[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: [amibroker] To compound or not to compound... that is the question



PureBytes Links

Trading Reference Links




More 
answers for Dingo:
<FONT face=Arial color=#0000ff 
size=2> 
<FONT face=Arial color=#0000ff 
size=2>1.  You asked which metrics I might use for selecting "best" 
parameters.   Things like Sharpe Ratio, K-Ratio (smoothness), come to 
mind.   I have more than 200 objective functions that are calculations 
like "RAR/MaxDD", etc.   I have my favourites that come and 
go.   I have tools to decide which objective function works best when 
the system is then carried forward out-of-sample.
<FONT face=Arial color=#0000ff 
size=2> 
<FONT face=Arial color=#0000ff 
size=2>2.  You asked about saving some data for out-of-sample.  This 
too is a complex issue.   To me, there are two aspects to your 
question.  When I'm trying to decide if the system has merit, I do two 
things.   I will optimize for the "best" parameters (various ways, as 
discussed previously) on data between 1994 and 2002.   I then will run 
that system/parameter set backward two years and forward one year.  If the 
results look good (hard to define, of course), then I will start to trade the 
system.   It is at this point that I will optimize up until the 
current date to decide on the parameter set I will use for realtime 
trading.   I may repeat this process every three months, or so, 
depending on how many trades are generated by the system.
<FONT face=Arial color=#0000ff 
size=2> 
<FONT face=Arial color=#0000ff 
size=2>3.  You said " I take it you're not a disciple 
of stocks having their own individual params?  ".    
There is no way that I would EVER use a different 
parameter set for individual stocks.   Yes, MSFT does look different 
than IBM.   But tomorrow, they can both look the same or completely 
different.    I want one set of parameters that is going to work 
on MSFT today and tomorrow, even if MSFT starts to look more like 
IBM.    We can all make systems that perfectly curve-fit to 
each stock in the universe.  To me, that's a good way to sell software, but 
not to trade my money.
<FONT face=Arial color=#0000ff 
size=2> 
<FONT face=Arial color=#0000ff 
size=2>4.  I told you how I would use AB to sort my signals by "something", 
such as P/E and then take only the top or bottom sorted stocks.   I 
think you are asking how I would do this outside of AB.   I only use 
AB to develop and test system ideas.   I don't think there is any tool 
better for the job and I have used most of them.   Once I have a 
system that I want to trade, I code the logic in a portfolio manager that I 
developed in Delphi.   That program knows exactly how much money is 
available for taking new trades tomorrow and will rank the orders by various 
means (proprietary but no mystery).  It will then connect to my broker's 
server and place the orders.   I do not look at what is being traded 
and I never look at a chart.   
<FONT face=Arial color=#0000ff 
size=2> 
<FONT face=Arial color=#0000ff 
size=2>Whew... I think I answered everything for you.   Again... just 
my opinions and everybody has at least one.
<BLOCKQUOTE 
>
  <FONT face="Times New Roman" 
  size=2>-----Original Message-----From: dingo 
  [mailto:dingo@xxxxxxxxxx]Sent: Thursday, April 17, 2003 4:41 
  PMTo: amibroker@xxxxxxxxxxxxxxxSubject: RE: [amibroker] 
  To compound or not to compound... that is the question
  <FONT face=Arial color=#0000ff 
  size=2>Thanks Chuck!
  <FONT face=Arial color=#0000ff 
  size=2> 
  My 
  further questions in red below:
  <FONT face=Arial color=#0000ff 
  size=2> 
  
    
    <FONT 
    face=Tahoma size=2>-----Original Message-----From: Chuck 
    Rademacher [mailto:chuck_rademacher@xxxxxxxxxx] Sent: Thursday, 
    April 17, 2003 4:00 PMTo: 
    amibroker@xxxxxxxxxxxxxxxSubject: RE: [amibroker] To compound or 
    not to compound... that is the question
    <FONT face=Arial color=#0000ff 
    size=2>Answers to Dingo's questions below:
    <FONT face=Arial color=#0000ff 
    size=2> 
    <FONT face=Arial color=#0000ff 
    size=2>1.  Which column(s) to use when deciding which parameters are 
    "best"?   I actually take the trade results to another program, 
    but given the columns that are available in AB, I would use total net profit 
    and/or average trade.   Hopefully, some of the parameters will be 
    same at the top of both of those lists.
    <FONT face=Arial color=#0000ff 
    size=2> 
    <FONT face=Arial color=#ff0000 
    size=2>I too have another program and am 
    curious as to what calculated results would lead to the "best" 
    params?
