Daily Speculations The Web Site of  Victor Niederhoffer & Laurel Kenner Dedicated to the scientific method, free markets, deflating ballyhoo, creating value, and laughter;  a forum for us to use our meager abilities to make the world of specinvestments a better place.

Home

The Trend Following Thread (November 2004-

11/29/04
Victor Niederhoffer on Trend Following

One of the difficulties of trend following is that the results are so variable that one can never gain statistical confidence in them. One or two good or bad trades is the difference between success and failure. In a book written by someone on the list, that purported to show that trend following works, if you took out the coffee trades, the trend following doesn't work. Because of the importance of one market and one or two trades to the results, it is so easy to come up with a system that works retrospectively. And to say that such a system worked out of sample would not be consistent with the 2004 results of the 3 heroes mentioned in the book, who last time I looked were down so many billions of dollars each in their trend following programs. But wasn't that guaranteed to happen.

11/29/04
Trend Following, by Richard Gula

I might suggest that the non-believers in trend existence find a copy of William Bretz's Juncture Recognition in the Stock Market...out of print probably, but likely on the shelves on a good public library... and Michael Zahorchak's The Art of Low Risk Investing...1972 originally and paperbacked in 1977,,,postulated the first "systematic" approach to trend exploitation...works best on volatile stocks..

The concept of group commonality is largely absent from most technical and quantitative works done since the 1970s...sadly...but it is a free and open game, and the fact that you think the forward pass is silly, dangerous and newfangled and won't last...is likely to relegate you to the the Woody Hays school of innovation and unsuccess. Been doing this stuff since 1975. No axe to grind with anyone...Most trend nihilists have been destroyed by trends at some point, unfortunately. The move in energy stocks is random? As I said, I have seen the Livermore model translated to a quantitative scoring system and it seems to add value... Lastly, required reading on this topic: Gerald Appel's several works on Trended vs Trading Range Models..simple in concept, fully back-tested data in them...

Would suggest that Appel and Zahorchak be digested before proclaiming there are not trends. The more recent world of asset management, where a single Fidelity mutual fund dwarfs all the emerging NY hedge funds combined...and the style boxed ERISA managers manage within their domains...give considerable insight into the roots of trends..which are always anticipating future fundamentals. The confirmation of SP500 index stocks, analyzed by sectors, by the trends of small cap stocks in the same sectors/industries, confirms the presence of the trends. The duration of the trends results from the concepts from Bretz.. the factors or CONDITIONS that encourage or even define the trends.
Most of these models work best when you REMOVE the price track from the second derivative model. Try the Zahorchak model without price plotted! And do it for all the stocks classified as semiconductors by Factset, for example. As tempting as it is to see the stock market as an elegant squash match with one of the Khan's, a tactical and psychological battle, the modern markets are much more like the NFL and the playoffs and the Super Bowl...not elegant, rude, dangerous, and played for keeps. When the stakes rise, the fatalities rise with them.

11/27/04
Peter R. Gardiner comments

>i might suggest that the non-believers in trend existence find a copy of William BRetz's Juncture Recognition in the Stock Market...out of print probably, but likely on the shelves on a good public library...

How about just showing us a reproducible test, with defined variables, and clear results so that we may see directly what the facts are.

For example, in SNP Futures, when t[close] > highest{high, last 10 days]

Buy close and hold for 1 day. Results since 1982: Avg return: 237/ 521 (45.48%) , mean return, = -0.19 pts, sd=5.93pts, se=.2597pts, z= -.73 . This result is completely random. Hold for 5 days? Fine: 159/299 (53.17%), mean return=-.26 pts, sd=15.38 pts, se=.8894, z= -0.29. Random. Everyone is entitled to an opinion, but not the facts. Please put the stats on the table and stop torturing us with ad hominem appeals.

> and the fact that you think the forward pass is silly, dangerous and newfangled and wont last...is likely to relegate you to the the Woody Hays school of innovation and unsuccess.

Sadly, when a forward pass is aimed toward the wrong end of the field it achieves a result entirely consistent with the results of trend following as described.

>Been doing this stuff since 1975.

That is quite a long time, although I am sure there are many more who have contributed their capital longer than 30 years.

>As I said, I have seen the Livermore model translated to a quantitative scoring system and it seems to add value...

Seeing is believing: please put the facts in front of us.

>Lastly, required reading on this topic: Gerald Appel's several works on Trended vs Trading Range Models..simple in concept, fully backtested data in them...

Would suggest that Appel and Zahorchak be digested before proclaiming there are not trends. The more recent world of asset management, where a single Fidelity mutual fund dwarfs all the emerging NY hedge funds combined...and the style boxed ERISA managers manage within thier domains...give considerable insight into the roots of trends..which are always anticipating future fundementals. THe confirmation of SP500 index stocks, analyzed by sectors, by the trends of small cap stocks in the same sectors/industries, confirms the presence of the trends. The duration of the trends results from the concepts from Bretz.. the factors or CONDITIONS that encourage or even define the trends.

