Systematic Trading research and development, with a flavour of Trend Following
Au.Tra.Sy blog – Automated trading System header image 2

Trend Following, “Monkey Style”

March 8th, 2011 · 43 Comments · Strategies


A while ago, I used a quote from Winton manager and trend Follower David Harding (found in this interview) saying:

If you put in stops and run your profits and trade randomly you make money; and if you put in targets and no stops, and you trade randomly you lose money. So the old saw about cutting losses and running profits has some truth to it.

The quote was used to illustrate a post stating that a large driver of Trend Following returns is based on the mechanics of those systems (“cut your losses short, let your winners run“) which therefore benefit from the right tail of market return distributions – which are “fatter” than the usually assumed normal distribution – and avoid the left tail.

“Trade randomly”? Like the proverbial dart-throwing monkey? It seems so…
In effect, Harding is saying that entry points do not matter so much: a random entry coupled with a smart exit strategy would make money.

Random Trading To the Test

I once met with a fund manager, who described his strategy as very similar to that random system in the Harding quote. What was really important to them was the position sizing for each new signal, as well as the exit strategy. The entry signal direction was “irrelevant”.

I found this puzzling at the time and have been wanting to test this idea since then, to verify whether a “random trading” system could indeed be profitable.

The system tested here is composed of random entries with additional “classic” components: a volatility-based fixed fractional money management and volatility-based trailing stop exits.

  • The system first “tosses a coin” to decide whether to go long or short the market.
  • An initial stop is set below/above the entry price at a distance equal to a fixed multiple of the volatility measure.
  • That entry-stop distance is used to calculate the position size, so that the risk per trade (amount lost if trade gets stopped out) is equal to the fixed percentage of account equity.
  • Every day, the trailing stop is adjused so that it is never further than the fixed multiple of the volatility measure. The stop always gets closer to the market and never gets adjusted further away from the market (i.e. if the market turns back toward the stop, the stop level does not change).
  • When the position hits the trailing stop level, it gets closed and a new position is open. The direction of that new position is again determined by a new coin-toss.

Test Parameters and Results

For this test, I used fairly standard parameter values:

  • Volatility Measure: 39-day (exponential) ATR
  • Stop Distance: 2 ATR
  • Risk per Trade: 1% of Account Equity

The portfolio used for this test is a subset of the one used in the State of Trend Following report, basically all those instruments that I have data for going back to the start of the test: in January 1990 (click for the exact list).

Since this is a random experiment, I generated multiple test outputs (200), all based on the same parameters, and averaged their monthly returns to create a composite equity curve, which performance summary statistics can be seen below:

Performance Stats
CAGR
18.11%
Max DD
33.57%
MAR 0.54
Monthly Std Dev
6.34%
Average Monthly Rtn
1.59%

 
The 2-ATR stop level is somehow an arbitrary choice and I wanted to check whether this bore an impact on the test results.

I ran a further test, stepping the ATR-multiple for stop calculation from 2 to 10. Each ATR-multiple set was run 200 times again and averaged to give a composite equity curve.
Normalizing these 9 composite equity curves (for equal monthly standard deviation) and averaging them produced a “super-composite” equity curve composed of 2000 random tests (equally split between ATR-multiples ranging from 2 to 10).

The performance summary statistics of this “super-random-composite” equity curve are below:

Performance Stats
CAGR
16.46%
Max DD
21.87%
MAR 0.75
Monthly Std Dev
5.67%
Average Monthly Rtn
1.44%

 
Note how the diversification and rebalancing over several ATR-multiple stop levels have a substantial impact on the Max Drawdown and volatility.

Both equity curves are charted below:

All in all, not too bad for “monkey-style” trading! It goes to show that signal entries, which most beginning traders/system developers focus so much on, are not so important after all…

Update: follow-up post tackling other aspects of randomness in trading systems and clarifying subjects such as averaging and commissions/slippage: Further Musings on Randomness
 
 
Credits/Additional Reading: The concept of random entries with trailing stops has actually been discussed before. It seems like it was introduced by Van Tharp in his Trade your Way to Financial Freedom book, and mentioned on this article by Chuck Le Beau, where he expands on the concept of “Chandelier Exit” (name for volatility-based trailing stops).
Thanks and credits also to user “sluggo” on the Trading Blox forum, who published a similar study four years ago, and some code which I reused most of for this study. Note that his study found an opposite result, showing a turn in profitability (downwards) of random systems after 1997 (portfolio and parameter values are different though), so you might want to run your own test to verify this concept for yourself…
 

Picture credits: Trevira via flickr (CC)
Related Posts with Thumbnails

Tags:

43 Comments so far ↓

  • docdan

    Great post. Thanks for your effort.

