Frosty's trading bot goes live part 2

Discussion in 'Automated Trading' started by frostengine, Jun 18, 2007.

  1. In this current sample the average loss is MUCH bigger than the average win. But historically the average win is much bigger than the average loss. In the 14 days its been live its the other way around, but long term it is not.
     
    #211     Jul 10, 2007
  2. Hmm ... Why did I get a different result?
     
    #212     Jul 10, 2007
  3. I can only go by what is posted. And granted the sample is small. Nonetheless, I have a difficult time getting my brain around the concept that it is coincidence this near 3 trading weeks sample contains approximately 33% outliers, 100% on the loss-side. Something is askew, and imo, it has to do with exits. Position sizing is not a factor as all trades are 1 contract.

    As always, all the best.
    Osorico
     
    #213     Jul 10, 2007
  4. nkhoi

    nkhoi

    can you also give how many r/t each day?
     
    #214     Jul 10, 2007
  5. It is averaging around 3.5 round trips a day during this sample
     
    #215     Jul 10, 2007
  6. surprising. must be a different implementation/data/test suite. did you run on eur/usd or something else? would you mind sharing with us which implementation of simple ma crossovers is a profitable automated signal? :D
     
    #216     Jul 10, 2007
  7. GTS

    GTS

    That's a faulty assumption without knowing the details of the individual trades (rather then just the daily totals)

    What if one of the days that ended up net positive actually started with the first trade being a $500 loss...simply stopping at -$500 would have turned that winning day into a losing day.

    Often times adding an artificial stop loss to a system causes it to be less profitable overall. I would surprised if Frosty didnt test something so simple (stopping after an $XX loss) in his backtesting.
     
    #217     Jul 10, 2007
  8. This is easy to discover and you are also making a faulty assumption that he did already tested something so simple. If he did, I doubt he would have had such major drawdowns compared to most days where even losses were still 3 digits. As of now, on a small data set stopping at $500 would have saved $1,500 so it is worth investigating further.
     
    #218     Jul 10, 2007
  9. GTS

    GTS

    I said I would be surprised if he had not done that testing; that is not an assumption.

    An example of an assumption is the unqualified statement that implementing a daily $500 stop loss would have saved him $xxxx in losses. You just repeated that statement in your last post and again I say you do not have enough information to know if that would have been true even for this small sample set.

    Do you understand that yesterday he could have had a $700 loss on the bot's first trade but ended up $250 for the day? How do you know that isnt the case? Again - you do not enough information to make the statement that a $500 daily stop loss would haved resulted in $1500 in savings over the past 3 weeks.

    Not trying to pick a fight but I have done lots of ATS backtesting so I know about what I speak. If only it were just a matter of slapping a stop loss into a system to make it more profitable, unfortunately that is seldom the case.
     
    #219     Jul 10, 2007
  10. The point is basically the same, he should focus on those two days in particular to see if there is something he can fix.
     
    #220     Jul 10, 2007