Long term trend following on equities a fool's game?

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.
ecritt
Roundtable Knight
Roundtable Knight
Posts: 134
Joined: Sat Aug 28, 2004 3:44 am
Location: Phoenix, AZ
Contact:

Post by ecritt »

AntiMatter wrote:
bobsyd wrote:What do you mean by "GARP-style filter"? Perhaps an example?

Thanks.
e.g. decent eps growth with not-too demanding p/e. Or even just simply eps growth of perhaps >20%.

I'm too inexperienced to give you anything more precise than that, sorry. It's just from my limited knowlege and experience, these types of screens provide some sort of edge, and it seems intuitive that they would complement some measure of momentum

Lots of people must have looked into this....
There is no shortage of people that insist this is the way to get an edge. There is no shortage of backtesting platforms (portfolio123, Zach's, etc.) that show incredible results from using such valuation criteria to pick stocks. But, for some reason(s), this approach has not worked in real life very well. From what I can see, firms that have implemented such have been worse off for it...in many cases dramatically worse off.
AntiMatter
Roundtable Fellow
Roundtable Fellow
Posts: 86
Joined: Sun Jul 05, 2009 11:10 am

Post by AntiMatter »

ecritt wrote: From what I can see, firms that have implemented such have been worse off for it...in many cases dramatically worse off.
Weird. Any ideas about why this might be the case?

In my short foray into studying the markets, they have done nothing but confound my logic :cry:
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

AntiMatter wrote:Weird. Any ideas about why this might be the case?

In my short foray into studying the markets, they have done nothing but confound my logic :cry:
I will leave your question to ecritt.

Whatever the reason may be, I would suspect that many, many people involved in trading and investment have never back tested and have no clue that some of their dearest held beliefs about markets may be fallacious.
ecritt
Roundtable Knight
Roundtable Knight
Posts: 134
Joined: Sat Aug 28, 2004 3:44 am
Location: Phoenix, AZ
Contact:

Post by ecritt »

AntiMatter wrote:
ecritt wrote: From what I can see, firms that have implemented such have been worse off for it...in many cases dramatically worse off.
Weird. Any ideas about why this might be the case?

In my short foray into studying the markets, they have done nothing but confound my logic :cry:
I could write a book about this topic. But for now let me give you a scenario and ask a question.

Scenario: You've identified a stock with nearly perfect fundamentals, wonderful earnings growth, great products, fantastic profit margins, etc.

Question: Why would you think this stock's price doesn't already reflect all of this stuff?
AntiMatter
Roundtable Fellow
Roundtable Fellow
Posts: 86
Joined: Sun Jul 05, 2009 11:10 am

Post by AntiMatter »

ecritt wrote: Question: Why would you think this stock's price doesn't already reflect all of this stuff?
Because I would be thinking that, for a variety of reasons, markets aren't terribly efficient? Couldn't we ask the same question of e.g. trend following? In fact, could some trends be related to periods of price re-discovery?

Anyway, I feel like I'm probably derailing this thread (and getting out of my depth....)
ecritt
Roundtable Knight
Roundtable Knight
Posts: 134
Joined: Sat Aug 28, 2004 3:44 am
Location: Phoenix, AZ
Contact:

Post by ecritt »

AntiMatter wrote:
ecritt wrote: Question: Why would you think this stock's price doesn't already reflect all of this stuff?
Because I would be thinking that, for a variety of reasons, markets aren't terribly efficient? Couldn't we ask the same question of e.g. trend following? In fact, could some trends be related to periods of price re-discovery?

Anyway, I feel like I'm probably derailing this thread (and getting out of my depth....)
In my experience, perfect fundamentals (using last quarter's financial statements) tend to correspond with the end of the price discovery process on the way up and the beginning of the price discover process on the way down. I believe people usually pay a premium to invest in stocks that make them "comfortable".

With respect to trend following, as I understand it, no weight is given to fundamentals. They do not influence the security selection...which tend to have trend followers buying/selling at times that make most people very uncomfortable. Which, is why I think trend following works. Most people must "lose" in this game. Alpha, while variable, is finite and must come at the expense of someone else.
AntiMatter
Roundtable Fellow
Roundtable Fellow
Posts: 86
Joined: Sun Jul 05, 2009 11:10 am

Post by AntiMatter »

OK that makes alot of sense, thanks.