    One question. Is the ATR value used for the trailing stop fixed at the time of entry or updated while the position is open?

  • Jez Liberty

    docdan – The ATR value is re-calculated every day

  • Pretorian

    Excellent post as usual. I get exactly the same feeling about entries, they explain a very small part of a TF system profitability. I think that trading with the trend is more than enough. I understood that when I read Mandelbrot.

    Exits, position sizing and (most importantly) learning to deal with DD is what counts. I have had the pleasure to speak with a few of the Market Wizards, they have given me some very useful advice….absolute none about entries though. It is also curious to know that in the seminars I have attended, entries are among the first and most asked questions.

    Thanks

  • Ali

    Indeed great post as usual. I came to the same conclusion during one of my back testing analysis. I fixed the entry point and tested several exit strategies. Then I fixed the exit strategy; and tested several entry points.

    The excellent exit strategy will always result into a profitable system. Unlike poor exit; it will mostly destroy any entry point strategy.

    By the way; MFE and the MAE are a very good tool to find/modify the system stop or exit point.

    Thanks

  • Mirec

    What were the parameters for trailing stop? Thanx.

  • Severus

    Thanks for the post. Did your system consider pyramiding entries ? That might change results a bit though I guess the overall takeaway will remain the same

  • Rick

    Hello,

    This is interesting but not enough information is provided to come to a conclusion. I would like that you kindly provide the following:

    (1) win rate
    (2) commission rate
    (3) Buy&hold performance for the combined set

    I highly doubt it that after inclusion of commission and slippage this system can beat buy & hold.

    Thanks

  • Jing

    I can’t fully agree exit is more important than entry. Take simple channel breakout for example. If you fix the breakout days, and change exit days to 15 days, 20 days, or 30 days, you will see the performance difference is not that dramatic. But if you change the breakout days from 20 days, 50 days, 100 days, and fix the exit days, you will see huge performance difference. Your article only show random entry can have profit. But it can’t show exit is more important than entry. In my opinion, if want to design an excellent system, the entry is the key, just use simple 20-days low as exit and some exit filter (high volatility exit), this simple exit is better than ATR trailing stop most of the time at least from my testing because it does a better job to make profit fly.

  • Jez Liberty

    @Mirec – trailing stops are based on 2-ATR for the first study and a range of 2 to 10 ATR for the composite curve

    @Rick
    I did not compute the exact Win rate as I only output the equity curves, due to the high number of tests and trades. However, I just re-ran a dozen random tests and all of them show a Win rate of roughly 36% – quite typical of a Trend Following (obviously the Win/Loss ratio is > 1).

    Simulation was friction-less (no commission, no slippage)

    Not sure if Buy-and-Hold is the best benchmark for this strategy since it goes both long and short, but it would be worth the comparison. I’ll try and see what I can come up with.

    @Jing, Agree that entries are not completely irrelevant. The goal of the post was not to show that exits were more or less important than entries. It was merely to illustrate Harding’s argument with a test, and yes, it seems random entries can be profitable when adding stops and no profit targets…
    But I do not think that entries are “the key”. As Pretorian was saying, not many Market Wizards give advice regarding entries, but rather on position sizing and money management, which – to me – also includes letting profits run (as your 20-day exit would).

  • Jing

    Yes, I fully agree position sizing, money management and make profit fly is more important because these concept is the most “robust”. For entries, you can optimize it to fit historical curve, but no one can guarantee this entry will work excellent in future, so excellent entry may be susceptible to curve fitting. But position sizing, money management and make profit fly are more robust and universal, not matter how markets change, these concepts are very robust. After 20 years, a previous excellent entry rule may not work well but money management and make profit fly always work.

