Portfolio size

Discussions about the testing and simulation of mechanical trading systems using historical data and other methods. Trading Blox Customers should post Trading Blox specific questions in the Customer Support forum.
Post Reply
tigertaco
Full Member
Full Member
Posts: 23
Joined: Fri Feb 27, 2009 1:50 am

Portfolio size

Post by tigertaco »

Let's say your system enters at the new N-high/low and exits after K bars (trend following system that is) and has no other exits. What 's an optimal portfolio size for such a system? Here is why I'm asking this.

Consider the situation in which you are fully invested and system gives an entry signal in some market. You didn't take it, the market kept trending, system kept giving signals and some M bars later (M<K) you can actually take the signal since you are not fully invested anymore. But, you are entering at much worse price. The problem here is that signals form a sequence (in a trending market) and other markets in portfolio interfered with your ability to use the earliest signal.

If you think that's a problem there are 2 solutions. First is to trade only a few markets, so that you can take first signal in the signal sequence in any market at any time. This is a simple solution, but I don't think it's bad. Second solution is to rigorously keep track of signal sequences, so each signal comes with a number (1,2,3,... ); the rule can be that the first signal in a sequence is the first N-high after a portfolio filter condition is satisfied, and each subsequent signal is N-high above previous signal bar. the sequence terminates when portfolio filter switches (same for shorting signals). all N-highs between consecutively-numbered signals receive the sequence number of lower-numbered bar. and with this in place you only enter on a signal if its sequence number is 1.

it seems that some solution to the above problem should be employed otherwise time-based exit system can get out of sync with a market by entering on late signals and lose a lot of money.
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

It's pretty common for trendfollowing futures traders to trade >100 instruments simultaneously. Search this site for the string "Transtrend", the poster-child of this trading style. They trade >300 instruments simultaneously. Is that "optimal"? They think so.

But futures are delightful beasts that allow huge leverage with no borrowing costs. If a trader wanted to trade other kinds of instruments having different max leverage and different cost-of-leverage, the optimum might be very different.
alp
Roundtable Knight
Roundtable Knight
Posts: 309
Joined: Mon Aug 27, 2007 8:09 pm

Re: Portfolio size

Post by alp »

tigertaco wrote:First is to trade only a few markets, so that you can take first signal in the signal sequence in any market at any time. This is a simple solution, but I don't think it's bad. Second solution is...
I wonder: if the system cannot take all signals, why would it trade a bigger portfolio in the first place?
tigertaco
Full Member
Full Member
Posts: 23
Joined: Fri Feb 27, 2009 1:50 am

Re: Portfolio size

Post by tigertaco »

I wonder: if the system cannot take all signals, why would it trade a bigger portfolio in the first place?
.

Well, it's probably better to have more signals than you can handle, then less. In fact, maybe it is better to have very large portfolio but keep track of signal sequence and take only the earliest signal. These are just guesses.

By the way, could someone please tell me what is a reasonable slippage+commissions combination for backtesting futures?
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

Commission is easy; it's the number that you negotiate with your broker. Depending on the service level that you select, and the number of contracts you promise to trade per month, and your skill as a negotiator, it could range from high-single-digit to high-double-digit dollars per contract per round trip trade. As you would expect, commissions on DIY execution methods are cheaper than paying a broker to execute your orders. In effect, you pay yourself the difference ... AND you transfer the "execution risk" (cost of mistakes) from the broker, to yourself.

Slippage depends on the markets you trade, and the types of orders you use (MOO/MOC/stop/limit/stop-limit/SCO), and the size of your orders relative to the liquidity present in the markets at the instant you execute, and the skill of your order execution person(s). Ted Annemann gave some excellent advice (HERE) -- don't assume slippage, measure it. You may also enjoy scrolling down that thread and looking at his equity curve.

Blox software implements a volatility based slippage calculation described in the manual (which anybody can download) and summarized in the image file below. The buy trade example shown assumes the user has set the adjustable parameter Slippage% to 25%, but of course you can set any Slippage% value you want ... or, if you wish, you can step Slippage% from 5% to 100% in steps of 2.5% and observe the impact of slippage upon system results.
Attachments
from Trading Blox User's Guide
from Trading Blox User's Guide
slippage.png (70.71 KiB) Viewed 5750 times
alp
Roundtable Knight
Roundtable Knight
Posts: 309
Joined: Mon Aug 27, 2007 8:09 pm

Re: Portfolio size

Post by alp »

tigertaco wrote:Well, it's probably better to have more signals than you can handle, then less. In fact, maybe it is better to have very large portfolio but keep track of signal sequence and take only the earliest signal. These are just guesses.
So, as a matter of fact, you're filtering out trade signals based upon the guess that the earliest signal has the best predictive value. In short it's a filter.

So, perhaps, you could trade a large portfolio and apply some or several rules to rank instruments and pick up trades which put the odds in your favour. I just wonder which rules are best.
tigertaco
Full Member
Full Member
Posts: 23
Joined: Fri Feb 27, 2009 1:50 am

Post by tigertaco »

Thanks, alp. I have a question, though. Can a slippage really be 5%? That seems unbelievable to me.

My current trading is 99.9% is in stocks on N-tick charts in Tradestation. I'm used to fighting for every tick (only exception is when it feels like a monster trend is coming then good fill could actually be a negative). Anyway, so when I recently decided to venture into system trading I naturally started with tick charts and set a slippage to .02 (liquid nasdaq) or .025 (liquid nyse) for stocks trading at prices > $50, and such assumption so far seems realistic in real-time trading although I don't have a statistically significant sample yet. Are futures that different? Then, could you estimate an intraday volatility and wait until it drops below average? I just don't see how can one overcome such slippages. I consider myself a novice as far as system trading goes, so maybe I'm missing something.
Roscoe
Roundtable Knight
Roundtable Knight
Posts: 250
Joined: Sat Jan 24, 2004 2:06 am
Location: Houston TX

Re: Portfolio size

Post by Roscoe »

tigertaco wrote:What 's an optimal portfolio size for such a system?
I seem to recall Ralph Vince giving the math to find the optimal portfolio size, and the optimal number was actually quite small. Try his most recent book for detail. In my testing with starting account size of $1M I came up at 12 MarketSystems as being the most useful: the peak of the curve.
alp
Roundtable Knight
Roundtable Knight
Posts: 309
Joined: Mon Aug 27, 2007 8:09 pm

Post by alp »

tigertaco wrote:Then, could you estimate an intraday volatility and wait until it drops below average? I just don't see how can one overcome such slippages. I consider myself a novice as far as system trading goes, so maybe I'm missing something.
I am also a novice...

Perhaps, you could implement a filter based on intraday volatility and see how it works in your backtests.

With regard to filters, say, if we were to choose which criteria upon which to rank instruments and pick up trades from a very large portfolio, which would we implement in order of importance, so as to put the odds in our favour: 1) a (long term) trend filter; 2) momentum (ROC, for instance); 3) correlation limits; 4) liquidity; 5) volatility; 6)...; 7)...; ?
Post Reply