On the other hand, I'm still of the belief that there of the simple fundamental scans, there is good evidence for Growth/GARP etc. Sorry, I have no references to hand - perhaps I need to revisit the evidence.

In relation to your arguments, there may be an "unrealised" element of growth - e.g. strong management will continue to make intelligent decisions, the company will move up through the indexes attracting institutional support, our screens may allow us to get in on the ground floor of a bubble, etc.

Surely we can still leverage some edge over behemoth institutions (with their procedural limitations) and the fishy retail investor?
Mark Johnson
Roundtable Knight
Roundtable Knight
Posts: 122
Joined: Thu Apr 17, 2003 9:49 am

Post by Mark Johnson »

I enjoyed Richard Bauer's book and its discussion of using fundamentals as filters for technical analysis (a/k/a "market timing") systems for trading stocks. He discusses his attempt to find good fundamental screens for trading the S&P 500 stock index, using 1984-1988 data for the "in sample" system development and optimization, then testing his system on 1989-1991 "out of sample" data.

His approach was ambitious:
To narrow the variable list in our search for attractive common stock timing rules, we ran regressions of all 167 different time series (in the "Stocks Bonds Bills & Inflation" database -MJ) and the changes in these series against the monthly stock return series. The appropriate data availability lags were always considered. For example, February housing starts were linked with May stock returns due to the data availability lag. Next, we examined the regression runs and narrowed the list of variables down to the ten data series having the highest correlation (magnitude, either plus or minus) with the asset series being considered.
The ten series he selected are shown in Figure 1. They are 1-7 and 9-11 on the list; number 8 was omitted because it was too similar to number 1. The list is sorted by decreasing magnitude of correlation.

He used the fashionable and trendy buzzword of those times, "Genetic Algorithms", to test millions of parameter variations (on the 1984-1988 "in sample" price data), and after examining the test results, selected 120 trading rules as being the best, most balanced, portfolio of systems. He wrapped these 120 rules into a SimulTrading Suite (as Blox would call it). Then he ran his Suite on the 1989-1991 "out of sample" data.

The results?
For the entire 1989-1991 out of sample period, the accuracy was 53.06 percent winners, and the average annual return was 14.96 percent -- less than that of a buy-and-hold strategy, which would have earned a 18.46 percent average annual return over the same period.
Attachments
Figure 1.
Figure 1.
bauer.jpg (88.57 KiB) Viewed 13980 times
AntiMatter
Roundtable Fellow
Roundtable Fellow
Posts: 86
Joined: Sun Jul 05, 2009 11:10 am

Post by AntiMatter »

For the entire 1989-1991 out of sample period, the accuracy was 53.06 percent winners, and the average annual return was 14.96 percent -- less than that of a buy-and-hold strategy, which would have earned a 18.46 percent average annual return over the same period.
I'm not a big fan of these drag-net approaches. Genome scans are one field that has had success, but even then, with their huge sample sizes, the results are often mired in statistical controversy.

I would much rather start with a simple, logical & testable hypothesis and work from there. This is a key attraction of Fatih' s approach as compared to, for example, the experimental chapters of "Evidence Based Technical Analysis". It's more an epistemological divide than Fundemental Vs Technical.
ecritt
Roundtable Knight
Roundtable Knight
Posts: 134
Joined: Sat Aug 28, 2004 3:44 am
Location: Phoenix, AZ
Contact:

Post by ecritt »

"I'm still of the belief that there of the simple fundamental scans, there is good evidence for Growth/GARP etc."

I think it would be more fair and accurate to replace the word 'belief' with the word 'desire' at this point. (no disrespect or sarcasm intended here)

"Sorry, I have no references to hand - perhaps I need to revisit the evidence."

I recommend you consider the source of any evidence and also look hard for the evidence that isn't being presented.

Example...

Presented evidence:
Countless newletters, backtests, subscriptions, marketing materials, websites, etc. that show 25% CAGR from selecting stocks according to some fundamental screening criteria.

Conveniently overlooked evidence:
Where are all the firms delivering 25% CAGR to their investors?

Followup evidence:
A few individual programs from a few firms did manage to deliver 25% CAGR to investors over some short period of time.