  • Jez Liberty

    @Rick
    I ran a quick test with simple Buy-and-Hold of the same 22 instruments.
    I normalized the results with the curves in the post for equal monthly returns standard deviation and the CAGR is 5.29% with a MaxDD of 80.34%.
    You can see the comparative chart there.

    Of course the comparison is not 100% realistic as the strategies in the post are tested without friction (whereas a real-life strategy like this would be subject to commissions and slippage unlike a simple Buy and Hold).

  • Mirec

    One more question. So actually you have just one stop? (the initial stoploss is the same as the trailing stop?).
    Isnt the standart ATR 14 days not 39?

  • Rick

    Hello,

    Is it so difficult to include commissions in your test? If you can backtest, it should be easy to include a fixed commission rate. Most backtesters have the option already, no need to do anything.

    There is no way for us to know how realistic are these tests unless there is commission included.

    How many trades were there on the average per run?

    Thanks

  • Ali

    Totally agree that the money management is the key….It is funny concept but indeed the money management is the CEO of multiple trading system. So let’s make sure the best CEO is always on our side ;)

    But going back to entries and exit strategies. In the post; random entry represented multiple entry strategies (Note that different strategies not changing only parameters) while fixing the exit point. And indeed the result was a profitable system.

    Now let’s consider the other way around; testing fixed entry (even basket of proven entry point’s strategies like in the state of trend following report) and work out random exit strategies. Meaning selling or holding randomly! The system will just result into losses for sure, which confirms the point of this post.

    Trend following is based on the concept of letting profit run and cutting losses short; and the only way to do that is by controlling the exit not the entry!

    But definitely; I would give importance to both entry and exit (in live trading) to give me as much as possible of positive mathematically expectancy. And then design a money management system that can give me as much as possible of geometric mean of return.

  • Michael Harris

    Hi Jez,

    I hope everything is fine with your website now.

    Your analysis is impressive but it only tells the average performance of “monkey style” trading for a large number of instruments. It does not tell what the percentage of traders is who will underperform this average and more importantly the percentage of those who will fail. This is because, what you calculated is not the performance of a specific system but an average of many systems trading many futures contracts.

    I have seen similar discussions in several forums recently and I have prepared an analysis to illustrate the issues involved:

    http://www.priceactionlab.com/Blog/2011/03/time-to-hire-a-monkey-not-really/

    All the best.

  • Jez Liberty

    @Michael,
    Apart from a few teething issues, website seems to be running fine now – thanks.

    @Mirec: 14-day is a standard value indeed. Note that the exponential moving average smoothing constant in Trading Blox is calculated in a different way from the more specific Wilder ATR moving average such that:
    TB/classic EMA Days = (Wilder Days x 2) – 1
    Wilder Days = (Normal EMA Days + 1) / 2
    So, a 39-day TB ATR is equal to a 20-day Wilder ATR, which is a fairly standard value also.

    @Rick: not difficult at all to include friction costs (with Trading Blox anyway). This (omission) is actually a choice to avoid making assumptions on commissions and slippage amounts, which is variable for every trader based on size, broker, market, etc. (and which can definitely have a non-negligible impact as previously discussed in here and here).
    So, I usually prefer to leave this out of the equation (at the risk of making the “raw” results less realistic) and let the readers interpret the results based on their friction costs assumptions. I agree that not much information was provided to do this, though. Number of trades and round-turns per million would be a good starter and I’ll check these for the test above.
    For this specific study though, I simply wanted to test whether random entries + money management with volatility-based sizing and trailing stops could exhibit a positive tendency (which they do), as opposed to a fully random system, which should exhibit random results (i.e. neutral tendency of it goes long as frequently as it goes short). This does not mean that the system is profitable/tradeable in real-life though
    I was actually thinking of writing a follow-up post that will address some of these issues. Stay tuned…

    @Ali, random exits, I’ll probably test in the follow-up post.

    @Michael: I purposely ran many tests to calculate an average to detect the tendency of such systems. The performance of an individual system/trader is not statistically representative of the underlying process at play. Same concept goes for the use of multiple instruments – as I was saying in a diversification post:

    Every trade/instrument can be seen as a particle composed of a (large) random element and a smaller edge that we try to extract via a mechanical system.

