risk control in all-in portfolio system

Discussions about Money Management and Risk Control.
Post Reply
Hubster
Contributor
Contributor
Posts: 3
Joined: Wed Jun 01, 2005 8:25 am

risk control in all-in portfolio system

Post by Hubster » Wed Jun 01, 2005 12:10 pm

Here is an issue i'm struggling with:

I developed a system that is always in the market, that is, each signal is a reversal signal. There also an algorithm that indicates how much to trade. It trades several markets at once, a portfolio, for example 10 futures markets at once.

However, how do i know if i am undertrading or overtrading, that is, trading too few or too many markets. The system is always in the market, thus the typical 1-2% (or other) of capital does not apply (you simply reverse the trade when the system says so).

Do I look at drawdowns?

In sum, how would one look at risk for an always-in-the-market system that trades a portfolio of markets?

Thanks. Hope the above is clear.

damian
Roundtable Knight
Roundtable Knight
Posts: 815
Joined: Tue Apr 15, 2003 8:43 pm
Location: dusseldorf

Re: risk control in all-in portfolio system

Post by damian » Wed Jun 01, 2005 3:13 pm

In what context do you refer to "the typical 1 - 2%". I thought I knew but I am not so sure as your question was :

"However, how do i know if i am undertrading or overtrading, that is, trading too few or too many markets".

Do you mean number of markets (C, CL, SF......)
or number of contracts (buy 1, buy 5, buy 100.....)

TrendMonkey
Roundtable Knight
Roundtable Knight
Posts: 154
Joined: Fri Apr 22, 2005 9:14 pm
Location: Vancouver, Canada

Post by TrendMonkey » Wed Jun 01, 2005 4:18 pm

Wouldn't you just want the return-to-drawdown ratio, aka. MAR, with some sort of minimum CAGR (to weed out the case of excellent MAR's on pathetic CAGR's of, like, 3%).

For what its worth, and to no great surprise, testing with VT I have found enormous differences depending what markets (commodities) you pick. Right now my philosophy is that a good system should work reasonably well across all commodities, but I have been reconsidering the wisdom of that approach and want to do some more research.

Hubster
Contributor
Contributor
Posts: 3
Joined: Wed Jun 01, 2005 8:25 am

Post by Hubster » Wed Jun 01, 2005 5:12 pm

damian,

By typical 1-2% i was referring to how it is often said that you shoudn't risk more than 1-2% on any trade. Such a rule doesn't seem applicable to an all-in system, since in such a system you don't abandon a trade once you lose 1-2%, but rather you take a loss until you get a reversal signal. Thus, your loss may be more than 1-2% (could be less of course).

As for markets v. contracts, i guess they are related. You can overtrade by trading too many markets or too many contracts. However, I like my position size (# contracts) algorithm, so given that algorithm, I was wondering how I might measure the optimal # of markets to trade.

TrendMonkey,

Yes, my first thought was also in terms of drawdown. But I have seen at leat one all-in system described as being designed to have a 1% chance of having a drawdown of more than 20% in any one month. So, I guess i was looking for ways to better "quantify" the risk associated with adding (or removing) a market to the portfolio.

At least so far, I agree with working across all commodities approach. However, this all-in system also works much better for some commodities, even better with position sizing, and drastically better in a portfolio. Moreover, at least so far, I grab the 10 markets that seem to trend best (at least by eyeball standards), design with them, and then later do correlation studies and test with other markets.

Take care and thanks.

Hubster
Contributor
Contributor
Posts: 3
Joined: Wed Jun 01, 2005 8:25 am

Post by Hubster » Wed Jun 01, 2005 5:22 pm

And TrendMonkey you led me into my 2nd question:

Has anybody ever researched the idea of trading a system across a portfolio of commodities, but adding or removing markets (e.g., corn) based on the trendiness of the particular market. This would assume you could somehow measure/quantify how well (or badly) a particular market trends, and that the trendiness of a market may change over time.

It was simply an idea in my head.

It goes somewhat against the notion that a system should be designed to trade reasonably well across all markets.

damian
Roundtable Knight
Roundtable Knight
Posts: 815
Joined: Tue Apr 15, 2003 8:43 pm
Location: dusseldorf

Post by damian » Thu Jun 02, 2005 9:46 am

Hubster wrote:damian,

By typical 1-2% i was referring to how it is often said that you shoudn't risk more than 1-2% on any trade. Such a rule doesn't seem applicable to an all-in system, since in such a system you don't abandon a trade once you lose 1-2%, but rather you take a loss until you get a reversal signal. Thus, your loss may be more than 1-2% (could be less of course).

As for markets v. contracts, i guess they are related. .....
Clearer now. I only asked as teh original wording said "too few or too many markets" as opposed to contracts.

