Combining multiple systems

mind,

I haven't fully proved it to myself yet. Ran into some snags on setting up the simulation earlier.

But, it's looking like everything he has been saying is true. I'm one of the biggest skeptics here, so for whatever that is worth, you might want to continue looking at it (at least for a mental exercise). Even some of the earlier references to parrondo's paradox that I said were practically impossible to work out in reality, seem to work magically under maestro's assumptions.

It's almost too good to be true so far, so I am looking for holes. All I can say is read back through the thread carefully-- all the clues are there.

There are a few potential drawbacks however, one being continuous outcomes ( I already mentioned), and two being potential requirement for IID returns, which market's do not necessarily conform to.
 
I'd just like to comment something.

1. I got a weird news that MAESTRO is not for real. (I have no idea.... not a personal attack and shit. Just something I heard during a real life chat. (I'm in Japan right now but I used to live the US and still connected) That got me posting this.

2. All this BS about Parrondo and shit. You have to change the rules of the game. Coming back later with all the "conditional" stuff after you state that the game is 50/50 is unfair. My example was obvious that attempt.

My figure about close to Fib. comes from changing the rules. Obviously, trading is not about all-in / all-out. You have a choice of holding, reducing, and etc..

3. So what's wrong with having smarter friends?

4. Back "1." if MAESTRO is offended by my info. (or my casual sources are wrong), it should get him post something more objective to prove his status, rather than keeping all his posts all subjective and trivial. If he ever does come out and proves some viable point, it's good for you... MAESTRO... stop running... we all know you're not sick. BWAHAHAHAHAHAHAHAHAHA Everyone... Stop being nice to a trivial writer.

It's pathetic.

5. dtrader... You're BS. Stop posting from your 1 bedroom apartment. You need help getting a job??? Problems with testing??? What the hell are you testing? Pure assumption. Do you even know what testing is all about? What the hell are you testing? Running assumptious codes and concluding it makes money is useless.

You sound like a dumb quant using matlab.

6. mind (and his other alias)... You're always cool.

7. Rest of ET. Stop envying institutions. It's pitiful. There's a reason we earn more money than you. We take the extra step to earn more money. Trading is about making money. Stop being all jealous.

Peace out and back to real life.
 
Anybody make any progress with this concept? (Monty Hall problem applied to trading). I'm really doubting that there is any applicability but I'm having fun trying to figure it out.
 
I'm surprised people are still kicking around Parrando and Monty Hall. To be honest I don't see any potential for market application.

On Sharpe : I don't think optimizing for sharpe on individual strategies matters as much as many people think. Sharpe is just a measure of reward per unit risk, and while one might say a strategy with a sharpe of 3.0 is superior to one with 1.0, there might be all sorts of reasons why you might wish to split your capital on both instead of just putting everything on the 3.0

Aside from the obvious issue of correlation, there are also issues of capital usage, expected longevity and all sorts of other things you might care about.

I think at a portfolio level, sharpe is a good yardstick to measure things by. But the individual components might be chosen for all kinds of different reasons.

As for rules that adjust individual strategies to make them perform better together as a portfolio on a dynamic basis ..... one can always find ways to tease out rules which improve overall performance by underweighting or overweighting strategies based on how they interact.

But I tend to be leary of such meta rules unless I can understand the mechanism.
 
student

i dare say sharpe is more than "reward per unit risk" in
the sense that it is in line with my personal utility function.
and unlike other measures of risk, like draw downs, it is
more robust since there are more observations. there is
only one figure that even better matches my utility function,
and that is sortino. since i have no problem with upside
volatility. but i am a natural sceptic and i tend to believe
that sortino is easily flawed by few great outlying trades.
and i'd rather be on the save side.

the good thing about both measures is the inherent
time component, which is completely absent in profit
factor, hit ratio and payoff.

nevertheless your point is absolutely valid. within my team
we always search for the highest additional contribution
to our portfolio sharpe. and we do not care if that contribution
is by itself high or low in sharpe. a low sharpe can add
more to the overall if it is uncorrelated or even negatively
corred to the rest of the portfolio.
 
I think you have to deal with risk holistically and avoid single number measures.

Sharpe is a good tool for evaluating one strategy over another under certain conditions. Kurtosis hurts the usefulness of the sharpe ratio for instance as is well known.

Another issue is dependency of observations or clustering.

Take a return string [-1,+1,+2,-2,0] with 5 terms.

The mean and variance will be invariant to the order of the terms in the string so the sharpe will be consistent across permutations of ordering.

However, the experienced drawdowns are obviously dependent on the permutations of ordering.

Sharpe doesn't capture this - the issue of dependency between observations or tendency to cluster.

And there are other examples, but the key thing is to understand that all tools have strengths and weaknesses.

In my mind the utility function of the portfolio manager is a completely separate topic.

It's essentially about your emotional responses and what is the appropriate "sleeping" level of risk of ruin you are willing to tolerate.

If measures like sharpe are the instrument dials, utility functions are about where the red zone on the dial starts.
 
Quote from TheStudent:

I think you have to deal with risk holistically and avoid single number measures.

Sharpe is a good tool for evaluating one strategy over another under certain conditions. Kurtosis hurts the usefulness of the sharpe ratio for instance as is well known.

Another issue is dependency of observations or clustering.

Take a return string [-1,+1,+2,-2,0] with 5 terms.

The mean and variance will be invariant to the order of the terms in the string so the sharpe will be consistent across permutations of ordering.

However, the experienced drawdowns are obviously dependent on the permutations of ordering.

Sharpe doesn't capture this - the issue of dependency between observations or tendency to cluster.

And there are other examples, but the key thing is to understand that all tools have strengths and weaknesses.

In my mind the utility function of the portfolio manager is a completely separate topic.

It's essentially about your emotional responses and what is the appropriate "sleeping" level of risk of ruin you are willing to tolerate.

If measures like sharpe are the instrument dials, utility functions are about where the red zone on the dial starts.
More...
in my experience - though there are theoretical weaknesses -
sharpe ratios pretty much reflect my taste in equity
curves. i cannot remember an equity curve that had
a high sharpe and felt like bad and vice versa. simply
does not happen. though i must admit that i take all
sharpes below 1 with big grain of salt. 0,75 is not
much better than 0,5. but above 1 things get in line.

i guess everybody has his/her favorite ways to look at
things. i have experienced that the better the trader
the more he respects sharpe as a valid concept. with
the occasional exception of course ... :).
 
You must log in or register to reply here.
Back
Top