    This is the way I see diversification: by adding a large number of mostly random elements, you can ensure that random moves have some cancelling effect on each other. All that is left is to collect the small edge from all the instruments via your preferred trading strategy(ies).

    I do think that entries have their importance in system design but I would tend to believe – As David Harding does – that they have less impact on overall system returns, in the context of Trend Following systems at least – something I’ll try to test out in a follow-up post.

  • MArco

    Hi Jez, excellent post. Thanks for posting it.
    Since you posted the average returns of the multiple tests in both cases, I was wondering if you could give some numbers on what were the worst cases, and best cases, or a chart with all the various performances of each run, for each case, in order to see how consistent is the return over time with a single test and an average. Alternatively, just to understand what is the worst and best drawdown and returns you found in your tests would be also useful.
    Thanks again for your work!

  • Michael Harris

    Hi Jez,

    Your wrote:

    “I do think that entries have their importance in system design but I would tend to believe – As David Harding does – that they have less impact on overall system returns, in the context of Trend Following systems at least – something I’ll try to test out in a follow-up post.”

    This is a very interesting statement but it needs to be proven, I think you have already agreed to that. My point is that, IMO of course, this cannot be proved by diversification and/or random entries because these are “many systems methods” and performance results are averages.

    I also think, as others have already commented, that you should include commissions in your studies. You can use IB commission levels as representative. This is important because as I have demonstrated recently in a response to an article by Bespoke Group about the performance of a simple system based on buying on the close – selling on the open of next day, for SPY, total ruin is possible due to commissions only, although studies without commissions show spectacular returns. You can find the study here: http://bit.ly/ib69sQ

    I am looking forward to more of your tests regarding trend-following entries.

  • Jez Liberty

    Michael,
    I do not understand your point about averaging. To me, averaging a random process is absolutely necessary to squeeze out the randomness and detect the central tendency (and other aspects such as dispersion, etc.).
    Without averaging many random runs, but instead relying on one random run, there is no way to understand how much randomness played a part in the results achieved, just by looking at the results.

    This is why I think it is vital to average when running random tests.
    If you think about it, this is exactly the same concept as when you run a test of a strategy on one single instrument: you still want to generate many trades for that market-system in order to detect a tendency and not be mis-led by the (mostly random) outcome of one trade only. A back-test on many trades will allow to determine the tendency of that market-system.

    I do not see the difference with running many back-tests and averaging their results to detect the tendency of the strategy, ie instead of “averaging” the partly-random outcomes of many trades on one instrument, I “average” the partly-random outcomes of many systems (which happen to be trading multiple instruments because this is the style of trading I want to test – but I do not see how this impacts the overall concept of “averaging a large number of random test”).

  • Michael Harris

    Hi Jez,

    I didn’t say that averaging is not useful. I just said it may be misleading in some context. I will give an example:

    Let us say that in a possible parallel universe there are only 4 fund managers with the following average yearly returns:

    Manager 1: 20%
    Manager 2: -1%
    Manager 3: – 2%
    Manager 4: -3%

    What is the meaning of saying that the average yearly return of managers in that world is 3.5%? Even if we state the standard deviation of 11.03?

    A more informative statement would be to say that 75% of advisors, or 1 out of 4 have failed.

    Now as far as my specific comment maybe you didn’t read my blog ( http://bit.ly/goBvuF).

    Along with the good tests you did I believe you should also calculate the number of systems that fail on the average because when you use random entries, you get many possible systems based on the tossing sequences obtained. Should we assume that the returns are normally distributed? I have not seen such study. What about if the distribution is skewed and a large percentage of those systems fail? Maybe systems that were lucky enough to have the right tossing sequence were profitable. That depends on the size of trades and of stop-losses.

    I do not know the answers. This is why I asked. I did not question the use of averages and standard deviations. I question the shape of the distribution of the returns of the many possible systems the tossing a coin generate for the specific data you used.

    Regards.

  • Pumpernickel

    The same burst of activity at the Trading Blox Forum 4 years ago, which created the code you ran for this blog entry, actually tested THREE hypotheses:

    H1. (Random entries) + (trailing stop exits) are profitable, even after $100/contract slippage and commission are charged.

