Page 1 of 1

portfolio stepping

Posted: Sat Mar 01, 2008 1:59 am
by ByllD
As you know, sometimes adding an additional contract to a portfolio may smooth the equity line even if it does not add to total return. I have been "stepping through" the combinations of contracts to trade as a portfolio. Is there a way to do this automatically, rather than changing the portfolio manually before each run? If I am looking at 10 possible contracts I'd like to compare results for any two, any three, any four, etc. Is there, in the most basic package, a way to do this automatically?


Posted: Sat Mar 01, 2008 7:29 am
by sluggo
I don't know whether that's available in the Blox Turtle Edition or not.

It is certainly possible in the Blox Builder Edition. You would just edit the excellent Blok supplied by Nickmar + Jake Carriker viewtopic.php?p=22321&highlight=purpose ... ndom#22321 and replace "random" by "systematic stepping".

However, in my opinion, this probably isn't a good idea. Optimizing the system by cherrypicking markets for the portfolio, smells feels and tastes like overfitting to me. Particularly if you're trying to build a smallish portfolio with two or fewer markets in each commodity sector.

AFJ Garner has presented many eloquent arguments along these lines; I urge you to search out his posts here on the Roundtable and read them thoughtfully.

The alternative to small cherrypicked portfolios is, naturally, large and inclusive portfolios. At first thought, it might seem that the larger your portfolio, the larger your account must be. But this is only true if you assume that you take every trade; you enter every time you get an entry signal.

If you've got any edition of Blox software, you know otherwise. You've got at least one system in your hands right now, that does not take every trade: the Turtle System. As you know from reading the manual and experimenting with Turtle, it has limits on how many positions you can take simultaneously (see image below). Once you're above these limits, Turtle simply ignores new entry signals. Skips em. So you can add 200 more markets to your portfolio and it doesn't increase your maximum exposure and maximum risk at all: the Turtle rules limit maximum risk.

Other systems use slightly different mechanisms to achieve the same goal: limiting maximum exposure without limiting portfolio size. One example that you can use with the Blox Pro and/or Blox Builder edition, is "Thermal BBBO" (here). Another, that you can think about and then program yourself, is the "Parking Lot" concept mentioned here by Dean Hoffmann and sketched out in preliminary form by Tim Arnold (here).

Using ideas like these, it becomes quite possible to have a portfolio of dozens and dozens of markets with a small account. You've greatly reduced the chances of accidentally overfitting or cherrypicking a portfolio, since your portfolio is now enormous. You don't have to agonize about choosing among Eurodollars, Banker's Acceptances, EURIBOR, Short Sterling, NZ Bank Bills, Fed Funds, or Euro Schatz. Choose them all! Give each of them a chance to present trades, and let your system decide which trades to accept and which trades to reject.

Key message: there's more than one way to make sure you only have a small number of simultaneous positions and hence a small total risk. Unimaginitive Method #1 is to have a small portfolio. Clever Method #2 is to have a large portfolio and a system whose rules explicitly control the number of simultaneous positions and hence control the total risk. AFJ Garner and I hold the opinion that Method #2 is superior. On the other hand, it takes a difference of opinion to make a market ....

Many markets

Posted: Sat Mar 01, 2008 10:49 am
by ByllD
Wow. There are folks out there charging money for less well-thought-out and less well-written reasoning than your reply here. Thanks. This Forum is worth the price of admission, without the software.

Over the years I have learned at my cost that what you say is true. My question was not as well presented as it might have been; I just wanted to make my need known. I have used CSI's correlation studies and tried to construct a portfolio with wide exposure. My fundamental algorithm is not very profitable in some markets and I don't change the parameters due to concerns over fitting. My desire to step a combination in a portfolio is not to restrict the total, but to find markets that, alone, may be uninteresting, but in the portfolio may provide additional negatively correlated diversification.

Back-testing is terrific. What it cannot replicate is time. One looks at the drawdown in a ten- or twenty-year backtest and is beguiled by the final result. Living through the day-to-day drip of drawdown is quite different. My goal, even at the cost of a somewhat lower return, is a relatively smooth equity line. I've been doing this for my income for over 15 years but I am now 60 and much less brave. (The "bravery" starting point was not high.) Discipline has always gotten me through, but a smooth line would be a help.

Thanks again, Sluggo.