Profitable months per year

Discussion in 'Strategy Development' started by insight, Mar 21, 2008.

  1. insight

    insight

    Could someone please point me in the right direction with this query?

    I would like to calculate the percentages of how many profitable months I can expect in a given period.

    I understand you only need profit factor and trade frequency for this calculation, though can't find the maths either here or on google.
     
  2. insight

    insight

    Nobody?

    Perhaps it would be accurate to say I would like to know:

    How many trades will it take to be profitable at the X % confidence level?
     
  3. If I understand your question correctly, which maybe I don't. I believe it depends upon how many times you can do the following correctly.

    1. correctly analyze market conditions

    2. match a correct strategy to said market conditions

    3. have the discipline to follow said strategy until market analysis indicates otherwise. Then, repeat process.
     
  4. insight

    insight

    Actually it's just a maths question.

    Was thinking I might get a few trading stats people on these forums.

    If your profit factor is 2 and you make 1 trade per day how many days would you need to trade to be profitable at a 95% confidence level? I'm looking for the math to work out questions such as this.
     
  5. You also need a win %.

    What percentage of your trades make 2, and do all the other trades lose -1?

    Ideally you have the following information from backtesting:

    Average win size and win%.
    Average loss size and loss% (loss % is normally 100-win%).
    Average trades per day.


    eg your system wins on average 2, 50% of the time.
    losses on average 1, 50% of time.
    Makes one trade a day.

    With this data it is fairly easy to write a monte carlo simulation, in you favourite language or MS Excel to find out all the stats you need. If you are good at stats you could probably work it using equations instead of monte carlo.. but i always use monte carlo.
     
  6. EricP

    EricP

  7. vikana

    vikana Moderator

    In addition to EricP's very insightful post, i suggest you simply generate the statistics as part of your backtest:

    Tables with years (rows) and months (columns), and for each cell calculate $profit, $DD, cps, %dd, length of DD , and any other piece of data you care for.

    I don't think there's any other way to know how your system actually behaved.
     
  8. TheoCap

    TheoCap

    Courtesy of ACrary.
     
  9. insight

    insight

    Excellent replies, thanks :)

    I have what I was looking for and some more avenues for research now.
     
  10. Acrary and a few others had a debate on this long time ago (not on ET).

    My conclusion is run a Monte Carlo.
     
    #10     Mar 25, 2008