    H2. (Trend following entries) + (random exits) are profitable after C&S are charged.

    H3. (random entry timing) + (random exit timing) but not random direction-of-trade, is profitable after C&S. On randomly chosen entry date, enter in the direction of a classical trend detection indicator such as MACD etc. Then exit at a random exit date. This is profitable.

    All three of them were found profitable, even with $100/contract C&S, when the trend following parameters / trailing stop parameters are set to very long term trading, and the stop widths are commensurately wide.

    These three test results suggest some conclusions:

    C1. “Good entries” are not required for profitability

    C2. “Good exits” are not required for profitability

    C3. Neither “good entries” nor “good exits” are required for profitability

    By the way, some of the charts of these results on the Blox Forum are awesome! http://bit.ly/i5LTNH ; http://bit.ly/dLMD2y

  • Rick

    Pumpernickel ,

    In the first link you gave only the exists are random. The entries are rule-based.

    In the second link, they are talking about a trade direction filter.

    Unless I missed something, there are no cases of random entry.

    None of your 4 conclusions appear to be derivatives of those links.

  • Jez Liberty

    @Pumpernickel: Thanks for the pointer on these TB threads. Seems I only caught the first instalment of what became a “randomness mini-series” on the TB forum (I had not seen these additional 2 posts, only an earlier one, which I linked to in the Credits section of the post, and where only random entries where discussed.

    Shame I did not find it earlier though, as the number of comments and questions on this post pushed me to prepare a follow-up post with more random logic testing, and that study/code could have been useful to look at/reuse the code. Well, the code for these other random systems was not too difficult to write and it is good in a way though, as this additional testing will not have been impacted by the hindsight knowledge of the results from my predecessors.

    My next tests are also slightly different (#1: random entries and profit targets, #2: random entries (direction) and random exits, #3: MA cross-over entries and random exits, #4: MA cross-over entries and target profits), so hopefully they will complement the findings on these TB threads.
    The one that seems to overlap is MA entries with random exits, but in my test the return is so small compared to the volatility that I cannot see much statistical significance in the results.
    The other difference is that I did not consider any commission/slippage, as discussed in comments above…

    @Rick: check the other thread (linked to in credits section of the post) for the study by sluggo on random entries.

  • Pumpernickel

    Rick, there were three conclusions (not 4 as you typed). Among the many useful Blox Forum posts on the topic, here are three specific messages by 3 different authors, that contain test results which support the three conclusions above

    C1: http://bit.ly/hS0ZcW (White Cube, 9/16/07)

    C2: http://bit.ly/fzPIOW (sluggo, 3/26/07)

    C3: http://bit.ly/dThD1i (svquant, 3/26/07)

    Please take a few minutes to surf around the Forum, looking at old threads and old topics. There is quite a bit of great stuff there, waiting to be (re)discovered. Heck, you may even decide to join.

    Note that H3/C3 use random entry *timing* and random exit *timing*, but on the randomly chosen entry date, they always trade in the *direction* of the (long term) trend. They use a bullish-or-bearish indicator to suggest whether it’s best to enter long or enter short on that date.

  • Rick

    Pumpernickel ,

    C1: One guy claims the system lost money 100% of the time and another claims it made money 100% of the time but CAGR is very small.

    I think the positive bias some people got may be attributed to several factors, including not properly rolling contracts.

    In C1 I do not see any conclusive evidence of random-entry/ATR exit making money. I see conflicting reports and low returns.

    In C2, it is claimed that pure random entry exits were profitable only 2 times out of 50. They used an ADX burst and other entry methods to fix that.

    In C3, I think the claims are wired. The input data are first filtered using moving averages.

    As a conclusion, I don’t see where you got your conclusions. These posts prove nothing, save the fact that in most cases exhibit a wrong approach. I also sense some desperate attempts to prove something is working the reason for which are not very clear.

    Neither C1 – C2 nor the posts by Jez have proved that random entry/ATR exit is a profitable trend-following system. Jez did not use commissions and never stated how many runs are unprofitable out of the 200 he ran.