    <FONT face=Arial color=#0000ff 
    size=2> 
    <FONT face=Arial color=#0000ff 
    size=2>2.  Which date ranges do I use?   I pretty much always 
    use 1992 to current date.
    <FONT face=Arial color=#0000ff 
    size=2> 
    <FONT face=Arial color=#ff0000 
    size=2>You go all the way to the current 
    period?  Do you not test out of sample?
    <FONT face=Arial color=#0000ff 
    size=2> 
    <FONT face=Arial color=#0000ff 
    size=2>3.  I use all 13,500 active and extinct stocks that have ever 
    traded since January, 1992 that were over $1 on at least one day and had an 
    average 50-day volume of 75,000 shares on at least one 
    day.
    <FONT face=Arial color=#0000ff 
    size=2> 
    <FONT face=Arial color=#0000ff 
    size=2>4.  Do I set aside any losing trades/stocks?  
    No.   Since my largest loss can only be $10,000 and the profit can 
    be in the millions, I keep in all losing trades.  As I explained in 
    previous email, the losses CAN BE more than the investment on a few short 
    trades.
    <FONT face=Arial color=#0000ff 
    size=2> 
    <FONT face=Arial color=#0000ff 
    size=2>5.  How do I determine that the best parameters (if I include 
    huge profitable trades) could be on the edge of the parameter 
    space?   I can't really determine that information, nor do I 
    really care.   I remove those stocks from my watchlist and 
    re-optimise.   If I end up with a new list of huge winners, I 
    might repeat the process.   I just don't want to choose a 
    parameter that squeezes in a couple of huge trades but is inferior on the 
    rest of the market.   
    <FONT face=Arial color=#0000ff 
    size=2> 
    <FONT face=Arial color=#ff0000 
    size=2>You method sure is counter intuitive 
    until I think about your reasoning.  I take it you're not a disciple of 
    stocks having their own individual params?  
    <FONT face=Arial color=#0000ff 
    size=2> 
    <FONT face=Arial color=#0000ff 
    size=2>6.  How do I decide which signals to take every day (if I had 
    more signals than cash)?   Using AB, I would place something that 
    I could sort the signals by that I have proven to myself increases the 
    likelihood of the signal being a good one.   For simplicity, let's 
    say that I was an advocate of using low P/E for buying stocks (I'm 
    not).  I would add a column to my Explore that showed the current 
    P/E.
    <FONT face=Arial color=#0000ff 
    size=2> 
    <FONT face=Arial color=#ff0000 
    size=2>If you were to make this decision 
    independent of AB then would you make the 
    decision?
    <FONT face=Arial color=#ff0000 
    size=2> 
    <FONT 
    color=#0000ff>I'm quite happy to discuss any of this 
    further.   Obviously, these are only my views and they are no 
    better than those you or others might have on the subject.   I 
    have the benefit of trading for (too) many years and the possible 
    disadvantage of "being past it".<SPAN 
    class=843033320-17042003> 
    <FONT 
    color=#0000ff><SPAN 
    class=843033320-17042003> 
    <FONT 
    color=#0000ff><FONT 
    color=#ff0000>Do you ever re-optimize for the purpose of calc'ing new 
    params?  If you do how do you decide when it is time to do 
    that?
    <FONT 
    color=#0000ff><SPAN 
    class=843033320-17042003> 
    <FONT 
    color=#0000ff><FONT 
    color=#ff0000>Thanks for your 
    answers!
    <FONT 
    color=#0000ff><SPAN 
    class=843033320-17042003> 
    <FONT 
    color=#0000ff><FONT 
    color=#ff0000>d 
    <BLOCKQUOTE 
    >
      <FONT face="Times New Roman" 
      size=2>-----Original Message-----From: dingo 
      [mailto:dingo@xxxxxxxxxx]Sent: Thursday, April 17, 2003 11:10 
      AMTo: amibroker@xxxxxxxxxxxxxxxSubject: RE: 
      [amibroker] To compound or not to compound... that is the 
      question
      <FONT face=Arial color=#0000ff 
      size=2>I can understand and appreciate why you use fixed trade sizes in 
      order to get the best parameters. But how do you get a reasonable measure 
      of drawdowns that way?  Do you use some other technique to evaluate 
      drawdowns?
