basic question on statistics

Discussion in 'Strategy Development' started by radolym, Apr 15, 2006.

  1. radolym


    I developed a swing trading strategy that makes 500+ trades per year and has 55% win/loss with good risk reward.

    However based on my money management I can't take every trade.

    Do I need to take every trade to achieve similar win/loss and risk/reward results or based on law of statistics ( I don't know any that's why I'm asking) it's fare to say if I trade long enough I should get similar results even I don't take every trade?

    Please help me to understand how statistics work. Do I need to take every trade or if I only take 200 trades out of 500 I should get similar results with let's say some % deviation.
  2. Consider tailoring it to Retail Spot Forex and open up an account that enables you to place smaller trades...

    Take every trade....

    consider the validity of your statistics in a bear and a bull environment and make sure your methodology can go long and short with equal results...

    Michael B.
  3. Your trades may not be evenly distributed over time. It's possible your profits are batched in a large number of trades during short periods, while the remainder might be marginally profitable -- or unprofitable -- as a group.
  4. radolym


    any other comments?
  5. waterboy


    you don't have to make every trade but,

    the more trades you make, the more expectation$
  6. bitrend


    Since your success rate is 55% which is greater than 50% then the more you trade the more you gain as well as your broker :D assuming your average win equals average lose. If the average win is not equal average lose then it's another story.

    If your success rate is below 50%, let say 40% than you have to have an average win greater than average lose in order to be profitable. In this case the money management plays a bigger part. For example, each single trade might have a worst lose at -10% and best win at 20%.
  7. If you took fewer trades and those were randomly selected from all of the trades your system signaled, then your expected percentage of winners should be similar to your large sample results (+/- some margin of error). If your subset of trades has a specific bias (e.g., you only kept those that met your money management criteria), then winners may be more or less prevalent in that new, smaller sample. You're selecting out observations based on a specific characteristic, not on a random basis, and that characteristic may have a direct impact on the qualities that define a winner. It's entirely likely that the two groups may show different winning percentages given that they have different characteristics.

    You would need to try to quantify the effect of that quality as a binary variable (yes, it meets the MM criteria / no, it doesn't) relative to expected profitably (model "realized profit" = "meets criteria"). Maybe you can include "maximum open loss" somewhere in the model as a co-factor. I'm sure someone here who is actually knowledgeable can suggest the appropriate statistical test (p-value from two-sample t-test? / I'm talking out my ass at this point, but it would be something similar to this).
  8. rws3z


    Your question is the bain of many traders, and is often the key to a trading system failure. The first posters are absolutely correct, in that you need to take a random sample period of time and see if you receive the 55/45 results that you indicated you had.

    Then if you had not done so already, remove the outliers. Those trades that produce the greatest profit and loss should, in long run, have zero effect on your system. There is nothing worse then watching a decaying system and waiting for that one big trade that makes the year.

    Finally, remove the winner outliers and keep the loser outliers; if your system still returns results, you got a keeper.

    So Follow The Leader Down, So Swallow Your Pride And Drown
  9. radolym


    Thanks rws3z

    Good advise Also as someone pointed on another post 10-15% of my trades account for my net profitability. So if I remove the 1 biggest winning trade I'm sure it'll be fine but if I remove top 10 or so it'll make huge difference. Any comments?

    By the way, this strategy I'm talking about is a swing/trend following and makes only about 30 trades per year. Pretty boring but makes good money...
  10. rws3z


    Then, in response to your orginal question, it would be of the upmost importance to get every trade possible. A one trade difference yields could be the difference between a losing system and a winning system, since each trade is about 3 percent of the system.
    #10     Apr 18, 2006