Drawdowns

[theasxgorilla]

Niz,

One of the things I plan to implement into the system I am working on is to tighen my intial stop after a given MFE. This is a mechanisation of the trailing stop idea that tech/a talks about. How many discreationary traders would move their intial stop to breakeven (inc. brokerage) after an MFE of greater than 10%?

If you don't do this and the stock fails the breakout then your trailing ATR or MA or whatever else you are using as your primary exit, may never overtake your initial stop and you effectively give back a 20%+ on that trade (from the peak of the MFE, assuming an initial stop of 10%).

ASX.G

 

[nizar]

Niz,

One of the things I plan to implement into the system I am working on is to tighen my intial stop after a given MFE. This is a mechanisation of the trailing stop idea that tech/a talks about. How many discreationary traders would move their intial stop to breakeven (inc. brokerage) after an MFE of greater than 10%?

If you don't do this and the stock fails the breakout then your trailing ATR or MA or whatever else you are using as your primary exit, may never overtake your initial stop and you effectively give back a 20%+ on that trade (from the peak of the MFE, assuming an initial stop of 10%).

ASX.G

Hi ASX.G,

Thanks for your comments.

What does MFE mean?
I assume some sort of return?

Anyway, yeah it seems like a good idea to increase win% and potentially decrease longest string of losses, therefore minimising drawdowns.

Profit giveback is painful but for trend followers its part and parcel of the game. Curtis Faith speaks about is in his book about how the hardest part of trading were the exits. Sometimes you have to watch 100% profits halve before they go up again (or hit your exit).

And tech, what do you mean introduce a trailing stop after X periods?
Couldnt the trailing stop be the exit and this is in place from the very beginning.

 

[theasxgorilla]

What does MFE mean?
I assume some sort of return?

Maximum favourable excursion/maximum adverse excursion. The maximum favourable and adverse distance a trade moved from your entry price. I noticed in testing a system recently that a handful of trades managed an MFE in excess of 10% but ended up being stopped out by the initial stop. I don't know if having a mechanism which moves the initial stop up before the trailing stop overtakes it would also take out good trades. I haven't coded it yet...its the next phase of my project

 

[bingk6]

Niz,

If you don't do this and the stock fails the breakout then your trailing ATR or MA or whatever else you are using as your primary exit, may never overtake your initial stop and you effectively give back a 20%+ on that trade (from the peak of the MFE, assuming an initial stop of 10%).

ASX.G

It is for this reason that your stop line should be the greator of either the initial stop (which may be 10% below your purchase price and is therefore fixed) or the trailing MA or ATR stop. That way, if at the start of the trade, your trailing stop is too far away from the price action, your initial stop will provide you with the protection until such time as the trailing stop will have risen sufficiently to take over (hopefully).

 

[tech/a]

It is for this reason that your stop line should be the greator of either the initial stop (which may be 10% below your purchase price and is therefore fixed) or the trailing MA or ATR stop. That way, if at the start of the trade, your trailing stop is too far away from the price action, your initial stop will provide you with the protection until such time as the trailing stop will have risen sufficiently to take over (hopefully).

This is not going to help Drawdown.
Initial Capital loss possibly----but not peak to valley drawdown which is the issue here.

would also take out good trades.

Highly likely.

 

[theasxgorilla]

This is not going to help Drawdown.

Why don't you tell us the answer...what helps peak to valley draw downs? Is it only exits, the part where the trend bends?? Giving back something at the end of a trend is the inevitable part of a trend following system, but I don't think this is the only factor that contributes to the depth of a peak to valley drawdown.

The next market phase could be non-trending and volatile. False breakout territory. How do you prevent your system from digging itself further into the valley 1R at a time. Moving the stoploss to prevent entire 1R losses every time your system gets suckered by a false breakout might be one way (the beauty is I get to test it). Keeping your system out of an unfavourable market with a filter of some kind, like an index filter or a momentum filter for example, might be another way.