      <FONT face=Arial color=#0000ff 
      size=2> 
      <FONT face=Arial color=#0000ff 
      size=2>Re your param selection method: Do I understand the steps 
      correctly: 
      <FONT face=Arial color=#0000ff 
      size=2> 
      <FONT face=Arial color=#0000ff 
      size=2>1. You optimize for the best params 
      <FONT face=Arial color=#0000ff 
      size=2>        a. Based on what column 
      or calculation?
      <FONT face=Arial color=#0000ff 
      size=2>        b. What date ranges 
      would you be using currently?
      <FONT face=Arial color=#0000ff 
      size=2>        c. What subset of stocks 
      would you be optmizing on?
      <FONT face=Arial color=#0000ff 
      size=2> 
      <FONT face=Arial color=#0000ff 
      size=2>2. You set aside the the top 100.
      <FONT face=Arial color=#0000ff 
      size=2>        a. Do you set aside any 
      at the bottom?
      <FONT face=Arial color=#0000ff 
      size=2>        b. How did 
      you determine that the first set of params would be at the edge of 
      the parameter space? 
      <SPAN 
      class=671315814-17042003><FONT face=Arial color=#0000ff 
      size=2> 
      <FONT face=Arial color=#0000ff 
      size=2>3. You reoptimize the resultant set from step 2 and those are the 
      ones you use.
      <FONT face=Arial color=#0000ff 
      size=2> 
      <FONT face=Arial color=#0000ff 
      size=2>Given the size of your trading capital how do you decide what 
      stocks to trade on a particular day?
      <FONT face=Arial color=#0000ff 
      size=2> 
      <FONT face=Arial color=#0000ff 
      size=2>I'm not trying to pick a fight here I'm intensely curious as I've 
      been struggling with these questions for quite some time 
      now.
      <FONT face=Arial color=#0000ff 
      size=2> 
      <FONT face=Arial color=#0000ff 
      size=2>Thanks for any comments you choose to make.
      <FONT face=Arial color=#0000ff 
      size=2> 
      <FONT face=Arial color=#0000ff 
      size=2>d
      
        
        <FONT 
        face=Tahoma size=2>-----Original Message-----From: Chuck 
        Rademacher [mailto:chuck_rademacher@xxxxxxxxxx] Sent: 
        Thursday, April 17, 2003 6:58 AMTo: 
        amibroker@xxxxxxxxxxxxxxxSubject: [amibroker] To compound or 
        not to compound... that is the question
        <FONT face=Arial color=#0000ff 
        size=2>Reply to Fred:
        <FONT face=Arial color=#0000ff 
        size=2> 
        <FONT face=Arial color=#0000ff 
        size=2>Yes... and no.
        <FONT face=Arial color=#0000ff 
        size=2> 
        <FONT face=Arial color=#0000ff 
        size=2>Absolutely, in real time trading I am 
        compounding.
        <FONT face=Arial color=#0000ff 
        size=2> 
        <FONT face=Arial color=#0000ff 
        size=2>To determine parameters via optimization.... not if my life 
        depended on it!   And, I guess my life does depend on it, as I 
        make my living managing funds for others.
        <FONT face=Arial color=#0000ff 
        size=2> 
        <FONT face=Arial color=#0000ff 
        size=2>I mentioned one trade (AOL) where my system made $1.5 million on 
        a $10,000 investment.  That's not bragging... I'm sure you 
        could come up with a system that could achieve similar 
        performance.   Since the average trade generated a profit of 
        $2,700 for every $10,000 invested, the AOL trade could cover up lots of 
        bad trades made using one parameter set.   Compounding that 
        trade would exacerbate the problem.   A minor tweak to the 
        parameters could cut out the AOL trade, yet that very tweak could 
        improve performance going forward.   
        <FONT face=Arial color=#0000ff 
        size=2> 
        <FONT face=Arial color=#0000ff 
        size=2>When choosing parameters, I want plain vanilla trades, each 
        standing on their own merit, with no compounding.
        <FONT face=Arial color=#0000ff 
        size=2> 
        <FONT face=Arial color=#0000ff 
        size=2>We may have to agree to disagree.   It's like absolute 
        gospel to me and I cannot see clear to do it any other 
        way.    
        <BLOCKQUOTE 
        >
          <FONT face="Times New Roman" 
          size=2>-----Original Message-----From: Fred 
          [mailto:fctonetti@xxxxxxxxx]Sent: Thursday, April 17, 2003 
          3:16 AMTo: amibroker@xxxxxxxxxxxxxxxSubject: 
          [amibroker] FW: [aaft_ta] Re: 
          TradingRecipesChuck,I'm sure you'd 
          agree, wouldn't you ?, that one way or another you compound.  