Most of these models work best when you REMOVE the price track from the second derivative model. Try the Zahorchak model without price plotted! And do it for allthe stocks classified as semiconductors by Factset, for example.

Please do not bore us with this drivel. You are speaking to an audience which contains some of the most mathematically sophisticated people out there, and we rarely see anything so unintelligible - or where intelligible, unsubstantiated - as this. State your hypothesis, test it, and show us the results.

>As tempting as it is to see the stock market as an elegant squash match with one of the Khan's, a tactical and psychological battle, the modern markets are much more like the NFL and the playoffs and the Superbowl...not elegant, rude, dangerous, and played for keeps. When the stakes rise, the fatalaties rise with them.

When you can lay claim to 1/20 the achievements of our host - in any endeavor whatsoever - you may tell a hushed crowd of specs. Otherwise, before you take a shot at him, I suggest you arrange a match in a racquet sport with him, and then tell us all about it, and how little you learned about yourself, risk, position management, expectation, variation, stamina, courage, imagination, winning, and finally losing.

Then we will all see who knows what. Otherwise, tell us what it was like to play in the Super Bowl; we are always in the mood for fiction.

Here is the point: it is not about sharp knives at all. It is simply having enough respect for the audience to engage in the most minimal attempt to substantiate (often wild) claims. Everyone is in favor of new profitable ideas. But, how, without some foundation, can we judge for ourselves what is of benefit? Surely not just a hearty say so. It is the entire point to help educate each other, but how, pray, can this be done without making clear the idea, the implementation, and the result, if only in a back of the envelope way. Thousands, nay, hundreds of thousands of serious tests have been run on a multitude of "tend-following" systems by members of this list, and the results are extremely robust. Nevertheless, there may be something new. However, how we would know if not by the signature of a repeatable, quantifiable result? This has nothing to do with "set ways:" it is simply the syntax of meaningful exchange among serious people with minds open and fully functioning.

11/27/04
Michael Stallings agonizes:

Once again trend following has raised its head upon these List(s). And, once again it is a newcomer to these List(s) which has raised it. And, once again, the usual hue and cry for "stats on the table" has been uttered.

Why, for the love of the all mighty dollar, don't these List(s) listen with an open mind, and accept the fact that there are people who are invited to this table, who have been successful in what ever form of trading that they do, be it trend, chartist, B-Bands, or randomness? Is everyone so set in his ways, that he can't or won't admit, that there may be other successful ways to trade, and that it is each to his own, as he sees fit.

I really think that in the true spirit of these List(s), that the "sharp knives" be put away, and we try to expand our financial horizons by listening to what each other has to say.

11/27/04
Peter R. Gardiner responds:

Here is the point: it is not about sharp knives at all. It is simply having enough respect for the audience to engage in the most minimal attempt to substantiate (often wild) claims. Everyone is in favor of new profitable ideas. But, how, without some foundation, can we judge for ourselves what is of benefit? Surely not just a hearty say so. It is the entire point to help educate each other, but how, pray, can this be done without making clear the idea, the implementation, and the result, if only in a back of the envelope way. Thousands, nay, hundreds of thousands of serious tests have been run on a multitude of "tend-following" systems by members of this list, and the results are extremely robust. Nevertheless, there may be something new. However, how we would know if not by the signature of a repeatable, quantifiable result? This has nothing to do with "set ways:" it is simply the syntax of meaningful exchange among serious people with minds open and fully functioning.

Gula previously stated:

Conditions are anticipated by TREND CHANGES...rising oil prices (from $9 a barrel a few years ago to $50 plus..is it a trend or a condition?) Having worked inside of a number of very large (asset wise) organizations (Putnam, Batterymarch in its heyday, and Fidelity in its heyday) I found that commonality of trend gave credence to conditions. That is, analyzing a pool of 250 plus energy related stocks, 80% of which reversed trend in the fall of 2002, coincided with the CONDITION of rising oil prices. At the moment, the stocks and their trends are anticipating a renewal of the energy condition and its trend in the crude futures. quietly, the plurality of energy stocks, especially service and drilling cos., are making new highs through this entire week while the contract bounces of its lows, but remains below its highs.

Sector concentration has become the key to investment returns and most of the hedge fund managers lack the experience to understand that they get entirely too wound up in quantitative risk reduction in a business where taking risk is the name of the game. Victor was and still is one of the best in this regard. I have not read anything of COVEL other than what the PH website would offer up, one chapter that suggested his was the insiders insider in talking to the various great traders. I usually dismiss such

s quickly if they offer up that "I know the Hollywood stars better than you do.." kind of tone.

