How to filter out losing trades

Discussion in 'Strategy Development' started by BrooksRimes, Apr 28, 2010.

  1. I would like to filter out some of the losing trades from an automated system. It’s a little more complicated than a straight stock/futures system because it is trading options on RUT (Russell 2000 index).

    Here is a snapshot:
    There are 170 trades evenly divided between longs and shorts. There are 27 losers (22%) also almost evenly divided between longs and shorts. The net profit is $26,800 trading 2 contracts.

    The goal is to increase the net profit by eliminating some of the losing trades (and fewer if any of the winning trades).

    Losers occur because the market becomes too volatile after entry. The situation is worse if the market becomes volatile immediately after entry.

    Some examples of trades I’m trying to eliminate:
    The system starts a bearish trade on 7/16/08 in the middle of an uptrend.
    Same with a bearish trade starting on 3/11/09.

    Starting a bullish trend on 10/17/07 during a downtrend.
    Same for 12/12/07.

    Attached is a chart.

    Currently the entries are based on X days to expiration.

    I was experimenting with a filter for Linear Regression Slope but it wasn’t working out well.

    Would appreciate any thoughts or ideas to explore.


  2. charts


    This sounds like adding a random noise generator ... :)
  3. I'm not sure what you mean.

    With option income strategies based on writing options (time decay, collecting theta), its common to open the days a set number of days to expiration.

    It IS different from waiting for a setup on a stock trade.

  4. charts


    Believe me: there is no such thing as "option income strategies based on writing options" ... They ignore risk, and something sooner or later will surprise you ... :)

    Opening a position at "a set number of days to expiration" is just a random thing, and over a large number of samples it evens out. But, if a surprise hits you early you might not live (financially) long enough to even out, even if you employ good money management. :)
  5. BR,

    The idea of a filter is flawed from the get-go. Essentially when people try to implement filters they are attempting to "filter" out bad trades while keeping good trades. Unfortunately it does not work like that. Any "filter" you apply will end up taking out good trades as well and in most cases will not make your performance any better. As a matter of fact, more often than not the "filter" will make performance even worse.

    With a 78% hit rate I'm assuming that winners are much smaller than losers and this is why you are trying to find the "filter."

    In reality a model either works or it does not. If it does not throw it out; if it does than you will not need a "filter." Furthermore trying to come up with a filter for a sample of 127 trades is all but guaranteed to fall apart going forward.

    Remember your building models here not making coffee. No filters needed!
  6. I really don't think you are very familiar with this. I know at least 4 guys making 6 figures plus per year writing options.

    This is a very reputable guy and program:

    You're right that there is risk, but there is in all trading. It a matter of managing the risk.

  7. Would it make any difference if I call it "refining the setup"?

    Whether you call it a filter or setup, it's improving the entry or the trade selection. People do it all the time. System traders spent a significant amount of time on this area.

  8. MGJ


    You could add a new Exit to your system. Something like
    If((position is not Flat) and (todays volatility > 1.1 * (60 day average of volatility))) then exit half the position
    If((position is not Flat) and (todays volatility > 1.5 * (volatility on the day of trade entry))) then exit all of the position
    You could fiddle around with the parameter values (1.1, 60, 1.5) and pretend you are Optimizing.
  9. Thanks a lot MGJ. This was the type of suggestion I was seeking.


  10. charts


    ... :) Just remember that somebody warned you.
    #10     Apr 29, 2010