually not representative of the outcome).
FWIW - the average winnings would be 18%, but this would not be a weighted average, which I believe is where the problem lies...
- Joined
- 6 January 2009
- Posts
- 2,300
- Reactions
- 1,130
- Joined
- 30 September 2012
- Posts
- 742
- Reactions
- 379
Man has infinite capital, LTI does not. Man survives.
- Joined
- 1 May 2007
- Posts
- 2,904
- Reactions
- 52
Here's a hint:
Double or nothing at 50% odds
50% *2.0 + 50%*0.0 = 1.0
However we all know nobody sane keeps playing this!
Double or nothing at 50% odds
50% *2.0 + 50%*0.0 = 1.0
However we all know nobody sane keeps playing this!
With a 60% win rate over 240 occurrences you are as likely as not to hit at least a losing streak of 6.
Loosing 90% of your account 6 times = 10%^6. So you have an average chance of drawing your account down to 0.0001% of whatever it was prior to the losing streak. (ie $1,000,000 becomes $1)
It’s a play on percentages. A 90% markdown requires a 1000% mark-up. If the man was offering you 10 fold each time you win to compensate for the 90% when you lose then you would actually have the an 18% expectancy.
Loosing 90% of your account 6 times = 10%^6. So you have an average chance of drawing your account down to 0.0001% of whatever it was prior to the losing streak. (ie $1,000,000 becomes $1)
It’s a play on percentages. A 90% markdown requires a 1000% mark-up. If the man was offering you 10 fold each time you win to compensate for the 90% when you lose then you would actually have the an 18% expectancy.
- Joined
- 25 February 2011
- Posts
- 5,688
- Reactions
- 1,231
Well spotted klogg and craft. It turns out that the trader would lose approximately 99.53% of his capital within the first 10 trades (i.e. 0.1^4 × 1.9 ^ 6)
Raise that to the power of 24 and he will be near enough to flat broke.
- Joined
- 25 July 2010
- Posts
- 1,120
- Reactions
- 284
Sorry to ask, but could you please post the math to get to the 18% expectancy? (i.e. how'd you figure out it was 1000%).
I'm failing to figure it out...
- Joined
- 25 February 2011
- Posts
- 5,688
- Reactions
- 1,231
(1 /(1-0.9)) × 100
- Joined
- 25 February 2011
- Posts
- 5,688
- Reactions
- 1,231
However, I believe the expecta cy to be considerably higher than 18 %. More like 67.8%edit (make that 61.5%) edit (no o second thoughts I will stick with 67.8 %)
Sorry Klogg you probably can’t make sense of it because it’s wrong – its only 900%. (quick posts on holidays - not always accurate
- actually I'm just make most of this **** up so its probably all wrong)The old mark-up / margin calculation.
Start $100 loose $90 = new capital of $10. Require $90 from $10 to regain position = 90/10 = 900%
Once you compensate for the geometric vs arithmetic effect the expectancy is generated solely by the 60% win rate.
1x.6 – 1*.4 = 20%
Its an extreme example of something that most should be aware of with their expectancy calculations but ....
Loss in original example on a geometric basis is 10 times bigger than the win. So expectancy that you gonna get in real life is 0.6*.9 – 0.4*9 =NEGATIVE 3.06 (give or take a bourbon or two)
- Joined
- 25 July 2010
- Posts
- 1,120
- Reactions
- 284
Sorry Klogg you probably can’t make sense of it because it’s wrong – its only 900%. (quick posts on holidays - not always accurate
The old mark-up / margin calculation.
Start $100 loose $90 = new capital of $10. Require $90 from $10 to regain position = 90/10 = 900%
Once you compensate for the geometric vs arithmetic effect the expectancy is generated solely by the 60% win rate.
1x.6 – 1*.4 = 20%
Its an extreme example of something that most should be aware of with their expectancy calculations but ....
Loss in original example on a geometric basis is 10 times bigger than the win. So expectancy that you gonna get in real life is 0.6*.9 – 0.4*9 =NEGATIVE 3.06 (give or take a bourbon or two)
Thanks craft - I didn't convert the relative to the absolute (line in italics), so was failing to calculate the correct expectancy.
Much appreciated.
Klogg, you're closer to the correct answer than craft this time. Remember, he is in holiday mode.
The expectancy in the story is actually POSITIVE 18%.
If these were straightforward even-money bets, the expectancy would be positive 20%.
You bet 100$. You win 100$ 60% of the time and you lose 100$ 40% of the time.
In the story you only win 90 rather than 100, but you also only lose 90 rather than 100.
Why then does LTI wipe out his capital?
Because he has to bet ALL of his capital every time. The Man says so, and LTI who knows about expectancy but can't apply correct position sizing according to The Ultimate Guide by Van Tharp gets wiped out.
If he could just bet a fraction of his capital 240 times, he would get rich, but he can't. The length of the losing runs don't help, of course, as craft points out, but even if there were never two consecutive losses, he would still wipe out.
DeepState has made up a very clever little story. I like the bit about "almost as good as Buffett", LTI confusing expectancy with annual growth rate.
The expectancy in the story is actually POSITIVE 18%.
If these were straightforward even-money bets, the expectancy would be positive 20%.
You bet 100$. You win 100$ 60% of the time and you lose 100$ 40% of the time.
In the story you only win 90 rather than 100, but you also only lose 90 rather than 100.
Why then does LTI wipe out his capital?
Because he has to bet ALL of his capital every time. The Man says so, and LTI who knows about expectancy but can't apply correct position sizing according to The Ultimate Guide by Van Tharp gets wiped out.
If he could just bet a fraction of his capital 240 times, he would get rich, but he can't. The length of the losing runs don't help, of course, as craft points out, but even if there were never two consecutive losses, he would still wipe out.
DeepState has made up a very clever little story. I like the bit about "almost as good as Buffett", LTI confusing expectancy with annual growth rate.
- Joined
- 25 July 2010
- Posts
- 1,120
- Reactions
- 284
Actually, I'm quite confident craft is correct. Converting to an absolute value is the right way of looking at it I believe.
If you use the 90% value, it's not an even bet each time, so the expectancy is thrown out. The same would occur if you went with a 5% value, but the change in capital wouldn't be so dramatic.