 

[tech/a]

ASX
Its a balance.
Your not going to get rid of drawdown you'll only keep it within acceptable limits.Then there is the possiblilty that your system and mine will trade in market conditions not seen in testing.
It is possible that the system will outperform in conditions more favourable and underperform ---even fail if unfavorable.

Is it only exits, the part where the trend bends?? Giving back something at the end of a trend is the inevitable part of a trend following system, but I don't think this is the only factor that contributes to the depth of a peak to valley drawdown.

Its the length (Time and number) and depth of these which determines the maximum drawdown and in the end un acceptable drawdown.
Limiting depth and number (consective) is the only way--I-- know to deminish it.
Thats one of the reasons I exited all long portfolio positions in 3 longterm portfolio's at 6170 XJO. This wasnt part of the system conditions!!!

The next market phase could be non-trending and volatile. False breakout territory. How do you prevent your system from digging itself further into the valley 1R at a time. Moving the stoploss to prevent entire 1R losses every time your system gets suckered by a false breakout might be one way (the beauty is I get to test it). Keeping your system out of an unfavourable market with a filter of some kind, like an index filter or a momentum filter for example, might be another way.

In a well performing portfolio the death by slow bleeding doesnt occur.
Testing has shown in a positive expectancy system that strings of losses and initial risk wont wipe it our.

Below is the current portfolio for the test portfolio of Tech Trader.
The "Initial stop in all cases is 10% of the initial purchase price.



Note stocks like QBE.
Initial stop or 1R 79c so far we have approx 25R in that one trade. Maximum string of losses was from memory 9 for the system.

I personally dont believe the answer or more to the point finessing relative to drawdowns--- is at this end (The buy end) of a system.

 

[yonnie]

I think it has to do with time frame.

the longer the time frame, the looser the stops will be and the bigger the potential drawdowns.

 

[theasxgorilla]

I personally dont believe the answer or more to the point finessing relative to drawdowns--- is at this end (The buy end) of a system.

It can seem this way, but I am discovering that system development is a bit like a rubiks cube, line up a few more of one color to see the side you have nearly completed go to bits.

To demonstrate this and as an example of how finessing down the entry end of a system can affect drawdowns I tried adding an entry volatility filter to a system I am playing with. Why? Because I noticed that on occassions (usually when tested on data that doesn't include the last 4.5 years of bull market, that should be a bit of a hint) my regular entry trigger was being drawn into stocks on bars with massive ATR. By adding a filter I managed to reduce the average DD from 23.7% to 21.4% and more importantly significantly bunch up the distribution from stdev of 7.24 down to 6. But my CAGR also fell from 18.35% to 17.3%, stdev was about the same. Hence my rubiks cube analogy....in this case it would seem that one can smooth the ride by tinkering with the entry at the likely compromise of some CAGR.

ASX.G

 

[stevo]

By adding a filter I managed to reduce the average DD from 23.7% to 21.4% and more importantly significantly bunch up the distribution from stdev of 7.24 down to 6. But my CAGR also fell from 18.35% to 17.3%, stdev was about the same. Hence my rubiks cube analogy....in this case it would seem that one can smooth the ride by tinkering with the entry at the likely compromise of some CAGR.
ASX.G

You are right, it is a rubiks cube! One side comes together and the other side crumbles. if developing a system was easy then I don't think that I would enjoy doing it as much.

If a trader wants less drawdown then one approach is to take less risk - as you found above. Less risk can be achieved, whilst not touching the system entry / exit criteria, by altering the position sizing strategy. Try 1% risk position sizing rather than 2% risk or take more smaller positions.

If a trader wanted to increase drawdown then use lots of leverage - it's very easy to have 100% drawdown with lots of leverage.

TradeSim looks only at closed trade drawdown and objectively that is the drawdown that is important, but emotionally the open trade drawdown (unless you are using lots of leverage) is the most difficult. Amibroker looks at open trade drawdown in it's standard report.