Trends are revealed through commonality studies no matter how simple the filter or model is. In fact, the  simpler the model is, the better in many cases, since that will encourage the observer to think through the relationships to the condition. Similar commonality existed in a 90 stock "silo" of semiconductor stocks in the fall of 2002. they reversed trend within several months of one another, leading to a 15 month rally. the condition was debated by the "good managers" who got it right...DRAM prices, market share, sales of boxes...or valuation attractiveness...INTC rallied from 13 to 35 and ARTI from 7 to 34....but the ALL rallied in concert. Getting commonality right is 70% of the investment battle, to my mind...

11/29/04
Mr. E responds:

Of course this idea of commonality is how the mutual funds increase their NAV's from year to year. Since the big mutual funds have become closet indexers, coincident with the loss of great risk takers at places like FIDO, so they take umbrage at convincing other mutual fund managers to adopt similar rules in masking their index games. As a reflexivist, like Soros, I watch how they set the rules and from time to time I drop a BIG ROCK that tests their theories of commonality and media mind numbness. We hedge funds can also test the rules and have some fun at the same time. After all to have a 50% compounded return since 1988 one has to be flexible about many things.

11/29/04
James Sogi adds to the Trend of Trend Following Talk

Whether at random or otherwise, some market prices appear to trend at times, and others trend at other times. Previously reviewed on this list, Lars Kestner, in Quantitative Trading, tested various markets for trending over time in which he has found good trendiness in some markets in recent times. Chris Cooper has done similar testing showing more trendiness in certain commodity markets than in equities. If one can time which markets are trending and get in with leverage, then you are a successful pro. The markets will trend for some underlying reason, so analysis of the fundamental conditions should help in addition to quantitative analysis. But assuming the amateur investor cannot time the trends, as does William Bernstein in Four Pillars of Investing, then an uncorrelated asset allocation model might allow the amateur investor to participate in a trending market, with at least a portion of the portfolio and can say at cocktail parties, "Oh yes I have been long long gold and oil on this big rally". Those portions of the allocation that do not trend, will suffer, but in the next cycle, it will be those asset classes that trend. The allocation is based on the investors risk profile.

Another aspect of trending is returns. At times, the asset values trend up, and hopefully never trend down. The measures of risk can be stated as variously as the draw down, the variation in asset value, the time underwater from asset highs, the length of draw downs to previous highs, largest loss per time frame. Another related aspect of trending is returns of different trading styles and time frames. Trendies suffered earlier this year, while rangers prospered. There are trends in conditions, or regime changes. Political trends, demographic trends. The key in all analyses of trend is to identify turning points. No trend lasts forever and will eventually end due to the eventual cycle change. Because whether a trend exists or not, unless one commits at the turning point or near the turn, very often a move will be missed. The analysis of turning points also allows capitalizing on the ranges as well as the trends. The issue of trend relates to the exit point or allocation of risk. The reason this discussion is attracting debate in this forum now is that we are at a turning point

11/29/04
Russell Sears responds:

Perhaps, I am torturing the data into a confession of being random... But using Mr. Cooper's 82 data points as a general distribution of trend following. I note 2 things that may fool many into assuming these results are not random. First, it seems the middle of the curve is thinner than a normal distribution. I count 22 occurrences within a + or - 1/2 a std dev and I would expect 31. So it seems the tails are "fatter" than a normal distribution. Second 3 positive positive data point are more than 2 std dev from the mean and only 1 is negative is less than 2 std. dev.

With such a distribution it is easy to see how, as the chair points out, "One or two good or bad trades is the difference between success and failure". After doing 10,000 random returns compounded 100 times based on historical returns (1 million random draws from the 82 data point) many returns seem highly abnormal...but both negative and positive. The best return 550% and worst -65%. But "abnormal" is different from "non-random". That people get rich using such schemes is to be expected. Yes, it has positive expectations but not much better than the market in general.

11/29/04
The Trend continues with Hany Saad:

Trend following is alive and well and wealth distribution will remain at 20 to 80. As Robert Bacon would say the public can never catch up to the form - or the game ceases. No worries in this case, the game is also alive, and the market upkeep is paid for in full.

Reminiscences of a Stock Operator
Amazon.com Sales Rank #987

Trend Following: How Great Traders Make Millions in Up or Down Markets
Amazon.com Sales Rank #868

Practical Speculation
Amazon.com Sales Rank #98,454

12/03/04
Michael Cook Jumps in with a Comment and a Question

The question is: why is it, do you think, that stocks do not appear to "trend"? Does it have something to do with the existence of earnings and valuation metrics? These would seem to take the appetite for stocks out of the realm of supply and demand, and overlay it with an analysis of what a stock is "worth".

