The Watch
I created a new monitor strategy called The Watch that sends an email whenever there are any changes to the closed trades table. I find this much more useful than specifically monitoring trade by trade.
Time Limit
Often I read about strategies that try to exit before the close of the day, or after a certain time period. I don’t have this in any of my strategies and I am curious if it will improve the results or not. Today I decided to look into this.
Firstly, I decided to print out the time in the market for each trade, and calculate the average for profit trades vs loss trades. I also plotting these on a histogram to try to see if there was any obvious thresholds that if exceeded resulted in more loss trades than profit trades. I couldn’t find any.
Never-the-less, I decided to implement a time based limit per trade, ergo, Time Limit. A new input has been created which allows you to enter a limit in days, supporting fractions. I.e., 3.333 would be 3 days and 8 hours. This is not business days but calendar days. I implemented the logic on the CONTINUATION and TREND_REV strategies and then fired up the optimizer on a couple of configurations to see if it yielded better results. Unfortunately, the most profitable result was always with the Time Limit disabled. The picture below shows the Final Balance on the y-axis and Time Limit on the x-axis.
Ok, so maybe we would be cutting ourselves off from some profit trades as well as loss trades, but has the overall Win ratio (Profit vs Loss count) increased?
Pass |
Final balance |
Time Limit |
Trade count |
Profit trades |
Loss trades |
WIN RATIO |
1 |
8,696.14 |
0.000 |
143 |
111 |
32 |
3.47 |
5 |
8,565.70 |
4.375 |
143 |
109 |
34 |
3.21 |
32 |
8,581.34 |
4.875 |
143 |
109 |
34 |
3.21 |
41 |
8,440.01 |
5.875 |
143 |
109 |
34 |
3.21 |
48 |
8,412.55 |
6.000 |
143 |
109 |
34 |
3.21 |
13 |
8,537.98 |
4.500 |
143 |
108 |
35 |
3.09 |
25 |
8,600.14 |
4.750 |
143 |
108 |
35 |
3.09 |
47 |
8,517.30 |
4.250 |
143 |
108 |
35 |
3.09 |
6 |
8,443.22 |
5.250 |
143 |
107 |
36 |
2.97 |
18 |
8,550.81 |
4.625 |
143 |
107 |
36 |
2.97 |
Pass |
Final balance |
Time Limit |
Trade count |
Profit trades |
Loss trades |
WIN RATIO |
49 |
7,910.16 |
0.750 |
65 |
40 |
25 |
1.60 |
14 |
7,928.58 |
1.000 |
64 |
37 |
27 |
1.37 |
21 |
7,765.09 |
1.125 |
64 |
37 |
27 |
1.37 |
1 |
8,135.92 |
0.000 |
63 |
36 |
27 |
1.33 |
38 |
8,117.65 |
5.000 |
63 |
36 |
27 |
1.33 |
31 |
8,077.70 |
4.875 |
63 |
36 |
27 |
1.33 |
23 |
8,048.85 |
4.750 |
63 |
36 |
27 |
1.33 |
16 |
8,026.07 |
4.625 |
63 |
36 |
27 |
1.33 |
37 |
7,940.96 |
4.125 |
63 |
36 |
27 |
1.33 |
36 |
7,918.63 |
3.250 |
63 |
36 |
27 |
1.33 |
The tables above are the top results for this test – TREND_REV and then CONTINUATION. We can see that for TREND_REV it did not improve the Win Ratio at all, but for CONTINUATION there were three sets of results that had better Win Ratios. So I guess it works, and sometimes it may be more feasible to do this, however in these cases the better ratio they did not make more money.
Perhaps the behavior of this functionality depends on the type of strategy and profile of the currency-pair. The implementation is easy enough to transfer between strategies so let’s just shelve this for now and periodically test it on strategies to see if it helps.