          If you are not compounding by increasing bet size then you are 
          compounding by increasing the number of stocks you'll potentially 
          take simultaneous positions in as equity grows, right ?  
          --- In amibroker@xxxxxxxxxxxxxxx, "Chuck Rademacher" 
          <chuck_rademacher@x> wrote:> For what it is worth, I 
          use fixed bet size for all backtesting purposes.   
          I> coudn't imagine backtesting/optimizing using any other 
          approach.  I even go> a step further if I'm doing any 
          optimizing.   I recently posted an equity> curve 
          showing something like $80 million in profit.   Within that 
          $80> million, the top 100 stocks (out of 13,500) generated 
          $20 million in> profits.  AOL, by itself, generated $1.5 
          million in profits.  In each case,> the original trade 
          was only $10,000.> > As I said, I go a step further than 
          just using a fixed bet size.  After my> first pass at 
          optimizing, I remove the top performing 100 stocks.  I 
          then> re-optimize without those stocks.  Granted, I could 
          end up with some new> "top" stocks.  However, my 
          objective is to remove the extremely large> winners so that 
          the profits from those stocks don't cause me to select> 
          parameters on the edge of the parameter space.> > I 
          don't bother removing the worst performers as the largest loss 
          might be> something like $16,000 (even though the original 
          trade was only $10,000).> This can happen if a short trade 
          goes against you.> > As I said... for what it's 
          worth...>   -----Original 
          Message----->   From: Bob Jagow 
          [mailto:bjagow@xxxx]>   Sent: Thursday, April 17, 
          2003 2:21 AM>   To: Amibroker>   
          Subject: [amibroker] FW: [aaft_ta] Re: TradingRecipes> > 
          >   Re the "portfolio level testing" magic 
          bullet.> >   Bob>   
          -----Original Message----->   From: Palmer Wright 
          [mailto:palmerw@xxxx]>   Sent: Wednesday, April 16, 
          2003 8:27 PM>   To: 
          aaft_ta@xxxxxxxxxxxxxxx>   Subject: Re: [aaft_ta] 
          Fwd: Re: Available Portfolio testing programs for> 
          TS2000i> > >   Since Michael forwarded 
          the two messages (see below), he added four> additional 
          ones. The issue about whether a "basket system" like 
          Aberration> is worth trading I will not discuss here (I 
          still trade it). The other main> issue is about the effect 
          of compounding when testing with TR (Trading> Recipes), and 
          I comment here on that.> >   Traders buy TR 
          because it can test portfolios of systems and markets 
          using> position sizing. A position-sizing strategy such as 
          fixed-fractional money> management brings two advantages: 
          it normalizes markets (eg., calculating> many contracts for 
          corn, but few for natural gas), and limits entry risk for> 
          each position to a fixed- fraction of current equity--thus 
          preventing> overtrading. If you do not use TR, I do not 
          know how you can get the large> returns that compounding 
          multiple markets can bring.> >   Leslie Walko 
          points to the potential danger of curve fitting caused by> 
          compounding. I agree, and have been concerned for years about how 
          one market> in a portfolio (commodity X) by being 
          dramatically profitable in a single> year can misleadingly 
          bias the results of the whole portfolio.> >   
          During a multi-year test in TR, starting equity is low, perhaps 
          $100,000,> but compounding raises equity to many million in 
          later years. The one-year> outperformance of commodity X 
          cand produce two kinds of curve-fitting bias:> early-years 
          bias and end-years bias. Mark Johnson's message describes 
          the> first, where X gives "a big turbocharged boost" to the 
          portfolio's equity,> which then gives a head-start boost to 
          the number of trades in all the> commodities traded. The 
          second occurs when X's monster trades occur in the> final 
          years of the simulated time period when the large number of 
          contracts> makes X's profit far larger than if its big year 
          came early. Here the> profits contributed by X dwarf what 
          they were in the first case.> >   As the 
          message from M points out, we can avoid such biases by 
          normalizing> with a fixed-dollar bet size in testing to 
          remove the galloping equity> effect. I proposed this method 
          in 1999, and still use it to compare with the> compounded 
          performance. I confess, however, that my testing has failed 
          to> find as much performance bias as I suspected I would find. 
