I have been running Forex Strategy Optimizations in NT7 where all of the eight parameters have a wide sweep of values. This results in optimization times that can be several DAYS of CONSTANT CPU use to finish. It does finish, that is very good. The CPU time use is expected, however, there are some issues that should be corrected.
First, the elapsed time indicator at the bottom of the Control Center only displays 24 hours maximum, then rolls over with no concern. The two digits for hour display can be allowed to display 00 to 99 and that would be an improvement, or alternately, a digit for days can be added. This problem is actually most annoying at the beginning of a long simulation, where the estimate of remaining time can be wrong because of the truncated DAY digit.
Second, a case study: I selected the option to optimize for "minimum draw-down" and set the results buffer to hold the best 5000 results (you would think that is big enough). When done, 4000 of these positions were filled with results that executed NO TRADES. Of course, no trades will result in no draw-down, but that is not useful news.
Third, related to the above, is a common result when I select the option to optimize for "max. profit factor" or "max. percent winns". There are often a lot of results shown that only have 1 or 2 trades. That is not very helpful when studying a year of historical data. I have to use very large buffer size and hope that there is room for the useful results.
There is no simple "one size fits all" solution, but just a bit of added logic can create a filter to remove the results that give less than one trade every two months or so. The filter can default to OFF for those who prefer to see everything.
Finally, I attempted some multi-tasking, and an error box appeared. I was running a long optimization, and at the same time, started the "New Indicator" wizard. I went through the series of steps, clicked "Finish", and the error box appeared. I am not sure exactly what it said, but it contained the name of the strategy that was being run in the optimization.
First, the elapsed time indicator at the bottom of the Control Center only displays 24 hours maximum, then rolls over with no concern. The two digits for hour display can be allowed to display 00 to 99 and that would be an improvement, or alternately, a digit for days can be added. This problem is actually most annoying at the beginning of a long simulation, where the estimate of remaining time can be wrong because of the truncated DAY digit.
Second, a case study: I selected the option to optimize for "minimum draw-down" and set the results buffer to hold the best 5000 results (you would think that is big enough). When done, 4000 of these positions were filled with results that executed NO TRADES. Of course, no trades will result in no draw-down, but that is not useful news.
Third, related to the above, is a common result when I select the option to optimize for "max. profit factor" or "max. percent winns". There are often a lot of results shown that only have 1 or 2 trades. That is not very helpful when studying a year of historical data. I have to use very large buffer size and hope that there is room for the useful results.
There is no simple "one size fits all" solution, but just a bit of added logic can create a filter to remove the results that give less than one trade every two months or so. The filter can default to OFF for those who prefer to see everything.
Finally, I attempted some multi-tasking, and an error box appeared. I was running a long optimization, and at the same time, started the "New Indicator" wizard. I went through the series of steps, clicked "Finish", and the error box appeared. I am not sure exactly what it said, but it contained the name of the strategy that was being run in the optimization.
Comment