The comment is: I have met several intelligent people who love craps because of the runs that occur, where a shooter seemingly has "hot hands" and keeps winning. They seem to believe that these runs are non-random. Or maybe, come to think of it, they just love the excitement generated by a table of people getting caught up in the action. The more I think of it, the more this is a picture of trend followers looking for a trend. And the contrarian - the guy who bets the don't pass line if I remember correctly, betting against the shooter - is usually hated and reviled, right?

The interesting thing to me is the cognitive limitation of humans making it hard for us to correctly perceive probability. If people are asked to write down a string of H's and T's "randomly", there will not be enough runs in the result. There seems to be a bias to expert "reversion to the mean" to kick in. This is similar to expecting a stock to go up after you have bought it because of an insight you just had.

So if trends are merely runs in data, and essentially random, then there's no system that will consistently be profitable. On the other hand, if a speculator can discern societal trends which result in price trends, that might be a different story.

12/03/04
George Zachar adds:

The "trick" is to carve out an area where one is an expert and can profitably acquire, evaluate and utilize information. Baldwin's tiny world of the pit, or Soros' macro stage, represent extremes along a continuum of opportunity. Personally, I never anticipated that my current "zone", options on 3 month interest rate futures, would become a dining hall. Study and trial and error (and error and error) brought me to this spot, and I suspect most specs must wander in the proverbial desert before finding "their" banquet table. I did. Have faith that your journey will be successful, and try to make the trek fun.

12/03/04
Shui Mitsuda on Fundamentals

As I see the failure by LTCM, and the most recent one, failure by China Aviation Oil in Oil Futures, I am reminded that being EXPERT IN THE FIELD, knowing and analyzing fundamental data and APPLY THEM TO FINANCIAL SPECULATION can be very DANGEROUS. Because financial market is full of uninformed participants, as well as participants who are unable to analyze the fundamentals and they often move the market. Floor trader, Tom Baldwin's comment echoes in my head. "I don't watch tv nor read news paper. But I just stand on the trading floor. By being the market, I get what I need for my trade. Problem with too much information is, the information itself could be right. But in next moment, it may not be right. Next it may not mean anything.  So what is important for me is what is happening on the trading floor at this very moment."

Since I am a mid-term spec, I don't fully follow Mr. Baldwin's advise but he has a very good point. I made a very bad investment 4 years ago by purely following fundamentals (cash flow, book values, P/L etc) that almost got me wiped out just because of negative public sentiment at that time, (which eventually affected the company's business so it was a double impact). The only reason I survived was I was not under time restricted speculation neither on LEVERAGE nor BORROWED MONEY. This sour experience reassured my life time speculation fundamental-- to follow Victor's Ed. Spec.

12/03/04
Philip J. McDonnell on Money Management

Steve Ellison reviewed some of the money management points from Stuart's book which I would like to rebut one by one (so I numbered them) because they appear to be based on several myths and half truths about probability. Here are some key points of Stuart's money management philosophy:

1.Most players are winning at some point and can remain winners by quitting at the right time.

There is no way to determine the right time. There is no guarantee that you will EVER be ahead at a casino. In fact the arc sine law says that a very large percentage of random walks will go above zero and stay above zero or go below zero and stay below zero indefinitely. See Feller for more on this. However, quitting is ALWAYS a superior strategy in a casino - even better is not to start at all.

2. Designate in advance an amount to put at risk and walk away if you lose that amount.

As I discussed in my earlier post today, this strategy will double your probability of being at or below that loss level. However it will neither give you an edge nor take one away. Walking away is your best bet.

3. If you get ahead, do not lose any more than the original amount at risk from that point.

This is a rising trailing stop idea. The same caveats apply here that your probability is going to double - only now you are ratcheting the stop up based on the recent high water mark. Thus the probability that you will get stopped out on a low is undoubtedly greater but I can't quantify.

4. If a streak of three or more is occurring, either bet with the streak or not at all. You can only lose once betting on a streak to continue. You can go bankrupt betting on a streak to reverse.

This is some more silliness. Dice and wheels have no memory. What they did on the last three realizations has nothing to do with the next outcome. You can also go bankrupt betting every time you see a streak of 3 or more.

5. If you win a bet, take the money off the table rather than pressing the bet.

Not betting and betting less in a casino is always a good thing!

12/03/04
Yishen Kuik adds:

Indeed, using a 2pt/-1pt money management strategy implies one believes that markets are positively auto-correlated over one's horizon, such that things that go up are likely to keep going up and things that go down are likely to keep going down. I believe one can even solve for how much the probability for the next move needs to be influenced by the prior move for this to be true by using a recombining binomial tree.

That is one condition that I can see which might cause the said money management strategy to have a positive expectancy. Many day traders seem to swear by some variant of this money management system, so I suppose the right question to ask is are equities positively auto-correlated over very short horizons?