View System Optimization Report
Previous  Top  Next

When an advanced study is optimized through UA, a system optimization report is produced that may be viewed via the right-click charting menu. When optimizing a study through MarketScanner, you can view the Optimization Report in essentially the same way. First run your selected study, then view the results in a UA chart. Next right-click your mouse on the chart (or MarketScanner spreadsheet). From the menu that displays, choose "View System Optimization Report." All variables will be shown, sorted by your chosen optimization field and ranked from lowest result to highest. If your goal was to find the maximum result (such as the most net profit), the best results (the best performers) will be at the bottom of the report. If your goal was to find the least drawdown or other low result, the best performers will be at the top of the report.

Interpreting the Report

The most valuable data provided by the optimization report is the progression of variables leading up to the single "best" result. In a valid optimization exercise, results will gradually become better and better as inputs are gradually increased or decreased to attain the optimal level. This produces a table that demonstrates similar inputs generating similar results. In other words, the input variables for the second, third and fourth best results should be very similar to the input variables for the best result. As a general rule, expect all members of the best-performing group to have similar inputs and to be consistently effective in real-time application.

If, however, the inputs that produce the "best" optimized result are not similar to the inputs for the other well-performing results, be suspicious. Also be suspicious if the "best" result is remarkably superior to the other results. Both of these situations would indicate a statistical anomaly or a fluke, as opposed to a valid optimization that is likely to yield similar results in the future. Beware of using optimized parameters that have been curve-fit to your test period. Such results are not likely to repeat in real-time trading. See the Trading System Performance Evaluator chapter for more information on appropriate levels of optimization.