Profit Target exits improve this system's performance

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.
sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Profit Target exits improve this system's performance

Post by sluggo » Tue Oct 05, 2010 2:48 pm

I received email from a friend who occasionally gets feisty and curmudgeonly when he goes off his meds. This time he boldly asserted "No trend-following mechanical systems are improved by profit target exits; PT exits always hurt performance. Why? I'll tell you why! It's because profit target exits violate the supreme religious axiom of trend following: Thou shalt cut thy losses and let thy profits run!"

Hmm? PT exits ALWAYS hurt performance? How about Anthony Gardner's tuned up Turtle system in Active Trader magazine (Feb. 2010) which gets a boost from profit targets? (link 1) , (link 2)

Or how about the system shown below (link to SOURCE CODE in the Blox Marketplace)

A prècis of profit target exits:
  • Having a position is riskier than having no position
  • Said the other way, having no position is less risky than having a position
  • Therefore, exiting a position removes risk
  • Similarly, exiting a part of a position removes some, but not all, risk
  • Therefore, exiting a part of a position when a Profit Target is hit, reduces risk
  • Therefore profit target exits reduce risk
  • This is especially beneficial when trading many simultaneous positions

The system shown here is a simple 3-moving average trend follower, with a couple of additional parameters that define the profit target exit. I ran it twice; once with the profit target exits enabled, and again with the profit target exits disabled. I adjusted the risk-per-trade until the Max-Drawdown numbers were exactly equal (red circles); this makes comparison of the other performance statistics very easy. Results were:

Code: Select all

Profit Target Exits     CAGR (%/yr)   MaxDD(%)  LongestDD   MAR ratio  Sharpe ratio
------------------------------------------------------------------------------------
Enabled                   43.00%       36.2%    7.8 months     1.19      1.41
Disabled                  32.57%       36.2%   11.7 months     0.90      1.36

So there you go, Tommy: a counterexample to your claim that no trendfollowing system benefits from profit targets. Put some garlic on that crow before you eat it.

Edit: formatted prècis using bullet-list
Attachments
eqcurve2.png
eqcurve2.png (31.75 KiB) Viewed 15697 times
eqcurve1.png
eqcurve1.png (31.5 KiB) Viewed 15697 times
params.png
params.png (31.13 KiB) Viewed 15697 times
Last edited by sluggo on Thu Oct 07, 2010 9:51 am, edited 1 time in total.

LeapFrog
Roundtable Knight
Roundtable Knight
Posts: 695
Joined: Mon May 17, 2004 4:18 pm
Location: Boston, MA

Post by LeapFrog » Tue Oct 05, 2010 5:53 pm

Should be enough proof to put Tommy back on his meds...

Jez Liberty
Roundtable Knight
Roundtable Knight
Posts: 123
Joined: Tue Nov 03, 2009 8:49 am
Location: London
Contact:

Post by Jez Liberty » Tue Oct 05, 2010 5:55 pm

sluggo,
First off - thanks for sharing this blox (on other thread) and study. Had some fun with it and it generated quite a few good thoughts...

I have to say I would intuitively lean towards your friend's Tommy' opinion (ie that one of Trend Following's main mantra is to let winners run and therefore breaking that fundamental rule might break the TF system as well).
Of course, we can always find exceptions to any rule (Never say never...)

If I may, isn't your comparison dependent on your choice of how to apply leverage to normalize both system returns (ie identical MaxDD)?
You chose to change position size in order for both systems to match on MaxDD and by doing so I feel that you are altering more than the "profit target" parameter in your test.

There's a very interesting chart (enlightening even) on the forum (that you posted actually: thanks!) which shows the evolution of various ratios plotted against position size: http://forum.tradingblox.com/viewtopic. ... 2169#32169
Basically, up to a point the MAR ratio keeps increasing along with position size (meaning for a given level of MaxDD, CAGR increases with position size).

As the position size for the profit target system is more than double the non-target one, it will inherently benefit from that increase in position size (and that's potentially the reason for its over-performance as opposed to the assumed inclusion of Profit Targets).

