Acrary is a genius!

Status
Not open for further replies.
I can't understand why diversification across uncorrelated strategies should not work.

Suppose you have a rigged coin. 60% chance of heads and 40% chance of tails. Let's define "heads" as profit, and "tails" as loss. So by tossing this coin sooner or later you will be in profit (more heads than tails). It's just a matter of time (number of tries). The probability of being in profit after 20 tries is bigger than after 10 tries, right?

Coin tosses are independent. Let's say we toss a coin 20 times a day. Take the first 10 and label them "strategy A", then the second 10 and label them "strategy B". The results of these two strategies will be uncorrelated, i.e. the strategies are uncorrelated. Now if you played only one strategy (A or B), you would get to toss a coin 10 times a day instead of 20, and the probability of you ending the day in profit will be lower.

So how did that guy proved this does not work?
 
countless examples can be given on this risk/reward and smoothness issue.
for instance think that you are betting on a rigged coin flip, chances are 1/3 that you loose all your money and 2/3 for doubling it. if you bet all your money on one coin flip, the chances of facing losing all your money is 1/3 . if you diverse your money on betting 4 coin flips, this chance is reduced to 1/81. if this what you aim to do, then there is no problem, your strategy will work.

there always can be unexpected events that can affect our investments, and by dividing our money on several investments we can reduce effects of this unexpected events. since we can't know everything, that our investments are affected by, in my opinion what greaterreturn has suggested is a wise thing to do, especially in times of uncertainties.
 
Quote from MAESTRO:

... The risk/reward ratio of the portfolio, therefore, is highly dependant on the correlation level observed in the set of the holding securities bur it does not depend at all on the level of correlation between the strategies implemented in this portfolio. I hope it clear things a bit.
More...
actually it is the opposite with me, it adds to the unclear
picture of what you are trying to bring across.

if i have a trendFOLLOWING strategy in the sp500 and
a trendFADER in the DowJones and we have a sudden
move in both of them, then one strategy triggers a long
and the other a short. if the two indices move on in the
same direction i win with one, i lose with the other. so
the ONLY thing i care about is strategy correlation. the
underlying market corr does not tell anything, unless we
are talking only about very similar strategies in the
first place.
 
Quote from man:

actually it is the opposite with me, it adds to the unclear
picture of what you are trying to bring across.

if i have a trendFOLLOWING strategy in the sp500 and
a trendFADER in the DowJones and we have a sudden
move in both of them, then one strategy triggers a long
and the other a short. if the two indices move on in the
same direction i win with one, i lose with the other. so
the ONLY thing i care about is strategy correlation. the
underlying market corr does not tell anything, unless we
are talking only about very similar strategies in the
first place.
More...


Yes, this thread was originally started to discuss Alan's ideas. With this in mind, when people are talking about correlation, it's correlation among results of strategies, that enter long and short trades at different times, implement stop losses and other trade management techniques. Which is a lot different than correlations among price data series of different markets. But I would probably be wrong to assume that Maestro does not understand this.
 
Quote from Indrionas:

Yes, this thread was originally started to discuss Alan's ideas. With this in mind, when people are talking about correlation, it's correlation among results of strategies, that enter long and short trades at different times, implement stop losses and other trade management techniques. Which is a lot different than correlations among price data series of different markets. But I would probably be wrong to assume that Maestro does not understand this.
More...

I wonder if things wandering so far off topic is a random or predictable phenomena?
 
Quote from man:

actually it is the opposite with me, it adds to the unclear
picture of what you are trying to bring across.

That is the point I was trying to make. If you have two (or more) correlated strategies (in your case your strategies are inversely correlated) then the combine result of running those strategies will reduce the exposure of your portfolio if your strategies (as you described) are taking opposite trades. However, the return on your portfolio will be also reduced as one strategy takes partially away the results from another strategy. This phenomenon as a whole will keep your risk/reward ratio constant. Notice, however how you used S&P500 index and Dow Jones index to illustrate your strategies. These indices are highly correlated and that is why your strategies are correlated (inversely) as well. Overall, if you run multiple trading strategies on a set of data the result of their performance will be another strategy that will exhibit risk/reward ratio that is less or equal to the risk/reward ratio of each of them separately.
More...
 
Quote from Indrionas:

Yes, this thread was originally started to discuss Alan's ideas. With this in mind, when people are talking about correlation, it's correlation among results of strategies, that enter long and short trades at different times, implement stop losses and other trade management techniques. Which is a lot different than correlations among price data series of different markets. But I would probably be wrong to assume that Maestro does not understand this.
More...

I definitely understand the difference. My point was related mostly to the multiple strategies scenario. However, correlation between the strategies usually stems out from the correlation of the data they are running on.
 
Quote from OddTrader:

I think we would have both correlated and non-correlated confusions here, whether positively or negatively associated, seriously! :D
More...

Exactly, all we need now is the definition of the positive expectancy of correlated and noncorrelated confusion .After that all is just a walk in the park .
 
maestro

(btw a great name. it is like talking to someone standing
on a podium ...)

i sense where you are coming from. yet i think in futures
the additional sharpe outweighs the cost of leverage by
far IMO. and i think this cost of additional leverage is
what your argument boils down to ...

one point comes to mind on the overall topic: if it is not
nonCorrelated strategies, what the hell is a systemdeveloper
looking for once he has something already trading?
 
Status
Not open for further replies.
Back
Top