          The method is> most important when selecting markets for a 
          portfolio.> >   Palmer 
          Wright>     ----- Original Message 
          ----->     From: Michael 
          Guess>     To: 
          aaft_ta@xxxxxxxxxxxxxxx>     Sent: Sunday, 
          April 13, 2003 9:14 AM>     Subject: 
          [aaft_ta] Fwd: Re: Available Portfolio testing programs 
          for> TS2000i> > >     
          This is for Pat Mazur & Palmer Wright. Others are invited to 
          comment. I> forwarded these two messages from another list 
          because we have discussed> these issues in the past. It 
          appears one of the posts is saying Trading> Recipes is in 
          error in the way it calculates. In fact, that it curve 
          fits> data in a particular case. Comments are invited.> 
          >     Michael> > > 
          >   Your use of Yahoo! Groups is subject to the 
          Yahoo! Terms of Service.> 
          >         Yahoo! Groups 
          Sponsor>               
          ADVERTISEMENT> > > > >   
          Send BUG REPORTS to bugs@xxxx>   Send SUGGESTIONS to 
          suggest@xxxx>   
          ----------------------------------------->   Post 
          AmiQuote-related messages ONLY to: 
          amiquote@xxxxxxxxxxxxxxx>   (Web page: <A 
          href="">http://groups.yahoo.com/group/amiquote/messages/)>   
          -------------------------------------------->   Check 
          group FAQ at:> <A 
          href="">http://groups.yahoo.com/group/amibroker/files/groupfaq.html> 
          >   Your use of Yahoo! Groups is subject to the 
          Yahoo! Terms of Service.Send BUG REPORTS 
          to bugs@xxxxxxxxxxxxxSend SUGGESTIONS to 
          suggest@xxxxxxxxxxxxx-----------------------------------------Post 
          AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx (Web 
          page: <A 
          href="">http://groups.yahoo.com/group/amiquote/messages/)--------------------------------------------Check 
          group FAQ at: <A 
          href="">http://groups.yahoo.com/group/amibroker/files/groupfaq.html 
          Your use of Yahoo! Groups is subject to the <A 
          href="">Yahoo! Terms of 
          Service. Send BUG REPORTS to 
        bugs@xxxxxxxxxxxxxSend SUGGESTIONS to 
        suggest@xxxxxxxxxxxxx-----------------------------------------Post 
        AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx (Web 
        page: <A 
        href="">http://groups.yahoo.com/group/amiquote/messages/)--------------------------------------------Check 
        group FAQ at: <A 
        href="">http://groups.yahoo.com/group/amibroker/files/groupfaq.html 
        Your use of Yahoo! Groups is subject to the <A 
        href="">Yahoo! Terms of 
        Service. Send BUG REPORTS to 
      bugs@xxxxxxxxxxxxxSend SUGGESTIONS to 
      suggest@xxxxxxxxxxxxx-----------------------------------------Post 
      AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx (Web page: 
      <A 
      href="">http://groups.yahoo.com/group/amiquote/messages/)--------------------------------------------Check 
      group FAQ at: <A 
      href="">http://groups.yahoo.com/group/amibroker/files/groupfaq.html 
      Your use of Yahoo! Groups is subject to the <A 
      href="">Yahoo! Terms of Service. 
      Send BUG REPORTS to bugs@xxxxxxxxxxxxxSend 
    SUGGESTIONS to 
    suggest@xxxxxxxxxxxxx-----------------------------------------Post 
    AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx (Web page: 
    <A 
    href="">http://groups.yahoo.com/group/amiquote/messages/)--------------------------------------------Check 
    group FAQ at: <A 
    href="">http://groups.yahoo.com/group/amibroker/files/groupfaq.html 
    Your use of Yahoo! Groups is subject to the <A 
    href="">Yahoo! Terms of Service. 
    Send 
  BUG REPORTS to bugs@xxxxxxxxxxxxxSend SUGGESTIONS to 
  suggest@xxxxxxxxxxxxx-----------------------------------------Post 
  AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx (Web page: <A 
  href="">http://groups.yahoo.com/group/amiquote/messages/)--------------------------------------------Check 
  group FAQ at: <A 
  href="">http://groups.yahoo.com/group/amibroker/files/groupfaq.html 
  Your use of Yahoo! Groups is subject to the <A 
  href="">Yahoo! Terms of Service. 







Yahoo! Groups Sponsor












Send BUG REPORTS to bugs@xxxxxxxxxxxxx
Send SUGGESTIONS to suggest@xxxxxxxxxxxxx
-----------------------------------------
Post AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx 
(Web page: http://groups.yahoo.com/group/amiquote/messages/)
--------------------------------------------
Check group FAQ at: http://groups.yahoo.com/group/amibroker/files/groupfaq.html



Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.