Tom, There are any number of formulas people use to evaluate results. Many follow the axiom of "no pain, no gain" and are expressed in the form of a ratio or as numerator over denominator. These simply describe how much pain had to be endured in order to accumulate the gain.
With that in mind, you should be able to figure out which part of the equation represents the pain, which part represents the gain, and whether a large number is good or bad. Hint; Variance, as expressed by standard deviation, is pain. Using one of your examples; A big CAR (gain) is only part of the equation. When you also taken into consideration the MDD (pain) you may not be as attracted to the strategy. A 50% CAR would be great, but not as palatable when the cost is a 75% drawdown. Too much pain for the gain. Most would prefer less pain (lower MDD) even if it meant a lower gain. Note that CAR/MDD can result in a larger number in either of two ways: A large CAR over a moderate MDD, or a moderate CAR over a small MDD. As such, trying for a smaller MDD is as benificial as trying for a large CAR, and trying for both gives even better results. Answering your specific questions: - 12 is bigger than 2, so yes a CAR/MDD of 12 is good. You have experienced 12 times more gain than pain. - The K-Ratio description is out of date. The formula has since been revised such that (according to Howard Bandy) 0.15 or better is good. Note that K-Ratio uses standard deviation and so penalizes positive variance as much as negative variance. Some people prefer to use ulcer index for this reason. See http://www.tangotools.com/ui/ui.htm - A lower ulcer index is better. As suggested by the name, a bigger ulcer means more pain. - I have found Bandy's approach to finding a fitness function usefull. He suggests that a good way to decide whether a measure suits your personal preferences is to look at the equity curve of a number of different strategies or parameter combinations and arrange them in order of best to worst (based solely on what you feel about the chart). Then look at the measurements and see whether the progression of values reflect your arrangement. For example; If the highest CAR/MDD was for your favorite chart, the next highest CAR/MDD is for your next favorite, and so on all the way down to your least favorite, then CAR/MDD would be a very good metric ("fitness function") for you to compare strategies with, since it so closely correlates to the type of equity curve that you favor. Generally, measurements that contain both a gain element and a pain element are good candidates for fitness. Looking at only one side (e.g. only CAR) or the other (e.g. only MDD) generally does not work well in out of sample data. Some candidates are CAR/MDD, RAR/MDD, Ulcer Index, UPI, K-Ratio, etc. Mike --- In [email protected], "professor77747" <profes...@...> wrote: > > Thank you for your replies. However, I have the way they are calculated, but > I don't know what they mean. For instance, CAR/MaxDD is good if bigger than > 2. My formula is over 12. Is that good? > > However, the K-Ratio should be over 1.0. My forumula is .06. So that must be > bad. > > I don't know how to judge my formula because I really don't understand how to > interpret the figures. > > What should the Ulcer Index be. Is higher better or worse. > > I have just compared my formulas using Profit and Downdraw. I take the > highest profit as long as the downdraw is close. > > Tom > > --- In [email protected], "Mike" <sfclimbers@> wrote: > > > > http://www.amibroker.com/guide/h_report.html > > http://www.investopedia.com/categories/formulas.asp > > > > Mike > > > > --- In [email protected], "professor77747" <professor@> wrote: > > > > > > > > > I have a very profitable formual that I have been autotrading for over a > > > year. However, it is also risky. I have another formula that is not as > > > profitable, but is also not as risky. My formula trade almost exactly as > > > a backtest except for the price which varies by so little that it is not > > > a factor. > > > > > > I don't understand any of the risk % factors in the top section and the > > > factors below the drawdown figures in the bottom section. > > > > > > Here is a link to the statistics for last year which are very similar to > > > this year except that there is more data. Statistics > > > <http://success101.biz/Backtest%20Report.htm> > > > > > > Please help me understand these statistics. Thanks > > > > > > Tom > > > > > >
