picking trades vs taking all signals

Discussion in 'Automated Trading' started by gnom, May 29, 2010.

  1. gnom


    My system generated about 3000 trades over last 6 months with good results. In real time I won't be able to trade this often and will be forced pick and choose. Is it fair to say that I will be trading with similar ratios as my system shows or because I'll be trading less my results could be much worse. What is your experience?
  2. LeeD


    I believe the best approach would be to alter rthe system in such a way that it takes the number of trades you can actually take. Perhaps, add restrictions on trading hours and frequency of trades. Does the system add to an existing position?
  3. NumLock


    what you can do is take the trade after a loss

    no this won't give you better performance but it can give you less DD for less trades that you will take.
  4. edbar


    3000 trades in 6 months is about 25 trades a day.
    That's about 1 trade every 15 minutes.

    That may or may not be excessive.

    I suppose if you have limited funds and the trades are over-lapping and your limited funds prevent you from opening all of the positions.

    I'll just take your word for it that 3000 trades is too much.

    The fact is, if you cut out some of the trades, the trades that get cut out will definitely yield different results than if you executed all of the trades. However, there is no way to know if the trades you eliminate will skew the results to the good or bad side.

    For valid trading tests, you must employ a money management scheme so that you do not open positions beyond what you can afford.

  5. It depends on the distribution of your winners/losers. In terms of size and frequency.
    So if you have some big winners in your backtest and without thoses the profitability of the system would be much worse then it's not a good idea to select trades.
    Same thing goes for losers. If you expect to have some big losers every once in a while then by selecting trades you could select too many of those.
    But if the size of you winners and losers is about the same and there are no fat tails in your distribution you can select trades. But still, it's not a good idea.
    If you have a backtest and you like the results, just test sampling from it. Suppose you have 10000 trades and you plan on trading only 1000 for every 10000. Then take some random sample into the 10000 and look if you still like the results. You could do some monte carlo simulation but I wont get into that.