Testing Trading Systems

biobuck

Testing Trading Systems

来自：MACD论坛(bbs.shudaoyoufang.com) 作者：biobuck 浏览：3949 回复：3

Testing Trading Systems:
Look at the Numbers that Matter Most

Many traders think that having a high Percentage Profitable is the main thing they're looking for.  But that's not the only factor that affects total profitability.  The other primary factor is the Ratio of (Average Win / Average Loss).  If you have a 90% winning percentage, but your one loser out of 10 trades is 10 times your average winner, you're not going to do well.  Similarly, I have seen numerous systems with low winning percentages do well, because they keep their losses small and let their winners ride, forming a very large Ratio of (Average Win/Average Loss).  For example, with a winning percentage of 30%, if the size of the Average Win is five times greater than the size of the Average Loss, this system will do well.

A way to evaluate systems is to create a baseline for comparison.  I like to start with a sample random system, say flipping a coin.  If you have a 50-50 chance of calling Heads, and you win the same amount as you lose on any particular flip, then your Percent Profitable is 50% and your Ratio of (Average Win/Average Loss) is 1.00.  So if you multiply these two factors together, you get a random Profitability Factor of 0.50 (.50 winning percentage X 1.00 average win/loss).  I like systems that can at least double the random baseline, giving me a Profitability Factor of 1.00 or higher.  This makes sure I've accounted for the impact of commissions and slippage (the cost of buying at the offering price or selling at the bid price).

Let's evaluate our futures systems Performance Summaries in this context.  I looked at the continuous futures contracts in each case:

      Percentage Profitable     Ratio Avg. Win/Avg. Loss       Profitability Factor

S&P:                .60                   X             0.49                             = 0.29

T-Bonds:         .57                   X            2.07                             = 1.18

Crude:              .56                   X            2.26                             = 1.27

Gold:                .29                   X            8.68                             = 2.52

In the above table we see that only S&P's offer worse results than our random coin-flip benchmark Profitability Factor of 0.50.  All other systems do more than twice as well as the random benchmark, with the Gold market doing 5 times as well as the random model.  As we noted earlier, even though Gold had the lowest percentage of profitable trades, its profitability overall was superior because it cut losses quickly while letting winners ride, resulting in winners which were 8.68 times greater in size than the losers, on average.

谁怕谁

测试交易系统？

谁来翻译一哈，俺看不懂。

飞网

同感!作为系统交易,胜率高低并不重要,重要的是盈利与损失之比.国外很多成功的trader都倾向这一比例较高的方法,他们一般认为是3左右,只有这样,他们才会下单交易!"专业投机原理"中经常提到这一点,不过他是用概率的统计方法!

慕融68

值得探讨！虽然文章不长，但对我启发很大，谢谢！