Conveniently overlooked evidence:
By my calculations, considering the thousands of programs, designed by hundreds of firms, there should be more than just a few that did well due to random chance alone. Where are they? Why don't they exist?
Rush
Roundtable Fellow
Roundtable Fellow
Posts: 50
Joined: Wed Dec 31, 1969 7:00 pm

Post by Rush »

Dear All,

I feel that 167 different time series (taken from "Stocks Bonds Bills & Inflation" database) are arbitrarly selected and it seems we are missing the most funny-mental indicators available to successfully time the SP500 Index.

If you use the 140 Countries International data series (published by United Nations) you may find a simple fundamental factor which has an R^2 of 75% (yearly data period form 1983- 1993) with S&P500 Index.
It is the Butter production in Bangladesh.
Can we do better?
How about combining in a linear fashion, the Butter production in Bangladesh with USA cheese production? A wonderful R^2 of 95%.

Taken from "Stupid data miner tricks: overfitting the S&P500" from D. Leinweber, PhD. (1995 Caltech)
AntiMatter
Roundtable Fellow
Roundtable Fellow
Posts: 86
Joined: Sun Jul 05, 2009 11:10 am

Post by AntiMatter »

Don't worry, no disrespect taken. I agree with everything you are saying to some extent - I'm pretty sceptical of everything, including my own beliefs (and yours). I'm sure you realise that I am playing devils advocate to try and understand the issues better.

Could we not re-write your last post, but replace the word "fundamental" with "technical"? In both fields, the analysis of real-life results is confounded by several issues - especially the fact that the primary motive of most of these organisations is deposits, rather than achieving the highest CAGR%. Moreover, as I alluded to earlier, many of the fundamentally driven-funds suffer from very obvious structural deficiencies.

It would be no surprise that in either route, your own results (after a period of research, testing, etc) will be better than the results of a caretaker.

I remain completely open minded about the fact that fundamental screening can provide some sort of edge.
ecritt
Roundtable Knight
Roundtable Knight
Posts: 134
Joined: Sat Aug 28, 2004 3:44 am
Location: Phoenix, AZ
Contact:

Post by ecritt »

You are absolutely correct. I'm not making a case for technical analysis. In over 10 years of full time effort the only edge I've found that's consistent across decades and continents has to do with the fact that the vast majority of a market's gains have come from a small minority of stocks. Likewise, the vast majority of a market's losses has also come from a minority of stocks. The "tails" are fatter and more consistent than people realize. One can attempt to capture the "tails" with a long-volatility and/or anti-martingale approach.

I admitt I'm biased against fundamental analysis. But feel that I'm justified in being so. I went to business school, learned MPT, financial statement analysis, etc. Went straight to work in the family office / hedge fund / investment advisory / fund of funds world straight out of college and have been here ever since. I have to respect the re-occuring evidence seen with my own eyes. People that anchor off last quarter's financial statements consistently own those left tails and miss those right tails. From what I can see, price almost always moves before the new "poor" fundamentals are communicated to the public. After all, that is the nature of a discounting mechanism.
AntiMatter
Roundtable Fellow
Roundtable Fellow
Posts: 86
Joined: Sun Jul 05, 2009 11:10 am

Post by AntiMatter »

Funnily enough, the first (and probably last....) epiphany I have ever had with regard to trading was when I noticed the fat tail effect. I remember getting really excited, ranting at my wife whilst drawing graphs in the air, lol.

It's interesting how coming at the problem from a gambling background such features present themselves more readily. I am still very excited by the prospect of essentially turning the marketplace into a simple numbers gambling game.

It's undoubtedly an elegant solution. At least in theory. Of course, the reality is always much messier, such are the problems I am struggling with now...

But I still don't believe it is the only way to win. Although as an aside, I think that often the problem with fundamental based approaches is the naivety with regard to these simple concepts. It should be possible to screen for small young companies with good growth prospects that are trending up, let the winners run & cut the losers.

One last question:
In over 10 years of full time effort the only edge I've found that's consistent across decades and continents
How about mean-reverting strategies and/or false-breakout type approaches? e.g. almost the polar opposite of what we have discussed; where the aim is to have lots of small winners, but at a high probability.
ecritt
Roundtable Knight
Roundtable Knight
Posts: 134
Joined: Sat Aug 28, 2004 3:44 am
Location: Phoenix, AZ
Contact:

Post by ecritt »

"It's interesting how coming at the problem from a gambling background such features present themselves more readily. I am still very excited by the prospect of essentially turning the marketplace into a simple numbers gambling game."


