Trend following

Discussions about the psychology of the markets and the masses as it relates to trading.
Post Reply
damian
Roundtable Knight
Roundtable Knight
Posts: 815
Joined: Tue Apr 15, 2003 8:43 pm
Location: dusseldorf

Trend following

Post by damian »

I have noticed more instances of 'profit targets' in trend following.

- I see chatter on this forum (I myself have investigated the concept)

- I see commercial systems hitting profit targets all the time these days. $10k seems to be a popular target (I found this also through optimisation, and dismissed it as useful)

- I see commercial systems that have no re-entry after a profit target exit

- I see commercial systems that use a fixed dollar target per contract, and then apply that target to a diversified portfolio: $X profit target for TY, O, C, NG. I am not joking. What is worse, some of these portfolio systems were designed and tested using Trade Station, with all its obvious limitations in that style of system development.

- I see commercial systems missing out on continuation of excellent trends. Trends that are designed to be followed (at last, her majesty The Trend has arrived, viz, NK, JY. Both instruments have triggered profit target exits of late in more than one commercial system and investors in those systems are now sitting on the sidelines and likely not sleeping so well as trends continue. Which is odd, as they likely weren't sleeping so well in times when they were giving up open profits. There are several other fantastic trends under way that have also triggered profit target exits).

Why am seeing more interest in profit targets? This is a serious question that I ask myself. Is its:

a) because only in the last year have I started to look at commercial systems. Had I looked earlier I would have found that they have always been there.
b) they are actually becoming more popular after a long period of trend following losses and/or sharp reversals in open profits.

It was (b) that motivated me to investigate profit target exits in my own system research. That and one of my family members who reached fever pitch when it came to continual trend following draw down whilst giving back open winners. The more he pushed to 'lock in profits!' and 'leave some for the next guy!', the more I knew he was not unique in the trend following universe and the more I resisted his pressure to protect open profits and take large profits. In fact, I have implemented the exact opposite.

It seems more and more people are forgetting how to follow trends: they have forgotten what a good trend tastes like and have altered their diet. I think this bodes well for trend following.

[post script: I think that there is room in a portfolio of multiple systems for a strategy that takes profit and attempts to re-enter on continued strength. However, this would only be a secondary system to compliment the base trend following strategy. I have a nice strategy that trades daily bars and like a normal trend following system 80% of the time. The remaining 20% of the time it progressively tightens stops as the market accelerates in favour of the trade. This system technique benefitted my old system with considerable improvement in times of rapid trends and no benefit at all in times of steady trends or whipsaw markets. I do not currently trade this system as I have designed a much better one and not enough capital to trade both]
TK
Roundtable Knight
Roundtable Knight
Posts: 167
Joined: Tue Apr 15, 2003 5:45 pm

Post by TK »

Sticking with winning trades for as long as possible is the only way to make big wins.

The above quote comes from the synopsis of the book recommended recently. The author is said to have earned big bucks and retired at the age of 34 by riding out huge long-term trends at a Bermuda-based hedge fund.

http://members.ozemail.com.au/~farleigh/taming.html

http://tinyurl.com/9hjj5
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

I doubt that Bill Dunn and John Henry and Jerry Parker and Liz Cheval and Robert Rotella have altered their approach towards profit targets. Since they represent 100X more market muscle than the sum total of all us retail scum on this message board system, I doubt that "new attitudes towards profit taking" will alter the course of market prices a/k/a change how trends behave.

Also I encourage owners of TBB to fiddle with the presupplied Blox named "Profit Target Exit". It may suggest some ideas for ways to build a multi-system Suite that might provide a bit of additional smoothing. One possibility is sketched below.
Attachments
ATR Channel Breakout + Profit Targets
ATR Channel Breakout + Profit Targets
pt.png (34.67 KiB) Viewed 30395 times
damian
Roundtable Knight
Roundtable Knight
Posts: 815
Joined: Tue Apr 15, 2003 8:43 pm
Location: dusseldorf

Post by damian »

Been there, done exactly that (but before TBB shipped with a pre-made profit target block). In fact, I prompted the idea of using multiples of initial risk as an exit trigger (at first it could not be done so easily as the required object properties did not exist, but now they do and now it is easy). All of my TBB beta testing was based on playing around with profit targets (for want of a another consistent theme on which to base my beta involvement)

I agree that the big guys of trend following will not change their dietary habits. But I keep in mind that Aberration was always just a retail system and was commonly quoted as being so popular that it ate itself (in terms of performance) and stability of trends in the markets in which it traded.

On reflection, I probably side with you Sluggo: trends will remain trends irrespective of the greed and fear of retail punters. I think I got a bit carried away in my above post.
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

Retail or not, Ab is really no more than a generic system mentioned in books from Kaufmann to Le Beau and Lucas under the general concept of trading envelopes. And it can be tweaked to show some very nice profits indeed over the past two to three years.

Re profit taking, Jerry Parker apparently does it or did it. Reluctantly so, according to an interview with him.