  • Money management: playing the equity curve « Neural Sniffer

    [...] saw about cutting losses and running profits has some truth to it. “ See the details here: http://www.automated-trading-system.com/trend-following-monkey-style/ And here: [...]

  • Alvantage

    Hello Jez, Good study. The take away is valid.

    However, I would like to emphasize one defect in the study as already pointed out by Michael.

    Your choice of averaging monthly returns over multiple random realizations actually BIAS the volatility (and hence the drawdown statistic) DOWNWARD significantly. This is just simple statistical artifact.

    As an illustration, I performed a simple simulation. Let’s say we have a strategy with monthly return which TRULY follows Normal distribution with mean 1.5% and standard deviation of 10%. We sample it for 100 months. Then we repeat the sampling 200 times. What you did basically was to average the each month return based on 200 sampling. This gives 100 averaged monthly returns. This averaged return, will still have the same mean value of the strategy. However, the volatility is greatly reduced by the averaging process. Consequently, the drawdown figure is reduced also.

    The picture shown in the following link highlight my point. The black bold line is the resulting equity curve from the averaged monthly return.

    http://img708.imageshack.us/i/effectofaveragingonequi.png/

    If you are into R programming, you can replicate the simple study with the following codes:

    ===========================
    n = 200
    dx = c()
    for(i in 1:n){
    x = rnorm(100,0.015,0.1)
    dx = rbind(dx,x)
    }

    dx.ave = apply(dx,2,mean)
    x.cum = apply(dx,1,cumsum)
    color = rainbow(n)
    plot(cumsum(dx.ave),type=”l”,main=”Random Equity curve”,ylab=””,xlab=”counter”,ylim=c(min(c(x.cum)),max(c(x.cum))),lwd=3)
    for(i in 1:n){
    lines(x.cum[,i],col=color[i])
    }
    lines(cumsum(dx.ave),lwd=3)

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

    Indeed, for the averaged monthly return, the sample average is 1.509% and the volatility is greatly reduced to 0.77%.

    ================
    mean(dx.ave)
    sd(dx.ave)
    ================

    Of course, the strategy you are testing do not produce monthly returns that is normally distributed. However, the impact of averaging randomly produced monthly returns on BIASING the drawdown downward is still valid.

    Cheers.
    Bryan.

  • Alvantage

    In light of that issue, I think the Performance Stats which remain valid are CAGR and Average Monthly Rtn.

    For the risks measure, I suggest you calculate the average of maximum drawdowns of each sampling. It will also be useful to get the distribution of the drawdowns, e.g. 10% percentile, 25% percentile, median, etc.

  • Zerks87

    Hi Jez,

    I wanted to clarify something about bet sizing technique.

    When I came across this post I went back to my copy of the complete turtle trader and looked up the mechanics of this technique.

    When you are using this ATR (or “N”) technique on futures, as I understand it, you are taking the cost of buying the future on a per point basis and multiplying it by the number of points that make up the ATR.

    Then you multiply that number by the size you want your stop loss (let’s say 2ATR’s) to be and then divide by the amount by % of your total portfolio that you want to risk.

    But what if you wanted to use this strategy on a stock instead?

    Would you just divide the amount of equity you are willing to risk (say 1%) by the 2ATR stop number or would it require some other calculation?

    Great post – look forward to hearing from you.

    Best,
    Matt

  • Jez Liberty

    Hi Zerks87,
    In essence, this is how you would do it to “translate” the rule directly to stocks instead of futures.

    Coincidentally, a post was created on the TB forum a few days ago, discussing a very similar issue (maybe started by you under a different avatar):
    http://www.tradingblox.com/forum/viewtopic.php?t=8768&highlight=
    which highlights potential problems linked to the fact that stocks do not have embedded leverage and this position sizing technique “could” result in total allocation going over the available equity.
    -Jez

  • stefan

    Making averages could be misleading. Sometimes it is best to visualize all 200-500 colored equity curves in one picture to see the robustness. Do you have something of this sort.

  • Jez Liberty

    Agree, averages could be the tree that hides the forest (although I really wanted to test for a central tendency here) – unfortunately I do not have these multiple equity curves available – sorry.

