Do markets change? Trend Following.

[Muskoka Joe]

"Personally, I like what Van Tharp has to say about developing not just one system to fit all market conditions, which is difficult, but to have at least three well-developed systems for various markets.

Here's a quick take from his book Super Trader, 2009: I like this idea of having different strategies for six market conditions 1. sideways stable, 2. sideways volatile, 3-4. bullish stable and volatile, 5-6. bearish stable and volatile. Tharp uses a 13-week window back-tested to 1950 - if the average change is less than 5.53% either way, he considers this a sideways market, great either way, bullish or bearish market condition. To measure volatility, he uses a 13-week ATR for the same period. Less that 2.87% ATR = quiet market, greater = volatile. Over the past 58 years, 41.35% of the time the Market is volatile, 11.91% it is bearish, 58.29% sideways, and 29.8% bullish. Curtis Faith (if you will pardon me!) makes similar points in his books."

================

^^^^ And how is this correlated to what strategy will work best over the NEXT 13 weeks? The answer is that this information provides zero edge in what system will work best over the right hand edge of the chart.

 

[salvador90]

Quote from Trend Following:

It all depends. There are plenty of trend followers who do not have this aim. The notion of equity smoothing is primarily designed to assuage nervous investors unfamiliar with or illogically unwilling to accept volatility.

BTW, humping people to another forum to hump them something else?

Believe me any trend follower would greatly accept a more smooth equity without sacrificing much or any of the returns...

AFJ Garner just answered it.

Smoothing out the equity is trying to preserve the same returns with less downside volatility...

I can name a few out of my head that you for sure interviewed in your book.

Salem Abraham is one... and I quote something he said back in 2006 I believe, that he was very happy to add new strategies to the main algorithm of his fund by reducing the volatility by HALF without much sacrifice of the returns.

If you go into his latest report you can see that already more than 30% of the fund's strategy is MEAN REVERTING algo's ... along with very short term trend following some other strategies and 35% being Long Term Trend Following.

Also look at Transtrend ! Do you want more smooth equity growth than that? In 16 years those guys had the biggest drawdown as 9% !! while achieving a compounded return of almost 19% per annum.

Running multiple systems for example... a long term trend following and another based on the same premises but for more shorter term... so when the market is chopping to bits in the Dailies, one cam take advantage of smaller shorter term trends.

My way of investing for instance is 50% long term trend following, 25% mean reverting and 25% short term trend following

 

[TSGannGalt]

More than 1/2 of you guys don't know what they're talking about... Do any of you actual implement a trend-following strategy? Have any of you done any reasonable research regarding TF on your own (rather than quoting some marketing material)?

And Mike, you need to do better than that... This thread is turning into something like a bunch of kids pointing fingers and calling each other's Mom / Dad whores and pimps....

See... this is very simple.

Markets Change = The probability of something happening is never constant (prob. density).
Trend Following = An outlier (fat tail) within a distribution.

Do markets change? Yes.
Do outliers occur? Yes.

The discussion here is trying to connect the same 2 things seperately. A simple basis of TF, you guys need to understand is:

Trend Following works (outliers happen to a point where it makes money) because the market changes.

 

[AFJ Garner]

Markets DO change. Trend Following DOES work. Yes.

Here are another couple of statements which will raise howls of anguish and disapproval:

Prediction IS necessary in trend following.

Discretion IS an integral part of mechanical systems trading.

It is perhaps not "necessary" to trade multiple systems, whether trend following or otherwise, to achieve some degree of smoothness. I do not think that Druz/Tactical does (judging from the disclosure document) and I like his results (apart from the 53% drawdown in the very early days). I think that most traders, given enough research and effort, can arrive at a single TF system which they will find bearable to trade in terms of win/loss ratio, drawdown, length of drawdown, volatility. Although they may not be able to approach the Transtrend example.

“Prediction” is necessary in a number of ways. For example, to take some sort of view on the likelihood of your system remaining profitable into the future (at least for a while). Prediction is an integral part of trend following: it says " I have spent much time back testing my plan and if in the immediate future the market does not change TOO quickly I EXPECT similar performance for the future”. Then, for a while, the trader can sit back and take his system’s trades without too much reflection. But to suggest that TF is purely reactive is abject nonsense.

“Discretion” is an integral part of so called mechanical systems trading because many decisions need to be made which (at the present time at least) can not be taken by a computer algorithm. The examples are legion but here a few thoughts: what portfolio will I trade, when and how will I alter my portfolio; what parameters will I set for my system and how and when will I alter these; a change in my data has caused a position to mysteriously appear/disappear – what should I do about it; 24 hour markets are changing the way people deal and the pools of liquidity – should I really leave my stops in the market any more;……………..and on and on and on.