I am quite interested in that topic of profit targets, so I ran both systems at identical position sizes (1.25% - I had to raise it to get closer to 32% MaxDD as my own "kitchen sink" portfolio contains about a third of your number of instruments) to compare the results:

Code: Select all

WITH PTs:  CAGR = 20.70%, MaxDD = 32.2%, MAR = 0.64
W/OUT PTs: CAGR = 29.41%, MaxDD = 41.6%, MAR = 0.71
If one decides to use the concept of notional funding to normalize the results (or rather "reverse notional funding" in that case, to reduce the volatility of the "Without PTs" system to that of the PT one), the performances become:

Code: Select all

WITH PTs:  CAGR = 20.70%, MaxDD = 32.2%
W/OUT PTs: CAGR = 22.76%, MaxDD = 32.2%
The WITHOUT PTs system comes on top (as was shown by the higher MAR figure for that system).

I then decided to run the test for various position sizes (so easy with TB) and the chart seems to be a good illustration/parallel to the enlightening chart you posted: MAR increases along with position size up to the optimal point where it starts to drop (note that this 3 D chart only displays profit-taking at 1.6R: front border of the 3D "sheet" and no profit-taking, aka 9999R, at its back border - all points in between are just interpolations).

The last chart is a plot of the ratio between the MAR ratios of both systems (W/OUT TP MAR divided by WITH TP MAR) plotted against the position size: this is to show of the under/over-performance of one system evolves with the choice of position size

Funny thing is that I was just working on comparing position size leverage vs notional funding leverage as you posted this...
I think it makes a good point of showing that there are 2 options to increase/decrease leverage, which are dependent on the system's characteristics.
One of my approach for a system's money management is to "loosely" optimise the position size wrt MAR ratio and further leverage up and down using a "notional funding slider" (of course if you want to leverage up you also have to start taking into account interest on margin...

PS: apart from the position size and different portfolio, all other parameters were identical to sluggo's runs
Attachments
MAR-ratio.jpg
under/over performance between both systems measured by their respective MAR ratio
MAR-ratio.jpg (38.67 KiB) Viewed 15652 times
stepped-3D-chart.jpg
stepped parameter simulation: position size stepping + With/Without TP system
stepped-3D-chart.jpg (36.38 KiB) Viewed 15652 times

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Tue Oct 05, 2010 6:31 pm

Jez Liberty wrote:isn't your comparison dependent on your choice of how to apply leverage to normalize both system returns (ie identical MaxDD)?

As the position size for the profit target system is more than double the non-target one, it will inherently benefit from that increase in position size
Yes, that's the point actually: Profit Target exits reduce portfolio-wide risk. They reduce risk so dramatically that they allow you to significantly boost the size of your initial bets while suffering equal or less portfolio-wide pain. This lets you move further to the right on the (gain/pain) versus (betsize) curves, which increases (gain/pain) as you observe. And that's the goal: to increase (gain) without increasing (pain).

Comparing equity curves levered to have the same max drawdown is an idea by Modigliani* and Modigliani. They reason: if you were willing to tolerate 41.6% drawdown without profit targets, you're also willing to tolerate 41.6% drawdown with profit targets, and why not avail yourself of the improved (gain/pain) ratios this entails? I agree with them.

*1985 Nobel prize in Economics

LeviF
Roundtable Knight
Roundtable Knight
Posts: 1428
Joined: Mon Dec 22, 2003 12:24 pm
Location: Des Moines, IA
Contact:

Re: Profit Target exits improve this system's performance

Post by LeviF » Tue Oct 05, 2010 8:28 pm

sluggo wrote:It's because profit target exits violate the supreme religious axiom of trend following: Thou shalt cut thy losses and let thy profits run!"
If trends were frequently long and smooth, we wouldnt want profit targets. But they are not. At least not in recent history.

nodoodahs
Roundtable Knight
Roundtable Knight
Posts: 218
Joined: Wed Aug 09, 2006 4:01 pm

Post by nodoodahs » Tue Oct 05, 2010 9:43 pm

sluggo wrote:Comparing equity curves levered to have the same max drawdown is an idea by Modigliani* and Modigliani. They reason: if you were willing to tolerate 41.6% drawdown without profit targets, you're also willing to tolerate 41.6% drawdown with profit targets, and why not avail yourself of the improved (gain/pain) ratios this entails? I agree with them.

