Multiple Strategy vs. Single Strat

syswizard

syswizard

I have a trading friend who wishes to eventually become a hedge fund manager using systematic trading...mostly with either index futures or equity index options.
He has this idea of using a blend of statistical indicators (TRIN, A/D, % of 52 week highs, put/call ratio, etc.) to develop one GRAND indicator...a composite indicator.
He then wants to use that composite indicator to develop a system with buy/sell rules.

I've been trying to talk him out of doing extensive research on the use of those statistical indicators because they seem to have low correlation with the underlying index price. Researchers like Colby have already done the basic work to show that.
http://www.robertwcolby.com/RelativeAdvantage.html

I am trying to talk him into developing several systems (I suggested three) instead of one and that the concentration of research time should be spent on position-sizing and portfolio management methods. Also, I want the three systems to have separate underlying premises.....i.e. trend-following, reversion-to-mean, trend-reversal, etc. Thus, the trick would be allocate the funds to the specific system/strategy which is showing the best performance based on the current market conditions. In this manner, we could possibly avoid the "John Henry" problem of poor performance in non-trending markets.

Does anyone want to comment on my ideas here ?
 
Quote from syswizard:

Thus, the trick would be allocate the funds to the specific system/strategy which is showing the best performance ...
More...

Allocate more to the best, or less - if you trade them all concurrently?

Quote from syswizard:
Also, I want the three systems to have separate underlying premises...
...i.e. trend-following, reversion-to-mean, trend-reversal, etc. ...

... based on the current market conditions.
More...

?
 
Quote from syswizard:

I have a trading friend who wishes to eventually become a hedge fund manager using systematic trading...mostly with either index futures or equity index options.
He has this idea of using a blend of statistical indicators (TRIN, A/D, % of 52 week highs, put/call ratio, etc.) to develop one GRAND indicator...a composite indicator.
He then wants to use that composite indicator to develop a system with buy/sell rules.

I've been trying to talk him out of doing extensive research on the use of those statistical indicators because they seem to have low correlation with the underlying index price. Researchers like Colby have already done the basic work to show that.
http://www.robertwcolby.com/RelativeAdvantage.html

I am trying to talk him into developing several systems (I suggested three) instead of one and that the concentration of research time should be spent on position-sizing and portfolio management methods. Also, I want the three systems to have separate underlying premises.....i.e. trend-following, reversion-to-mean, trend-reversal, etc. Thus, the trick would be allocate the funds to the specific system/strategy which is showing the best performance based on the current market conditions. In this manner, we could possibly avoid the "John Henry" problem of poor performance in non-trending markets.

Does anyone want to comment on my ideas here ?
More...



The best current market conditions can only be determined after the period , not during.
Same remark for non-trending markets. By the time you know the market is not trending it might become trending again.
So in fact you want to use information that will be only available later on. So you need to be able to look in the future.

If it would be that easy to see if the market is trending or not, John Henry ( and others) wouldn't have such a drawdowns. They would also always use the best perfoming strategy. In other words: they would always make money.
 
Your friend will have difficulties obtaining and using historical data on options prices, even equity index options prices.

The problem with multiple systems is, they seldom do what you hope, which is to protect you against big adverse moves by having partially offsetting positions. In fact the only way I've found to guarantee that systems are anti-correlated, is to make some of them long-positions-only and others short-positions-only. Which gives off an odor that some people call "the stench of curve fitting." I disagree but I'm in the minority.

The standard fantasy is to have a short term system, a medium term system, and a long term system. Then you hope / believe that this arrangement provides non-correlated positions. Unfortunately, if the underlying tradeable goes into a prolonged unidirectional move, all three systems will eventually get aligned with that move (profitably so!) and now you've got 3 perfectly correlated (identical) positions, just exactly what you don't want. A price shock will kick you in the balls with 3X the force.

Test your system ensemble and see how well it handled the oil embargoes of 1973-1974, the crashes of 1987 and 1989, the Thai Baht crisis, the Iraqi invasion of Kuwait, the "shock and awe" bombing campaign of Gulf War II, and so forth. See how it reacts under stress. Then expect that Murphy's Law will produce something in the future, far worse than anything you've seen in the past.
 
Quote from syswizard:


Does anyone want to comment on my ideas here ?
More...

What's your objective? You first need to define that, and think carefully about it. Are you wanting to control risk, or maximize your gains? What's your emphasis on the balance of the two?

If your objective is risk control, then you really aren't helping much with systems that are all exposed to U.S. equities, unless you are balancing long with short.

And in my experience, even that is problematic, as I've very seldom found that hedging a good long strategy is worth the capital outlay. Yes, I do it, just out of need for some portfolio insurance, but don't count on it helping much in day to day management.

Ideally, IMO, you would want to focus on your best long strategy if you are in U.S. equities or indexes, which will hopefully be the one that produces maximum gains with minumum drawdown during difficult market conditions. The idea here is to party heavy when things are good, and accept some small defeats when conditions turn against you for the time being. Otherwise you'll find yourself trying to throw a lot of sh*t against the wall and hoping that it sticks once in a while.

I'm a very big believer in diversity in trading positions, but in applying multiple strategies it is a different story as they often just water each other down and draw on capital when you really need it.

But this is just my opinion.
 
Wow, guys....thanks...and that was FAST !
The best current market conditions can only be determined after the period , not during.
Same remark for non-trending markets. By the time you know the market is not trending it might become trending again.
More...
The best that can be done is to attempt to detect market conditions as witnessed by the performance of one of the systems, and then to adjust capital committment to that system ASSUMING the market will remain in the new condition.
I had intended for that determination to be done on a rolling 3 month basis.
What is your objective ?
More...
Good point. Initially, it would be to have a proven strategy or set of strategies that would provide a risk/reward measurement (Sharpe, sortino, etc) that would be competive with the major hedge funds.
I'm a very big believer in diversity in trading positions, but in applying multiple strategies it is a different story as they often just water each other down and draw on capital when you really need it.
More...
Hmmm....food for thought...and testing as well. I wonder, should I be thinking MULTIPLE MARKETS instead of MULTIPLE STRATEGIES ? (i.e. Interest Rate, Commodities, Equity Indices, etc ) If MULTIPLE MARKETS, then can a Composite Indicator be fashioned for each market ? Obviously, commodities have much different stats than the equity markets i.e. there's no "breadth" available, etc.
 
Quote from syswizard:

I wonder, should I be thinking MULTIPLE MARKETS instead of MULTIPLE STRATEGIES ?
More...


Ideally, yes. The trick is in getting that done effectively assuming you have somewhat limited resources for trading, and for support.

Initially, feel fortunate if you can nail one good strategy that turns consistent profits without getting you in trouble. Unless you are very heavily capitalized and have staff standing by, that alone will fill your days and nights with enough challenge.

Don't underestimate the hurdles to developing a solid strategy for any market. Multiple markets will increase those hurdles correspondingly.

But again, diversification in any form is a very noble goal, and will payoff assuming you aren't wasting resources.
 
System development is not as easy as it looks, nor is it impossible. But developing the holy grail of trading systems is impossible -- very few systems will work as your friend wants. If it was so easy, others would have already done it (maybe they have???).

Remember, no matter how smart your friend is, many many people much smarter than him have tried and failed to do it. Doesn't mean he won't be able to, but the odds are against it.

SSB
 
Amen Brother....and thanks for the post.
The original "concept" was to create this "super composite indicator" ?
But with "rules" that say 5% max in a position, ONE INDICATOR driving ONE SYSTEM in ONE MARKET says that 95% of our client's money will be in CASH.
This is not good.
 
You must log in or register to reply here.
Back
Top