Profit taking does indeed "smooth" but in my testing at least it smoothes by preventing the equity curve going so high in the first place. In other words it has less far to fall when it inevitably does. And it reduces CAGR.

So I don't -take profits that is.
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

In this example, a profit target (run #2) increases CAGR, decreases drawdown, increases MAR ratio, and increases Sharpe ratio.
Attachments
ptx.png
ptx.png (7.95 KiB) Viewed 30222 times
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

I should have stated that my testing was of Bollinger Band Breakouts using long term MAs. With and without profit taking, testing a single version at a time.
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

Hmm, using R Multiples as profit taking targets rather than an arbitrary target certainly produces some interesting improvements. IE looking at the biggest R Multiple profits your system produces and adjusting the profit taking target so as not to cut these short..................but so as to also not let them run away from you.
TK
Roundtable Knight
Roundtable Knight
Posts: 167
Joined: Tue Apr 15, 2003 5:45 pm

Open Position Risk

Post by TK »

Speaking of smoothing equity curves, is it possible in TBB to test the following rules?

Code: Select all

If the risk (difference between the last close and the stop) of any open position exceeds 5% of the total equity, reduce the size of this position on the next open to keep the risk within the 5% limit.

If a 10xATR move against an open position represented a loss of more than 5% of today's total equity, reduce the size of this position on the next open to keep the risk within the 5% limit.

That's something Chuck B. wrote about some time ago in these two posts:

viewtopic.php?p=530#530
viewtopic.php?p=709#709
Last edited by TK on Thu Nov 24, 2005 5:44 am, edited 1 time in total.
Austrian
Roundtable Knight
Roundtable Knight
Posts: 223
Joined: Sat Oct 25, 2003 12:35 pm
Location: Austria, Vienna

Post by Austrian »

Yes, it is possible. There is a instrument property instrument.totalPositionRisk.
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

I eat my words - with more aggressive targets on a longer term system I am seeing higher CAGRs and lower drawdowns. On the parameters I am using the differences are not so huge but material nonetheless.
damian
Roundtable Knight
Roundtable Knight
Posts: 815
Joined: Tue Apr 15, 2003 8:43 pm
Location: dusseldorf

Post by damian »

AFG - indeed using R multiple profit targets rather than $ triggers is the better option for portfolio trading. Dollar triggers are rubbish if you trade C and SP in the same portfolio.

However, I am very concerned that optimising for maximum the perfect profit target is a dangerous game. Potentially the most risky parameter you can optimise. To reduce this risk I decided that rather than simply take profits at the trigger (i.e., picking tope with the benefit of hindsight), I implement a much tighter stop once a certain R multiple has been hit.

It is pretty easy to improve backtest results by telling your system to, in retrospect, take windfall profits at the peak in the market. And that is exactly what optimising for the best profit target is doing, particularly if the best results came from the largest R multiple instance in the parameter step range and that trigger level was hit very infrequently (define frequent) and only in the bigger trending markets. I have a variable which counts how often the profit target exit was actually executed. Have a look at each instance: you will be very proud to see that the exit was right at the top of a huge rally, just before a large, sharp and rapid reversal in the market. What a clever system! ;)
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

I have been fiddling with my own profit target blox for the Bollinger Band Breakout system. It certainly gives food for thought and I am clearly guilty of not paying enough attention to this area and in particular of not experimenting with scaling out at higher risk multiples - the longer term the system, the higher risk multiple to use for scaling out.

So on BB BO 80/2/0 scaling out at 1 to 3R provides some interesting food for thought. For a BB BO using say 300/1/-0.5 scaling out at R Multiples of 2 to 10 provide further interesting thoughts. For the latter I use a hold stop at the moving average so that I can use a still reasonable bet size while ensuring enough contracts are put on to give a healthy CAGR.

And I have been looking at scaling out 10% of a position at a time up to selling out completely when the target is hit.

While I have found that there CAN be slight improvement in both CAGR and Max DD using scaling out (as opposed to letting each trade run its course) what is of greater interest to me is the way scaling out completely alters the win loss ratio. And of course as Sluggo says, some potential reduction in volatility of the equity curve.

A couple of examples will suffice. On a portfolio of 34 well trending markets (which include some stock indices) on the longer term BB BO outlined above, scaling out at 50% each time the target is hit and using a consistent bet size of 2% produces the following:

Scale out at R Multiple of 2: CAGR 28.2%, Max DD 22.2%, Wins = 68.6%
Scale out at R Multiple of 4: CAGR 42%, Max DD 30%, Wins = 54.5%
Scale out at R Multiple of 8: CAGR 45.9%, Max DD 29.7%, Wins = 45.3%
No Profit Taking: CAGR 46.2%, Max DD 29.8%, Wins = 40%

Now personally I would tend to rule out taking profits at the 2x R Multiple but even so, look at the win loss ratio: over 68% of trades are winners. This makes it a comfortable sort of system to trade. And I should add that percent profit factor is a respectable 2.17% with expectancy at 0.33.

Using the R Multiple of 8 looks attractive and produces a percent profit factor of 3.86 and expectancy of 1.53.