*1985 Nobel prize in Economics
I disagree with them.

DD is path-dependent. Do some MC sims on trade order or segments and you'll generate dozens of different max drawdowns, all of which will be less than your theoretical max DD and more than your theoretical min DD.

Now, normalizing to either a specified annual (or annualized monthly) standard deviation, or to a specified CAGR? Those I could get behind.

[Edit added for clarity: my issue isn't to the optimization for risk per se, it's to the use of DRAWDOWN as the risk optimization measure. As mentioned, I could also see an optimization to match CAGR and then measuring the smoothness of the equity growth ... ]

And seriously, the Nobel? Puh-leeze. So much crap has been awarded the Nobel that, as an "appeal to authority" argument, it's beyond worthless to me.

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Wed Oct 06, 2010 8:27 am

I take profits. To me it is meet and right so to do for the reasons outlined by Brother Sluggo above. So endeth my contribution. Although I would rather normalize for std dev than max DD.

nodoodahs
Roundtable Knight
Roundtable Knight
Posts: 218
Joined: Wed Aug 09, 2006 4:01 pm

Post by nodoodahs » Wed Oct 06, 2010 9:15 am

More clarity on my POV -

I don't have a dog in the fight re: profit targets. While I don't have targets; and while I haven't personally investigated whether a combination of setting targets + increasing R + optimizing leverage would work with my systems; I hold open the possibility that I might in the future set targets and do such an optimization.

My beef is purely with the use of optimizing for DD, as stated above. I think the other measures of risk (vol, downside vol, heck, even skewness and kurtosis) are much more robust than DD.

I also still think that optimizing for matching CAGR and comparing the risk measures would make an interesting mental exercise.

Jez Liberty
Roundtable Knight
Roundtable Knight
Posts: 123
Joined: Tue Nov 03, 2009 8:49 am
Location: London
Contact:

Post by Jez Liberty » Wed Oct 06, 2010 5:27 pm

After a good night's sleep, I realised that some of my comparison above is somewhat flawed.

The linear adjustment used to normalize both system performances (linear calculation of CAGR when setting MaxDD equals) is "theoritically interesting" but does not offer much practical value.
Indeed, "notional funding" does allow for increase of CAGR (ie trading a 100K notional account with a real account of 50K will effectively increase the CAGR) but in no case does the same apply for MaxDD.

I tried to normalize the two systems using a practical adjustment ("reverse notional funding" of the W/OUT TP system to reach a MaxDD figure of 32.2%) and the CAGR came to... a minuscule 0.37%!!

It does not make much sense (especially as I did not account for any sort of rebalancing during the 15 years of the test, and also this is path dependent) but it gives an idea of the flaw in my initial post... Apologies for this!

As for the pain measure (MaxDD vs. Vol or Std Dev), I have to say I do focus mostly on the MaxDD figure too but as they say it is mostly a matter of personal preference (the proverbial "as many bliss functions as there are different traders").

In any case, that paper on "drawdowns as useful statistical measure", written by Winton/David Harding might be interesting for some, so I am attaching it below
Attachments
Pros and Cons of Drawdown as a Statistical Measure.pdf
Winton paper on Drawdowns
(40.07 KiB) Downloaded 584 times
Last edited by Jez Liberty on Fri Oct 08, 2010 11:30 am, edited 1 time in total.

nodoodahs
Roundtable Knight
Roundtable Knight
Posts: 218
Joined: Wed Aug 09, 2006 4:01 pm

Post by nodoodahs » Wed Oct 06, 2010 5:37 pm

Jez Liberty wrote:"as many bliss functions as there are different traders"
Actually multiply that result by some non-prime integer and you'll get closer to the truth ...