Precious few of such questions strike the trader until he has spent a while trading mechanically and has lived through such moments.

The hapless beginner (and we are all beginners at some point in time) will spend much time reading books, buying courses and contributing to threads. He will only later learn that much of the “wisdom” he reads out there comes from those who make their money from the fringes of the trading world selling trading advice but not necessarily trading for themselves.

In a hilarious exchange on Seykota’s website, he suggests that Van Tharp should change the name of his best known book to “WRITE Your Way To Financial Freedom”.

Beware the BS. You will come to recognise it. Unfortunately it is not so easy until you haver a little trading experience under your belt.

In the meantime, those contemplating trend following or any other form of mechanical rule based trading have a very powerful anti BS spray at their disposal: back testing. Use it.

 

[TSGannGalt]

Quote from AFJ Garner:

Markets DO change. Trend Following DOES work. Yes.

...

In the meantime, those contemplating trend following or any other form of mechanical rule based trading have a very powerful anti BS spray at their disposal: back testing. Use it.

Everything you mention in the post comes down to a sematical issue.

1. What is the "mechanical" in mechanical systems trading? Depending on your own definition, I can agree and disagree on alot of points.

2. I know the guys at Transtrend in Holland (The Nether...) quite well. And they do trade multiple systems. They automate, but equally, they do add discretion to what they trade. Of course, "most" of their trading systems are trend oriented and mostly EOD stuff. The edge over other trend funds that they have is that they are very good at finding (discretion / macro-economics and other stuff) trends "globally". In another words, they're a combination of both "systematic" and "discretion".

3. I agree with the prediction part. Semantically, prediction gives people the wrong impression. In definition, "Probabilitic Outcome\Outlook" is also considered a prediction. On a trade-by-trade basis... you react but you're predicting that your reactive basis will continue to have significance.

4. Everything doesn't have to be automated. Sure. In the end, discretion is applied somewhere within your trading or development process, where decisions are required. In any trading process, it can be starting from "Whether you want to go Long/Short", "Whether you implement the system", "Whether you use some statistical criterion for deciding on a system", "Whether you decide on which development paradigm to implement"... etc. etc.

Anyways...

AFJ Garner, you sure like to use words that can get a lot of nit-picking nay-sayers like "Mechanical", "Prediction" and "Discretionary".

 

[AFJ Garner]

Quote from TSGannGalt:

AFJ Garner, you sure like to use words that can get a lot of nit-picking nay-sayers like "Mechanical", "Prediction" and "Discretionary".

[/B]

Unusual to get an intelligent discussion on this forum, so thank you for your input. It is good to provoke some thought for a change - that is what I have so appreciated about the TB forum. I have learnt a lot from various posters there in the past and continue to learn from intelligent posts by others. It is easy to separate the wheat from the chaff.

 

[TSGannGalt]

Quote from Trend Following:

Bottom line, why would one be switching among systems (or trying to predict "market conditions") if they did not think they could avoid losing periods or down months? And if this was something that was possible...why do the best in the business (folks like Harding, Eckhardt, Aspect, Abraham, etc.) regularly have losing months? A good learning exercise: read carefully the disclosure documents of trend following traders. Decades of month by month performance. Much more indicative of the future real world over any one man's theory.

1. I can't speak for Abraham and others, but in terms of Eckhardt... The traders there do a lot more things than TF. They have Mean Reversion models and have a good team of Quants in their firm. For some reason, I know (Poker buddy... ) a guy working at Eckhardt as an Options trader... and he adds discretion over the pricing models... Is there another Eckhardt in Chicago?

2. PPMs, Disclosure Docs??? Do you seriously take them literally? Look at Amaranth, their marketing pitch was that they were a Quant. Fund with strictly followed top notch risk management... Does Eckhardt's PPM/Disclosure mention that they "ONLY" trade Trend-following models that was developed 20 years ago line by line? NO. I can guarantee you that their models have evolved with the "market condition".

Don't get me wrong, I have nothing against Trend-Following... I run models based on it (along with others). But you really need to stop concluding points based irrelevant material. Personally, I think you need to start studying a bit more about the OPM industry so that people can start taking you more seriously. It seems like no one is willing to providing you with relevant info, beyond the same ol' pitch you get from those public marketing docs.

If there is a difference between you and Curtis... He's got more inside friends in the industry.

 

[JezLiberty]

Quote from TSGannGalt:

Trend Following works (outliers happen to a point where it makes money) because the market changes. [/B]

This is exactly my point of view too: markets change and fat tails ensure trend following is a succesful strategy http://bit.ly/6bsiS

Quote from AFJ Garner:

Prediction IS necessary in trend following.