Using no scaling out gives a larger per cent profit factor of 4 and an expectancy of 1.79 and in this case the highest CAGR. But one may be far more comfortable using SOME scaling out to enjoy the comfort of a much higher percentage of wins. In this example there is little to be gained in terms of the percentage of positive months . I am aware I should have quoted some measure of curve smoothness (other than simple max DD). No doubt also people will notice differences in the time taken to recover from a DD. But enough.
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

Incidentally, many of the better test results displayed on this forum (including my own outlined above) seem to have been constructed by overly fitting the portfolio and using very high portfolio heat.

I deliberately chose the elements in the portfolio to show an attractive MAR and I boosted this by a combination of a relatively high bet size (2%) and a fairly large portfolio of futures which have trended well and fit together well (or at least have done so historically). I have seen results posted by others using a bet size higher even that I chose and with a far , far greater overall portfolio heat.

I chose as part of the portfolio a few stock indices which have done wonderfully well over the past couple of years and added a few other well chosen items. All with the benefit of hindsight - wonderful how it can improve one's hypothetical trading!

Most of us will realise that some markets have shown good trending ability over the past 20 years (energies, interest rates, currencies for instance), some have shown almost no abilty at all (OJ, PB, CC perhaps) and some have been marginal for many systems (GC, SI).

Designing a portfolio to trade may require a different set of skills.
BARLI
Roundtable Knight
Roundtable Knight
Posts: 650
Joined: Sat Jan 17, 2004 6:01 pm
Location: USA

Post by BARLI »

AFJ Garner wrote:

Most of us will realise that some markets have shown good trending ability over the past 20 years (energies, interest rates, currencies for instance), some have shown almost no abilty at all (OJ, PB, CC perhaps) and some have been marginal for many systems (GC, SI).

Designing a portfolio to trade may require a different set of skills.


Garner, what currencies did you find that have good trends? From my backtesting I found that Japanese Yen has trends, where canadian Dollar is not trending. To the non trending markets I will add Cotton...

Now tircky question:

Has anyone here did some studies on when the markets alert that there's a trend on the way? If so please share some thoughts what you guys did.
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2987
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo »

It is actually a wonderfully instructive little exercise to put together a portfolio of ten or twelve "poorly trending" markets, then find a parameter set for (lets just say) the Dual Moving Average system, whose performance is "optimum" for the Crappy Portfolio. Now take this same system and parameter set and run it on All Markets. Wow! Son of a (word)! It performs (verb)ing GREAT on All Markets!

Our friends and colleagues on the Trading Recipes list, did this exercise about two years ago. Their Crappy Portfolio consisted of
  • Australian Dollar
  • British Pound
  • Cocoa
  • Canadian Dollar
  • Canadian 10-Year Bonds (Montreal Futures Exchange)
  • Long Gilt (LIFFE exchange in London)
  • Copper
  • Lean Hogs
  • Orange Juice
  • Pork Bellies
  • Rough Rice
  • Silver
Some of these are, imho, inspired choices. If ever oh ever a Crappy market there was (for trend following), Silver gets my vote. Closely followed by Cocoa, Pork Bellies, and Orange Juice.

It might be worth a little of your time.
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

Indeed, indeed. I had that excellent discussion from the TR list in front of me last week as I wrote the above. In essence Neal Stevens (now of Ancile Capital Management - www.ancilecapital.com) stated his preference for widening out the stops for a more reliable system and trading everything under the sun. The very opposite of curve fitting a portfolio.

In terms of sizing positions, widening out the stops need not reduce the number of contracts you trade on each signal. It is well worth comparing for instance the bollinger band parameters I mention above. The distance between 2 standard deviations of closing prices over an 80 day period and the 80 day moving average give a very similar risk AT ENTRY to that of 1 standard deviation of 300 day's worth of closing prices over a 300 day moving average. Use a stop at the 300 day moving average and hold that stop if you want to set the exit at -0.5 - which is what I did. I used similar filters for both to ensure that trades were only taken in the direction of the longer term trend.

What you find is that each system will trade a similar number of contracts(and indeed will often have very similar entry dates) but that since price rapidly moves away from the longer term MA in a trending market, the stop rapidly drops away from the price, so that such a system has traded well in the choppy markets of the last two to three years.

But of course you can achieve similar effects by widening out the stops with any other half decent system.

In view of the similar entry points I concluded that combining the two systems did not produce an ideal mix to trade together however.
AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2071
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner »

Sluggo reported above: "In this example, a profit target (run #2) increases CAGR, decreases drawdown, increases MAR ratio, and increases Sharpe ratio. "

It varies from portfolio to portfolio as to whether a profit target mechanism improves all these measures at once over and against no profit target.

However, what perhaps one should look for in a stepped parameter test, where no such increase in all these measures initially presents itself, is the risk adjusted returns using a finer filter such as std dev of daily returns.

Profit taking can dramaticallydecrease the standard deviation of daily returns. So you can increase the leverage on the less risky system which does take profits to give far better stats all round than on the system with no profit taking..........
Post Reply