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Thu Oct 07, 2010 10:38 am

... Profit Target exits reduce portfolio-wide risk. They reduce risk so dramatically that they allow you to significantly boost the size of your initial bets while suffering equal or less portfolio-wide pain.
Students of Ralph Vince's book The Leverage Space Trading Model will recognize the above as nothing more than a straightforward application of his "Fundamental Equation of Trading" found on pages 50-54. Well-crafted profit target exits reduce the second term more than they reduce the first, thus Total Wealth Relative increases. (Eureka!) Chopping off the fat tails of the returns distribution can be very beneficial. Even chopping off the tail on the right hand side, the tail with the outlier positive returns, can be beneficial. IF you reduce your standard deviation more than you reduce your average returns as RV is careful to state.

Students of Ralph Vince's earlier books will note that he's been saying this for twenty years (cf. Portfolio Management Formulas, C 1990, p.245)
Attachments
combo.jpg
Excerpt from Ralph Vince, "The Leverage Space Portfolio Model", p.53
combo.jpg (94.4 KiB) Viewed 15417 times

kianti
Roundtable Knight
Roundtable Knight
Posts: 335
Joined: Fri May 02, 2003 6:10 am
Location: Florence - Italy

Post by kianti » Fri Oct 08, 2010 4:04 pm

sluggo wrote: ...Eureka!
Attachments
pitagora.png
pitagora.png (47.44 KiB) Viewed 15097 times
Last edited by kianti on Mon Oct 11, 2010 4:28 am, edited 1 time in total.

tovetski
Senior Member
Senior Member
Posts: 46
Joined: Fri Jan 13, 2006 10:49 am
Location: Montréal, Québec, Canada

TMA Blox with multiple profit targets

Post by tovetski » Sat Oct 09, 2010 10:12 am

Hi all,

I added 2 blox in the blox marketplace that builds on sluggo's TMA with Profit Target. I achieve similar results with a portfolio of 134 markets. These blox allow for many profit targets to be tested (maximum of 5 for now). They also incorporate a volume cap.

Below are 2 samples. One without and one with a profit target similar to sluggo's.

Have fun and let me know how it works for you.
Attachments
TMA_0_0.png
No Profit Targets
TMA_0_0.png (29.53 KiB) Viewed 15215 times
TMA_10_1.png
1 Profit Target of 10 ATRs
TMA_10_1.png (29.37 KiB) Viewed 15215 times

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Sat Oct 09, 2010 2:53 pm

Pierre, thanks for sharing your work! It looks beautiful.

LeviF
Roundtable Knight
Roundtable Knight
Posts: 1428
Joined: Mon Dec 22, 2003 12:24 pm
Location: Des Moines, IA
Contact:

Post by LeviF » Wed Jan 05, 2011 9:58 am

I was experimenting with my profit targets last night. For the last couple of years, my system was programmed to take off 1/3 of the position at 1R, another 1/3 at 2R, and let the last 1/3 ride. This smoothed the equity curve and improved my performance stats vs. no profit targets.

After more tinkering, i'm learning that taking off even more (ie - most) of my position at 1R is even better. It allows me to increase my bet size and get better risk adjusted returns. But I dont like it from a fundamental viewpoint. Everything I've been taught says I'm supposed to embrace and profit from the multi-year trends and not cut profits short... :cry:

AFJ Garner
Roundtable Knight
Roundtable Knight
Posts: 2040
Joined: Fri Apr 25, 2003 3:33 pm
Location: London
Contact:

Post by AFJ Garner » Wed Jan 05, 2011 10:19 am

One way of looking at it is that you are diversifying your systems rather than cutting trades short. This would be my preferred view. Imagine that you trade one very long term system - perhaps trades are entered all in one go, perhaps entry is staged.

By taking profits at successive target levels you are in effect trading a number of different systems capturing different length trends. Some of your trades will partially exit after a relatively short period but the rump of the trade may go on for a year or two.