This is something that I realised this weekend as I was pondering the "philosophies" of mechanical systems design and testing. For example, if you take the argument that price distributions are mostly random with fat-tails, and since we cannot predict, trend following is the best way to profit from these fat-tails... Deciding to trade a TF system contains a huge contradiction in itself:

This is a bit of a devil's advocate twisted way of thinking.. But the (rhetorical) question is how/where do you draw the fine line between what is predicting and what is not... How much of past information can you use to "safely plan" for the future.

Quote from AFJ Garner:

Unusual to get an intelligent discussion on this forum, so thank you for your input. It is good to provoke some thought for a change - that is what I have so appreciated about the TB forum. I have learnt a lot from various posters there in the past and continue to learn from intelligent posts by others. It is easy to separate the wheat from the chaff.

I agree about the TB forums - I am new there (here as well actually...) but the level of quality seems top-notch (it's mostly TF though, but thats what I am interested in).

 

[TSGannGalt]

Quote from JezLiberty:

This is exactly my point of view too: markets change and fat tails ensure trend following is a succesful strategy http://bit.ly/6bsiS


This is something that I realised this weekend as I was pondering the "philosophies" of mechanical systems design and testing. For example, if you take the argument that price distributions are mostly random with fat-tails, and since we cannot predict, trend following is the best way to profit from these fat-tails... Deciding to trade a TF system contains a huge contradiction in itself:

This is a bit of a devil's advocate twisted way of thinking.. But the (rhetorical) question is how/where do you draw the fine line between what is predicting and what is not... How much of past information can you use to "safely plan" for the future.



I agree about the TB forums - I am new there (here as well actually...) but the level of quality seems top-notch (it's mostly TF though, but thats what I am interested in).

Personally, I'm not biased and don't have a personal preference between trading something "normal" and "outlying". Only difference is that you're exposing different distribution profiles within the market. From a very generalized sense:

Stat. Arb. (Mean Reversion) assumes that the distribution is normal and you trade using the normal.

Momo (like the old WorldCo and Schony...) assumes that the distribution is normal and you trade using what's outlying.

Trend-following assumes that the distribution is random and trade using the outlier.

Most forms of Price Action (Swings, Flow, Pricing model based trading etc. etc.) assumes that the distribution is random and trade using the normal.

It's a matter of understanding the "concepts / tendency / character / etc. / etc. (in another words... whatever you want to call what the market does)" behind the trade and quantifying (test and automation) them as much as possible.

Prediction... That's a real toughy... Fine line historically... The question is... both random and normal is equally viable and equally has underlying risks. My philosophical conclusion comes down to being extremely paranoid about risk management. You make sure you don't get hit hard when the prediction is wrong... Kinda like position sizing (portfolio weight) and stops (cut off point) for a single trade... only done within a system to system basis...

Let me note again that the examples I gave are OVER-GENERALIZED.

----------------------------------

And I like the TB forums... all the old school guys from Chuck's Traders Club used to post early on but don't see much of them now....

 

[salvador90]

Quote from TSGannGalt:

Everything you mention in the post comes down to a sematical issue.

1. What is the "mechanical" in mechanical systems trading? Depending on your own definition, I can agree and disagree on alot of points.

2. I know the guys at Transtrend in Holland (The Nether...) quite well. And they do trade multiple systems. They automate, but equally, they do add discretion to what they trade. Of course, "most" of their trading systems are trend oriented and mostly EOD stuff. The edge over other trend funds that they have is that they are very good at finding (discretion / macro-economics and other stuff) trends "globally". In another words, they're a combination of both "systematic" and "discretion".

3. I agree with the prediction part. Semantically, prediction gives people the wrong impression. In definition, "Probabilitic Outcome\Outlook" is also considered a prediction. On a trade-by-trade basis... you react but you're predicting that your reactive basis will continue to have significance.

4. Everything doesn't have to be automated. Sure. In the end, discretion is applied somewhere within your trading or development process, where decisions are required. In any trading process, it can be starting from "Whether you want to go Long/Short", "Whether you implement the system", "Whether you use some statistical criterion for deciding on a system", "Whether you decide on which development paradigm to implement"... etc. etc.

Anyways...

AFJ Garner, you sure like to use words that can get a lot of nit-picking nay-sayers like "Mechanical", "Prediction" and "Discretionary".

Very interesting what you say about Transtrend--- so what they do is nit-pick economically and fundamentallly the best potential universe of assets at a certain point and then apply their systems onto the assets they think economically and fundamentally that have more potential? Is that it?

Regards,