  • Paul

    Hi Jez,
    It appears that the trailing stop distance of 2 x ATR was arbitary. The comparison against the mix of stop distances shows that 2 x ATR is better than a mix between 2 and 10 x ATR. However, it doesn’t show whether 2 x ATR is optimal.
    Has anyone run the same test with suitable sample sizes to compare the performance of 1 x ATR vs 2 x ATR vs 3 x ATR etc to try to pinpoint the optimal trailing stop distance?
    Thanks,
    Paul

  • Jez Liberty

    Paul,
    Not that I am aware of, but that would be a good idea for a next post! I did do something similar with Profit Targets a while back though: http://www.automated-trading-system.com/profit-targets-trend-following/
    Cheers,
    Jez

  • Juston

    Hey Jez,

    I was wondering if you would be able to supply a simple bell curve of CAGR and MAXDD. I would like to see where the range and majority of the CAGR and MAXDD lies.

    Thanks,
    Juston

  • Jez Liberty

    Juston,

    The methodology I used for this test was actually to average all the random simulation iterations on a monthly basis (ie to simulate re-balancing) instead of running all simulations on their own and averaging their end results. As such, I do not have such CAGR/MaxDD data… But if you are interested you might want to check that post: http://www.automated-trading-system.com/trading-diversification-free-lunch/
    which shows the spread of CAGR/MaxDD for another randomized test (to do with portfolio selection that one).

  • Stuart

    Hi, I am having trouble working out how you use the ATR. When I look at an ATR indicator, the figure is something like 0.0019, so how do you convert this into the number of pips for the stoploss?

  • Jez Liberty

    Stuart,
    Usually an ATR-based stop is just a way to place the stop price N number of ATRs from the entry price.
    Example:
    - Buy Long EURUSD at 1.31 (this is not a recommendation! ;-)
    - Calculate the ATR, maybe get a value like 0.01
    - Set the stop so that stop price is Buy_Price – N x ATR: if N = 1, stop price = 1.30 (1.31 – 0.01), if N = 3, stop price = 1.28, etc. (note that for Short Positions you’d have Sell_Price + N x ATR)
    This is a fairly neat way to adjust the stop taking into consideration the recent volatility of the instrument (as opposed to a fixed % amount, etc.)

  • Stuart

    Thanks for the reply Jez. So, right now EUR/USD is 1.32338. ATR is 0.00092. So:

    Stop price = 1.32338 – (2 x 0.00092)
    Stop price = 1.32338 – 0.00184
    Stop price = 1.32154

    That’s a stop loss of just 18.4 pips I think?

    It doesn’t seem enough… Or is it just that the ATR is low right now?

  • Jez Liberty

    ATR value sounds low but it depends on the period used to calculate it (ie an ATR(5) on 1-min bar will be much slower than an ATR(20) on daily bars).
    Otherwise calc looks good

  • Newone

    Being new to trading i am a bit puzzled with these stops. Does adapting the trailing stop daily mean setting a new order each day with a new trailing stop? Doesn’t this mean that the trailing stop will only be met if the intraday change exceeds f.e. 2 ATR? So, you create each day a new high from which to the trailing is done. Is this meaningfull?
    I suppose the stop and trailing stop also meant in this text is always 2 ATR? Or is this only the initial stop?
    Thanks for your help

  • Jez Liberty

    Newone, the trailing stops typically do not “go down” (ie it basically trails the price by being at most 2/3/5 ATRs away: every time the price makes a new high – for a long position – the stop is placed 2/3/5 ATRs away but when the price goes back down, the stop does not)

  • Newone

    Jez, thank you very much for this information. I understand u probably place new orders daily, but keep the stops in place in case the price dropped, and u raise the stops in case the prise rose. How about doing this on a monthly basis? Isn’t this more safe in order to prevent unnessasarry stop-out due to whipsawwing?
    I’m trying to follow trends using turbo’s. Do you know turbo’s or speeders? We have them in europe, it’s basically following for example a stock using a hedge (leverage). I think this is more understandeable then options and futures. You can never loose more than you invested either. What is your idea about trading such products? I don’t find much information on the net about opinions on trading these.
    Thanks for your information. I think your site is very informative and inspiring to people new to the business like myself :)

Leave a Comment