Page 2 of 2
Posted: Wed Nov 15, 2006 9:11 am
Thatâ€™s not really it. The way I am doing things is quite a bit more involved than that (not that complicated is better!). However, thanks for taking the time to read it and comment.
My goal was simply to stimulate thinking on the subject and it seems as though it did that. I know personally I am always happy when something stimulates a new testing idea (or reminds me of an old one). Hopefully this will be that to some.
Posted: Wed Nov 15, 2006 9:18 am
Our shop employs a systematic approach to market selection. The only real constraint is market liquidity. Each market is categorized by sub-sector (e.g. short-term US interest rates) and exposure to each sub-sector is capped. Back-testing such an approach was not trivial since ALL markets (both active and inactive) needed to be examined.
Question: Is a systematic Manager that arbitrarily decides which markets to include/exclude in their portfolio truly systematic?
Dynamic Portfolio Selelction
Posted: Wed Nov 15, 2006 12:49 pm
I still think Dean is pointing to a good idea here. What I am interested in is designing some tests for it. One thought might be, if I have a large enough universe of products to trade (cars) and a limited number of parking spaces (due to small risk capital) I might want to have in my portfolio positions with tight stops (if based on volatility, let's say) versus ones with large stops. A dynamic approach could swap out existing positions for the tighter (better) positions. That might be one test.
There would be a whole lot of tests/ideas one could pursue with this approach. It seems to me to be more than just not adding a position if one is already concentrated in a sector (the Turtle approach) but actively managing existing open positions by changing them on the fly.
Dean showed an example of the results of one approach to this - I wonder if TB can do this as well? I don't yet own TB so I'm just asking. I am assuming it either does or could be programmed to do so.
Posted: Wed Nov 15, 2006 1:34 pm
Trading Blox can do anything you can think of here.
You can filter trades before entering orders. You can adjust positions after the fact based on risk or any other criteria you can come up with.
You could implement an algorithm that exited your worst positions based on time in market or profitability, or sector risk levels. You have access to all the historical trades, all the potential portfolio instruments, and all the open positions when making these determinations.
The way that Trading Blox handles simulation matches with what you can do in real life so the limitations are essentially the same. If you can come up with an algorithm you want to test I am sure that Trading Blox can test it.
- Forum Mgmnt
Re: Dynamic Portfolio Selelction
Posted: Wed Nov 15, 2006 2:24 pm
LeapFrog wrote: ... approach could swap out existing positions for the tighter (better) positions. ... actively managing existing open positions by changing them on the fly.
Leapfrog, that idea is presented in Mike Covel's Turtle course, what he calls Switches. However I don't think Dean is doing Switches in the trading system he showed on his "Single Biggest Mistake Investors Make" video; there just aren't enough trades. See image file framegrabbed from the video and zoomed, below. This kind of rejection rate points more toward the earlier remark "big portfolio and rules that skip lots of trades".
Posted: Tue Nov 21, 2006 3:13 pm
There are over 100 tradable commodity markets worldwide
Small accounts are able to efficiently trade markets that would be far too illiquid for large accounts
(Dean Hoffman offers) a trading program that monitors and trades over 70 diversified commodity markets
Why? It's not because of illiquidity; statement 2 takes care of that.
Not sure about Dean's situation, but I do take issue with "small accounts" trading markets far too illiquid for large accounts in one aspect -- you may be able to take positions, execute orders, etc., with a small account without moving the market, etc.. All well and good. However, what this concept doesn't take into account is that these markets, at times, may be dramatically impacted by a big-a__ trader or traders such that you are destroyed trying to get out. In other words, if a big trader tries to get in or out of your small market, his impact may blow your intended stop price out by a large margin.
Hence expecting large outliers in stop order fills, either from large overnight gaps or intraday movement, is something you have to account for as market volume, liquidity and participation go down. You may have a small account and be "able" to trade these markets, but that fact doesn't prevent a big swinging **** from bashing that market around like a ping pong ball. There's two sides to the story of market liquidity.
Posted: Tue Nov 21, 2006 4:29 pm
Chuck B wrote:
You may have a small account and be "able" to trade these markets, but that fact doesn't prevent a big swinging **** from bashing that market around like a ping pong ball. There's two sides to the story of market liquidity
Yes... Best to avoid big swinging ****
Now that's funny
Using MOO orders are a much better option in the thin stuff than a stop. Typically the open is a time of higher two way liquidity vs an intraday pressure point. Not immune, but safer.
Posted: Tue Nov 21, 2006 9:40 pm
Below is a post from Old European from 2004. It describes a method of portfolio optimization that, at first, set off all the curve fitting warning bells
! But, in fact, it is a very good approach.
Joined: 20 Apr 2004
Location: Old Europe
Posted: Sat Dec 18, 2004 1:59 am Post subject:
It is definitively possible to come up with single systems with a MAR ratio of 2 or higher without optimizing too aggressively.
The procedure I use is the following. I first define what I call a 'generic market universe'. It is simply the universe that consists of all the futures markets that meet a number of basic selection criteria (like sufficient liquidity). It contains about 50 markets in my case. Then I choose a simple system and look for a set of parameters that lead to stable results (flat area in parameter space). It isn't too difficult to find robust systems for such generic market universe and tested over 25+ years (with reasonable assumptions for commissions and slippage) with a MAR of say 1.25 or so. It is then very easy to increase the MAR dramatically (even without touching the parameters) by cherry picking the best markets. This is however something that goes way too far in my opinion. Instead of choosing the best performing markets I just throw out the very worst markets (a small number of markets). In order to figure out whether a futures market is good or bad, I compare the test results on the generic universe with and without this specific market (and thus automatically take correlations with the other markets in the portfolio into account). Even by prudently optimizing the investment universe in this way (and still using the same parameters for all markets), it usually turns out that the MAR ratio increases easily from say 1.25 to 2+.
I don't think there is anything wrong with optimization (why would you after all prefer parameter values that lead to inferior historic test results over better parameter values?) if you don't do it too brutally and if you don't count on repeating the optimized results in real trading. To be on the safe side I have set my personal real life performance target (MAR) as only half the MAR of the historic tests.
Never try to catch a falling knife!
I had been working with the concept that one constructs a balanced portfolio of futures markets based on diversity and low correllation, and that's it. Don't fiddle with it any more.
After testing his idea, and getting similar back-tested results, I now consider portfolio optimization to be another method I will use to achieve my target results. He cautions as to how taking this too far could turn into curve fitting.
This approach is different from Dean's. It's not dynamic. But, if done manually, say on an annual basis, it could be.
Anyway, as a novice with TB Professional, this seemed a good approach to take.
Posted: Wed Nov 22, 2006 5:37 am
By far the most problematic characteristic of LTTF systems is their intrinsic instability. And by far the most challenging parameter is the composition of the portfolio.
Since writing the above mentioned 2004 post I have continued to further research portfolio selection, but I failed to come up with a better method. Luckily I can live with it. The technique is very pragmatic, tries to avoid over-optimization and looks at portfolio selection and market correlation through the eyes of the system I want to trade.
I am however not inclined to (prudently) re-optimize my portfolio on a yearly basis. The only change I'm prepared to make is to add in again a few markets that I had thrown out in 2004. I prefer to go the way of less instead of more curve-fitting.