I wanna use some statistical tools here is the pic I use the trend function in Excel to judge wether it is profitable or not ? If I am wrong ,plz correct me
How to judge a system no longer profitable? === I suspect you are experiencing a draw down and want to know when the system will show profit. No one can consistently correctly and reliably predict the nonexistent future. I know of no way to determine when in the future a draw down will end. Unless profit is a historic high value then judging system profitability is a personal decision. If my method shows 8 % profit for the week, 10 % loss for the month, 15 % gain for the year and 20 % loss from account equity two years ago is my system profitable? Sometimes I use the house indicator: <img src=http://towleroad.typepad.com/photos/uncategorized/mansion.jpg \img> This person owning this house might be very profitable. <img src=http://wyll.oneplace.com/Subpages/FOOD%20FOR%20THE%20POOR-PIC%20OF%20HOUSE2.jpg \img> This person owning this house might not be as profitable.
Thats not a bad yardstick, but what if the guy with house #1 has a $zillion in debt and the guy with house #2 has zero debt, which is more profitable?
check out this thread: http://www.elitetrader.com/vb/showthread.php?s=&threadid=36083&highlight=confidence
It's a good question, but this has been discussed many times in the past: when the system's live trading drawdown has exceeded the worst drawdown experienced in the historical backtesting period, then it's "done", "cooked", "fini", etc.,etc. Others have argued to use a mean and standard deviation approach for the max drawdown measurement. That approach would have the system "cooked" before reaching the largest drawdown in the backtest period. So if the mean drawdown was -$5000 and the std dev of drawdowns was $2000, then at 2 std deviations, the system would be cooked when it hit a drawdown of $-9000.
First you would need to see the probability of your profitable results being just luck. once you do that, you will have a p value. for example, lets say the hypothesis test shows you that there is 20% chance (p value of .2) of your results being random. Then if you decide to trade this strategy, you can monitor this p value to see if it increases or if it decreases. if it is increasing (p value going from the initial .2 to .25, .30, etc.) then you know that the strategy is becoming less and less profitable, or at least, that the results of your strategy are coming closer and closer to results that are attainable by luck. Which means that your edge, assuming you had one in the first place with a p value of .2, is disappearing. this also has been discussed before on ET, i would do a keyword search for "randomness", or "edge test". Hope this helps. shane
Your results also have the characteristics of a strategy that has been over optimized. of course it is hard to tell by just looking at an equity curve.