|
|
|
|
|
|
|
and NG. Long and short entries are on 48-day highs and lows, and exit stops are trailing 12-day lows and highs. Sixty dollars per contract is provided for commissions and slippage. |
|
|
|
|
|
|
|
|
A full year of out-of-sample data is now available. Test-48 was run with 6/30/97 data using the original parameters. The out-of-sample period shows a somewhat enhanced winning percentage and a lower average win to average loss ratio, resulting in a slightly lower Kelly. The risk-reward curves were calculated for the entire 1/91 to 6/97 period. |
|
|
|
|
|
|
|
|
Neal: Would you explain Table 9-1 a bit more? |
|
|
|
|
|
|
|
|
C.V.: The table shows the relationship between risk per trade, which is a function of the number of contracts and stop, return (CAR, or compound annual return), and drawdown. Up to a point, increasing risk increases return, while also increasing drawdown. |
|
|
|
|
|
|
|
|
When risk is increased past the system's ability to absorb it, return decreases and drawdown increases. |
|
|
|
|
|
|
|
|
Neal: In reality, how much does this system make? |
|
|
|
|
|
|
|
|
C.V.: Traded on a conservative basis, the system would probably produce a return in the 15% range. |
|
|
|
|
|
|
|
|
Neal: You do use some commonsense trading concepts, like trading into a report, etc. |
|
|
|
|
|
|
|
|
C.V.: I've learned from experience that it is hard to quantify the effect of major reports in historical testing. Therefore, I do not use an entry order on a big report day until after the report is out. |
|
|
|
|
|
|
|
|
Neal: What is your definition of a trend up and trend down? |
|
|
|
|
|