yet another coin flipping experiment

[dr_trick]

hi everyone,

i wrote a blog post (and paper) on a coin flipping experiment. i was testing a simple strategy for bet sizing / money management for a coin flipping game. the results look interesting, i have not applied the strategy to trading yet. i still need to look at variable returns and other real world stuff.

i am interested in finding out what the forum thinks about this. the blog post is at http://thetulipfund.wordpress.com/2014/11/29/a-little-experiment/

cheers

 

[Forexmospherian]

Hi Dr Trick

Another recipe for disaster

I can only relate it to FX trading - as thats the only trading I have undertaken over 13 yrs - with approx 7 yrs as full time.

Why dont you learn the science part - then that art and then all the skills needed and if you are doing it part time you will be far better off when you understand how the market really moves - which might take 3-7 yrs - depending how many times you get taken off track.

The FX market is not totally random - its based on logic - gameplay - manipulation - deception etc etc - forget coin flips - fridge magnets - weather and all other silly methods - Yes they might make you money for a few days - or a few weeks - or if your are extremely lucky a few months - and then you will lose it all - or if not all - much of your gains

I would then say carrying on with the method is about 95% certain to fail within a year or two - unless you only do day 5 trades a year ;-)))

Good Luck - and buy a lottery ticket with one of your wins - as that might be more successful

Regards


F

 

[tomorton]

Interesting puzzle for an afternoon when the markets are closed but I stick to the principle that random actions will give random results.

Your 3rd and 4th specimen games, with random H or T tosses, prove this - whether you are in the black or in the red after any given toss is random. There is no way to know that you should exit the game after 1 toss or 10 or 7 or 4 etc. But is your conclusion that at some point a series of tosses will always give a positive balance, and trading should then cease with profits banked?

 

[dr_trick]

Interesting puzzle for an afternoon when the markets are closed but I stick to the principle that random actions will give random results.

Your 3rd and 4th specimen games, with random H or T tosses, prove this - whether you are in the black or in the red after any given toss is random. There is no way to know that you should exit the game after 1 toss or 10 or 7 or 4 etc. But is your conclusion that at some point a series of tosses will always give a positive balance, and trading should then cease with profits banked?

hi,

thanks for your reply.

you are right, i was looking at the case that one can only exit the game after all 10 coin flips.

thanks for the feedback again.

 

[dr_trick]

Hi Dr Trick

Another recipe for disaster

I can only relate it to FX trading - as thats the only trading I have undertaken over 13 yrs - with approx 7 yrs as full time.

Why dont you learn the science part - then that art and then all the skills needed and if you are doing it part time you will be far better off when you understand how the market really moves - which might take 3-7 yrs - depending how many times you get taken off track.

The FX market is not totally random - its based on logic - gameplay - manipulation - deception etc etc - forget coin flips - fridge magnets - weather and all other silly methods - Yes they might make you money for a few days - or a few weeks - or if your are extremely lucky a few months - and then you will lose it all - or if not all - much of your gains

I would then say carrying on with the method is about 95% certain to fail within a year or two - unless you only do day 5 trades a year ;-)))

Good Luck - and buy a lottery ticket with one of your wins - as that might be more successful

Regards


F

Hi,

i think i should rephrase my initial post. the experiment is not a trading strategy.

what motivated me was this: given any trading strategy, how can one manage bet/trade sizes so that it maximises the probability of being profitable after n trades.

a simple way of doing this is to do Kelly sizing or betting only 2% of your portfolio on each trade or for the more adventurous a martingale sizing strategy

the simplest strategy to test this with was with the coin flip experiment - where the coin flip is the outcome of a trade i.e. profit or loss.

thanks for your comments.

cheers

 

[Trader333]

You could also use a monte carlo simulation Excel spreadhseet to see how it performs.

 

[counter_violent]

Fletcher Formula

No idea if this formula is any use...but the author claims it's great

Designed for blackjack....but at the end of the day...a hand either wins, loses or breaks even.
The formula however is a strict set of rules for a betting sequence and could probably be adapted for the coin toss experiment.

http://www.gamblersbookcase.com/FletcherFormula/dnFletchFormBJ11151rfTxq0/FletcherFormula-book.pdf

 

[dr_trick]

You could also use a monte carlo simulation Excel spreadhseet to see how it performs.

hi...

I did some simulations. I put results in a paper (http://ssrn.com/abstract=2530595 if you are interested).

cheers

 

[verial]

1. Martingale strategy.

2. The 1/4th strategy.

Good luck.

 

[dr_trick]

No idea if this formula is any use...but the author claims it's great

Designed for blackjack....but at the end of the day...a hand either wins, loses or breaks even.
The formula however is a strict set of rules for a betting sequence and could probably be adapted for the coin toss experiment.

http://www.gamblersbookcase.com/FletcherFormula/dnFletchFormBJ11151rfTxq0/FletcherFormula-book.pdf

hi..

i was not looking for a fixed set of rules for a betting sequence. what i was trying to study was, the effect of risk size on performance for any possible sequence of wins and losses.

the kelly criterion is often cited as the most optimal sizing strategy that maximises returns - this is related to that. there is a paper by Edward Thorp on using the Kelly Criterion for trading.

i guess what i am getting at is that in trading it is not enough to have a good signal generation method, risk/money management is also a critical component of any good trading strategy. if one risks too much money each trade then one runs the risk of being wiped out quickly if there is a run of losing trades. on the other hand if one is timid with the trade sizes then one does not fully gain on the upside on a run of good trades.

the balance is defining what too much or too little is. that is what the kelly criterion answers. i was looking at the kelly criterion for money management on a trading strategy i was working on before stumbling on this. i have only used coin flipping as an example to communicate the idea.

cheers

 

[dr_trick]

1. Martingale strategy.

2. The 1/4th strategy.

Good luck.

hi,

the martingale is a very risky strategy, i would not be encouraging anyone to try that.

could you elaborate on the second point please.

cheers

 

[bredin]

Any system where the lotsize (bet) increases after a win and decreases after a loss, with a 50% win rate, will result in a negative equity curve in the long run.
There will always be runs within that are positive, but being random there is no way to predict their location.

This is the simplest example i can place with a zero expectancy game of two coin flips either gaining or losing 10% per flip. for the same of math we'll start at $100
W +10 -> 110
L -11 -> 99
-or-
L -10 -> 90
W +9 -> 99

You'll need to take a look at geometric averaging for a more detailed proof.

B.

edit: I would guess that such a system as you propose would tend to a value approaching 1005 in the long run since the first bet is 5, rather than 10.

 

[dr_trick]

Any system where the lotsize (bet) increases after a win and decreases after a loss, with a 50% win rate, will result in a negative equity curve in the long run.
There will always be runs within that are positive, but being random there is no way to predict their location.

This is the simplest example i can place with a zero expectancy game of two coin flips either gaining or losing 10% per flip. for the same of math we'll start at $100
W +10 -> 110
L -11 -> 99
-or-
L -10 -> 90
W +9 -> 99

You'll need to take a look at geometric averaging for a more detailed proof.

B.

edit: I would guess that such a system as you propose would tend to a value approaching 1005 in the long run since the first bet is 5, rather than 10.

hi,

thanks for your reply. I was going to write a proof but I haven't gotten round to it yet.

the key feature of the approach is the level. i found that adjusting the bet sizes too frequently was not optimal. the level means that one has consistent bet sizes either side of it irrespective of whether one is winning or losing.

thanks again for your comments.

cheers



PS: the first bet also makes a huge difference.