My drawdown won't be too healthy for August - probably double digits when I get around to balancing everything up!

regards

 

[nizar]

If a trader wants less drawdown then one approach is to take less risk - as you found above. Less risk can be achieved, whilst not touching the system entry / exit criteria, by altering the position sizing strategy. Try 1% risk position sizing rather than 2% risk or take more smaller positions.

I have personally found 1.5% risk to be a good trade-off.

Just a question for everybody.

When your portfolio is experiencing a period of drawdown, how do you know whether its just another drawdown or whether the system no longer works because market dynamics/conditions have changed?

Howard mentioned on the Robustness thread that he thinks trend following (or at least that of the Donchian channel style) worked well in the 1970s and 1980s but does not work well now. He included John W. Henry as an example of a renowned trend follower who has struggled thus far in 2007.

Couldnt it just be John W. Henry's funds are just going through a period of drawdown? Surely 8 months isnt enough to say a system has failed?

Michael Covel in his book goes through some of Henry's performances. Its happened a few times through the history of his fund that he was in the red until October, then still finished the year nicely as some huge trends took off in the last 2 months of the year.

Trend following, well at least is my understanding, isnt supposed to be a slow and steady method. Of course this doesnt mean that adding extra conditions to your system wont smoothen out the equity curve.

 

[tech/a]

I have personally found 1.5% risk to be a good trade-off.

Just a question for everybody.

When your portfolio is experiencing a period of drawdown, how do you know whether its just another drawdown or whether the system no longer works because market dynamics/conditions have changed?

More so the market dynamics are outside of the tested period long enough to effect the positive expectancy of the system.
How do you know.
The system trades outside the maximum or minimum parameters returned by montecarlo testing,which will adversely affect the performance of the system.
IE String of losses/Maximum Peak To Valley Drawdown/Win rate.
Well thats my opinion.

Howard mentioned on the Robustness thread that he thinks trend following (or at least that of the Donchian channel style) worked well in the 1970s and 1980s but does not work well now. He included John W. Henry as an example of a renowned trend follower who has struggled thus far in 2007.

Couldnt it just be John W. Henry's funds are just going through a period of drawdown? Surely 8 months isnt enough to say a system has failed?

Could be HIS system has failed---cant tell dont know it.

Michael Covel in his book goes through some of Henry's performances. Its happened a few times through the history of his fund that he was in the red until October, then still finished the year nicely as some huge trends took off in the last 2 months of the year.

Trend following, well at least is my understanding, isnt supposed to be a slow and steady method. Of course this doesnt mean that adding extra conditions to your system wont smoothen out the equity curve.

Hmm disagree.
Longterm equity curves get smoother the longer they run---generally.
Open equity Curves are smoother as trades run often for years---yet accumulated open profit can be quite smooth.
Tradesim will show you an open equity curve.

 

[wayneL]

Longterm equity curves get smoother the longer they run---generally.

While not disputing this, it seems counterintuitive to me.

Do you have evidence of this, and thoughts as to why this might be so?

 

[stevo]

When your portfolio is experiencing a period of drawdown, how do you know whether its just another drawdown or whether the system no longer works because market dynamics/conditions have changed?

Good question! I don't know the answer but do like the idea of a system turning itself off if the market conditions don't suit it.

Equity curve smoothness is quite subjective. Is there a reasonable metric that measures equity curve smoothness? Sharpe ratio is part way there but if a curve has the occasional upward spike the Sharpe ratio drops even though the system is ok. I probably should do a Google search on the topic.

Where would the challenge and excitement be if we had no drawdown.

stevo

 

[howardbandy]

Greetings --

Regarding John Henry, trend-following systems, and equity curve behavior.

John Henry had a great record and earned a lot of money for himself and his clients in his CTA accounts.

According to the president of the CTA firm that I worked for a few years ago, who was a friend of Henrys, Henry's technique was primarily Donchian-style breakout -- that is, buy new highs, sell new lows.