I’m with you on that. Imagine there’s a casino that offers a game with the following:

A 55% chance of a 46% gain and a 45% chance of a 34% loss. Virtually no table limits or restriction on bet size. That’s a positive expectancy of 9.8%...which almost any gambler in the world would consider a significant tail wind. These numbers are from the individual annual returns of the 3000 most liquid stocks (including delisted) in the U.S. over the last 18 years; or 54,000 individual annual stock returns.

The interesting part is that if you omit the top 5% performing stocks each year (150 out of 3000) the whole thing turns negative; you actually lose money over 18 years if you are unlucky enough to miss the best performing 5% of stocks each and every year. That means there are some big returns in that right tail each year.


“How about mean-reverting strategies and/or false-breakout type approaches? e.g. almost the polar opposite of what we have discussed; where the aim is to have lots of small winners, but at a high probability.â€
ecritt
Roundtable Knight
Roundtable Knight
Posts: 134
Joined: Sat Aug 28, 2004 3:44 am
Location: Phoenix, AZ
Contact:

Post by ecritt »

"many of the fundamentally driven-funds suffer from very obvious structural deficiencies"

Would you elaborate on this statement?
AntiMatter
Roundtable Fellow
Roundtable Fellow
Posts: 86
Joined: Sun Jul 05, 2009 11:10 am

Post by AntiMatter »

ecritt wrote:"many of the fundamentally driven-funds suffer from very obvious structural deficiencies"

Would you elaborate on this statement?
I'm sure you know more about this than me, but what I'm getting at are the majority of the big players - insurance companies, pension funds, mutual funds etc don't seem to be particularly intelligently run (unless I am missing something?):

1) Their portfolios aren't necessarily constructed with CAGR% as a primary goal. Rather, they might aim to come within a certain % of the index.
2) They suffer from corporate structure problems with rather arbitrary limitations e.g. have to be 90% invested at all times, only S&P500 companies, long only etc. These rules seem to have been dreamed up by an out-of-touch manager (something anyone who has worked in an office can relate to ;))
3) problems of sheer size whereby they can't take up a significant holding without moving the market.

This results in them all appearing to have a portfolio which comprises a random assemblage of 1% of portfolio on blue-chip companies. It seems that these players are contributing somewhat to the fat-tail affect, which we aim to exploit. However, such inherent problems suggest they should actually be very exploitable by a small player, using a variety of different strategies which aim to gamble on who might be included in the indexes in the future.
ecritt
Roundtable Knight
Roundtable Knight
Posts: 134
Joined: Sat Aug 28, 2004 3:44 am
Location: Phoenix, AZ
Contact:

Post by ecritt »

"the majority of the big players - insurance companies, pension funds, mutual funds etc don't seem to be particularly intelligently run (unless I am missing something?)"

No, you've got it pretty much right. Those entities are the payers of what we call the "convenience premium"...the largest and most consistent source of extractable alpha in the markets.
danZman
Roundtable Knight
Roundtable Knight
Posts: 117
Joined: Thu Mar 10, 2005 10:07 am
Location: Newport Beach,CA

Post by danZman »

When it comes to equities, I have proven to myself that
what I'm trading against is large funds < $1 billion.

I believe there's a considerable edge there because
most managers learned the same BS at the same
school with a different name.

In other words, they all react to the same data at the
same time. Pretty worthless education one might think.

My professor drove a beat up old Honda, so that makes
sense.
ratio
Roundtable Knight
Roundtable Knight
Posts: 338
Joined: Sun Jan 15, 2006 11:07 pm
Location: Montreal, Canada

Follow up information, I got a question about what happen

Post by ratio »

since I posted this:

In real life, with a smaller account ;). My real result are as last week: CAGR: +8.12% Drawdown: 29%


I ran the test, starting with 30,728,588 $


End balance: 37,117,344.17,
CAGR: 4.5
Largest DD: 12.8%

Here is the Chart
Attachments
Post20120521.jpg
Post20120521.jpg (65.96 KiB) Viewed 10277 times
Post Reply