If you have a pure reversal system and you do not know the reversal point (or an estimate) inadvance.... which is quite difficult: then yes, youa re right, your ultimate loss may be greater than the desired x% maximum that you set for yourself.

In TBB do a big test that generates lots of trades, go to the trade details tab and sort the trades in the bottom window by % win/loss (cant remember the exact name of the field, I am at work sans-TBB). It will show you the largest % loser. Now, you have to position size somehow even if you are operating in a reversal environment. BAck-off some setting in your position size algo until you achieve a max % loss that is than the maximum amount. Use that setting for real trading. If history repeats itself, you will not have a loss of greater than X%.

Sorry I can't be more helpful.
Last edited by damian on Thu Jun 02, 2005 11:22 am, edited 1 time in total.

TrendMonkey
Roundtable Knight
Roundtable Knight
Posts: 154
Joined: Fri Apr 22, 2005 9:14 pm
Location: Vancouver, Canada

Post by TrendMonkey » Thu Jun 02, 2005 10:01 am

Hubster,

I have been thinking that very thing a lot lately. Conventional wisdom seems to be that a good system should work across all time periods and all markets, but I am thinking that all time periods is far more important than all markets.

There are things about specific markets, eg. position limits, price limits, tendencies of governments to intervene, seasonality, correlation to other markets, renewableness of the commodity, etc. that would seem to make it foolish to paint, say, SF with the same brush as W.

I think ideally you would want to determine some rule by which to dynamically pick which markets you are going to trade, maybe by reviewing the prior 5 years results at the start of each year, or something like that. As opposed to statically deciding that C trends more than CL (or whatever) and never looking at CL after that.

Right now I am testing different UA contract pricing algorithms (ho boy, there's some fun in there) and intend to look at more cleverly picking markets next.

Tigerline
Full Member
Full Member
Posts: 10
Joined: Mon Jun 23, 2003 12:30 pm
Location: Cumbria, UK

Post by Tigerline » Fri Sep 30, 2005 11:11 am

I have never understood the wisdom of trading a set number of futures or stocks exclusively. I have always thought that it would be better to pick the things that are trending whether it is a stock, commodity or whatever. I look through several hundred charts every week (surprisingly this only takes a few minutes) and choose a few to monitor more closely. I know which types of trends my system works best on and I trade those.

damian
Roundtable Knight
Roundtable Knight
Posts: 815
Joined: Tue Apr 15, 2003 8:43 pm
Location: dusseldorf

Post by damian » Fri Sep 30, 2005 11:26 am

As a long term trend follower, I would love to pick the things that are trending whether it is a stock, commodity or whatever.... that way I would never lose money. Simple. ;)

Tigerline
Full Member
Full Member
Posts: 10
Joined: Mon Jun 23, 2003 12:30 pm
Location: Cumbria, UK

Post by Tigerline » Fri Sep 30, 2005 12:33 pm

hi Damian

I lose plenty of money thanks :D

I know my comment sounds a bit simple. And it is. But it seems to work.
If it stops trending, then so be it. What usually happens is I get whipsawed a few times, then get a few open contracts and the chance of some profit.

I've just been reading your posts damian. How are things going this year so far?

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Reversal System

Post by AFJ Garner » Fri Sep 30, 2005 12:55 pm

Hubster, you can still accurately manage risk and position size on a reversal system. You can still have and "hold" stops so long as you have a mechanism to get back into the market if you are stopped out. Take an N Day breakout: 90 day high to enter, 90 day lo to exit. You can attach a 2 ATR (or 3 or whatever) stop. If you get stopped out, your basic system gets you back into the market if a new hi/lo is set. OK, you are not in the market 100% of the time, but risk control is more important and position sizing becomes very hit or miss unless you do this.

Somewhere on this forum Sluggo posted a dual MA system with have and hold stops. If you are stopped out, the re-entry trigger is that price must move back above or below both MAs.

Excuse me if I have misunderstood your problem
.

Roscoe
Roundtable Knight
Roundtable Knight
Posts: 250
Joined: Sat Jan 24, 2004 2:06 am
Location: Houston TX

Post by Roscoe » Fri Sep 30, 2005 6:50 pm

Hi Hubster.
Hubster wrote:Has anybody ever researched the idea of trading a system across a portfolio of commodities, but adding or removing markets (e.g., corn) based on the trendiness of the particular market.
This is the subject of ongoing research here. The conceptual key is to define the probability of success for each trade and use that probability to rank the MarkeSystems in the portfolio then assign risk capital based upon that ranking.

One thing that has become appearant is that basing the probability measure on trade-by-trade results produces relatively poor results compared to basing the probabilty measure on bar-by-bar dynamics.
Hubster wrote:It goes somewhat against the notion that a system should be designed to trade reasonably well across all markets.
My approach to this is that yes, a 'good' system should perform well across many markets but at any point in time some markets will have a higher probability of performing well than others.

Post Reply