CTA firms are required to disclose their methodology to the public and to the regulatory agencies. Henry's disclosure documents also state that their systems are primarily trend-following.

Most of the CTA firms that used breakout-style systems have been doing poorly lately, not just Henry's.

Trend-following systems are designed to take every signal, losing on most trades (70% or so), winning on a few (30% or so), with small losers and big winners. The smoothness of the equity curve depends on having enough capital to trade many markets, and to have those trades be relatively uncorrelated. While the many commodity markets traded would appear to be uncorrelated, readers of Lowenstein (When Genius Failed) and Taleb (Fooled by Randomness and The Black Swan) will not be surprised to hear that the trades made are often very correlated.

There are statistical tests can that can be performed to determine whether a trade system is in a drawdown or broken. Two examples are a "runs" test that evaluates the number of wins and losses, and a "t" test that compares the mean of one sample (when the system is working) with another (recent performance) and gives a probability estimate of whether those two samples were drawn from the same distribution or not.

Thanks,
Howard
www.quantitativetradingsystems.com

 

[It's Snake Pliskin]

Greetings --
Most of the CTA firms that used breakout-style systems have been doing poorly lately, not just Henry's.

Trend-following systems are designed to take every signal, losing on most trades (70% or so), winning on a few (30% or so), with small losers and big winners. The smoothness of the equity curve depends on having enough capital to trade many markets, and to have those trades be relatively uncorrelated.

Thanks,
Howard
www.quantitativetradingsystems.com

Howard thanks for the posting.

Linda Rashke touched on the subject in one of her interviews I read.

 

[tech/a]

Greetings --


Trend-following systems are designed to take every signal, losing on most trades (70% or so), winning on a few (30% or so), with small losers and big winners.

Surely you not saying ALL trend following systems take or SHOULD take every signal?

If so I cant DISAGREE more.

The smoothness of the equity curve depends on having enough capital to trade many markets, and to have those trades be relatively uncorrelated.

I see this as a general statement that would give a smoother curve but not a necessity in generation of a smooth eqity curve.

While the many commodity markets traded would appear to be uncorrelated, readers of Lowenstein (When Genius Failed) and Taleb (Fooled by Randomness and The Black Swan) will not be surprised to hear that the trades made are often very correlated.

Cant disagree here.

There are statistical tests can that can be performed to determine whether a trade system is in a drawdown or broken. Two examples are a "runs" test that evaluates the number of wins and losses, and a "t" test that compares the mean of one sample (when the system is working) with another (recent performance) and gives a probability estimate of whether those two samples were drawn from the same distribution or not.

Are these tests in your book.
IE how to set them up and how to apply them?

Thanks,
Tech

 

[nizar]

Are these tests in your book.
IE how to set them up and how to apply them?

Id be keen to know as well.
Thanks for dropping by Howard.

 

[stevo]

Trend-following systems are designed to take every signal, losing on most trades (70% or so), winning on a few (30% or so), with small losers and big winners.

I think that this is dependent on the markets traded. I cannot take all the signals given when I scan 500 stocks using a long term system since I don't have that much money. If I was only following 20 stocks would definately be able to take all signals.

I am reading Way of the Turtle at the moment and it looks like it will offer some good insights in terms of drawdowns and trend following systems.

 

[Nick Radge]

Donchian style systems are too simplistic and is why they've failed over time. You need to go beyond simplistic. Anyone can bang out a breakout system in minutes these days. Some research on the way Campbell & Co have changed over time will lead to some great insights on how to improve the quality of a trend following system. I think you'll find that price and volume are now almost secondary inputs.

I should also say that many of these top fund managers being quoted look at risk adjusted returns and not necessarily absolute returns like many of us are attempting to achieve. This portends directly to the drawdown equation being discussed herein. A 25-yo who has the maturity and patience to look at a trend following method over the longer term should have greater tolerance for drawdowns than say someone who is older and more risk adverse.

Nick