The rump represents your LTTF system. Earlier profitable exits represent your short and medium term TF systems.

LeviF
Roundtable Knight
Roundtable Knight
Posts: 1428
Joined: Mon Dec 22, 2003 12:24 pm
Location: Des Moines, IA
Contact:

Post by LeviF » Wed Jan 05, 2011 10:32 am

That was the intended purpose of my 1/3, 1/3, 1/3 system.

But now it appears that getting out quicker is better so that the exposure to the VLTTF portion is minimized.

sluggo
Roundtable Knight
Roundtable Knight
Posts: 2986
Joined: Fri Jun 11, 2004 2:50 pm

Post by sluggo » Wed Jan 05, 2011 10:35 am

LeviF wrote:Everything I've been taught says I'm supposed to embrace and profit from the multi-year trends and not cut profits short...
Your are beginning to suspect: Everything you've been taught is wrong. And your suspicions are correct.

That's because your teachers are old-school thinkers whose worldview is limited by their old-school analysis methods and their old-school software tools, notably Tradestation. These people confine their thinking to { one system, which trades one instrument, with constant position size } , and they analyze in terms of profit per trade and so forth. They have no conception of multi instrument portfolios traded simultaneously, and they cannot imagine multi system SUITES traded simultaneously.

And as AFJG has just said, partial exits (at profit targets) are exactly equivalent to multi system suites. Which your teachers never even imagined. What they did not imagine (and could not simulate), they did not analyze.

Fortunately you have two things your teachers did not: (1) an open mind; (2) more modern backtesting tools. To paraphrase Isaac Newton, you can see farther than they did, because you are standing upon their shoulders.

Have a look at the vendor system "Aberration Plus" as a source of new ideas and stimulating thoughts. AP scales out "X percent of the remaining position" at each profit target. For example, if X% = 40%, and the initial position is 125 lots, then:
  • At the first profit target, exit (40% of 125) = 50 lots, remaining position is 75 lots
  • At the second profit target, exit (40% of 75) = 30 lots, remaining position is 45 lots
  • At the third profit target, exit (40% of 45) = 18 lots, remaining position is 27 lots
  • etc
Aberration Plus attempted to place its profit target levels such that half of profitable trades hit the first profit target, (1/4th) of profitable trades hit the first two profit targets, (1/8th) of profitable trades hit the first three profit targets, (1/16th) of profitable trades hit the first four profit targets, etc. I find this an extremely interesting idea. Whether or not it is desirable, acceptable, or optimal ... is for you to decide. Sweeney's MFE rears up its head again!

LeviF
Roundtable Knight
Roundtable Knight
Posts: 1428
Joined: Mon Dec 22, 2003 12:24 pm
Location: Des Moines, IA
Contact:

Post by LeviF » Wed Jan 05, 2011 11:02 am

For me, the conflict is that getting out of 99% of the position at 1R (leaving 1% in so the system doesnt jump back in the trade) is better than letting it ride. Again, this goes against all the theory of why LTTF works: returns arent normally distributed - you need to capture the 20+R fat tails, most people arent disciplined enough to cut losers and run winners or deal with lots of losing trades, yada yada. So then what is my edge?...

drm7
Roundtable Fellow
Roundtable Fellow
Posts: 96
Joined: Sun Apr 20, 2003 9:02 pm
Location: Richmond, VA

Post by drm7 » Wed Jan 05, 2011 11:26 am

Great thread on a great forum!

Levi,

While it may be tedious to do so with 20 different markets, take a look at the overarching "regime" of the backtest period by staring at the appropriate charts. The multi-year blockbuster trends of the late 70's and early 80's may have shortened up a bit or have become "steeper" and/or more volatile.

Looking at the mechanics of MA crossover systems, you want a nice shallow "landing" at the end of your trend, to give your stops time to catch up. (An L-shaped downtrend vs. a "V"). Profit targets catch the point of the "V", but only get you part of the way down the "L".

I don't know for sure, just thinking out loud here.

Post Reply