June 2012 Signals
| Date | JT Composite | TNA | J-Trader | CCI |
| May 29 | Short | 50.45 | Short | Short |
| May 30 | Cash | 47.57 | Buy | Hold |
| Jun 1 | Long | 43.07 | Hold | Buy |
| Jun 4 | Long | Hold | Hold | |
| Jun 5 | Long | Hold | Hold | |
| Jun 6 | Cash | 47.73 | Sell | Sell |
| Jun 7 | Cash | Hold | Hold | |
| Jun 8 | Half-Short | 48.61 | Hold | Short |
| Jun 11 | Cash | 45.33 | Buy | Hold |
| Jun 12 | Half-Short | 47.09 | Sell | Hold |
| Jun 13 | Cash | 45.53 | Buy | Hold |
| Jun 14 | Cash | Hold | Hold | |
| Jun 15 | Short | 48.87 | Short | Hold |
| Jun 18 | Short | Hold | Hold | |
| Jun 19 | Short | Hold | Hold | |
| Jun 20 | Short | Hold | Hold | |
| Jun 21 | Cash | 47.67 | Buy | Hold |
| Jun 22 | Cash | Hold | Hold | |
| Jun 25 | Long | 47.03 | Buy (#2) | Buy |
| Jun 26 | Long | Hold | Hold | |
| Jun 27 | Cash | 49.67 | Sell | Sell |
| Jun 28 | Half-Short | 49.53 | Hold | Short |
| Jun 29 | Half-Short | 53.87 | Hold | Short |
Bought Tza at open doubled up around 1pm and bought some OTM jul calls around 330. Need to fill that gap
ReplyDeleteJ, here's something I read that might be a useful tool to gauge the robustness of your systems:
ReplyDelete"A system may not work well if the system is not robust.
That is, the system has been optimised without considering if he has chosen "islands of stability". The system designer have chosen specific values for variables that produced great results. But then they change their variables a little bit, there’s a significant difference in the system results such that the system no longer really performs. They key is to design a robust system that performs well with a range of values for variables, near the value you’ve chosen."
In other words, the suggestion is to test the systems with settings that vary from the ones that you've optimized the systems for, and then see if the results vary by a lot. If a range of values for your variables produces results that are all good (though not as good as the ones you've settle on.), then the systems are probably robust.
Aex
How does one quantify if the results "vary by a lot", or how much difference is a "significant difference"?
ReplyDeleteGood question. The short answer is that I don't know. I would actually be very interested in a way to quantify this, since it would be a great tool in system building.
ReplyDeleteBut I can give you an impressionistic answer. You can take a look at an optimization report. If there is a small group of variables that produces great profits and then the rest fall off precipitously, then the system is probably not robust. If a lot of variables produce results that are pretty similar, then it probably is.
So, if one is optimizing for a system and using $100,000 per investment in order to instantly see percentile returns, if the results are in the same ballpark (within 10 percent of each other or so), then the system is probably robust. If one set of variables produces great results and then the rest are all over the map (from mediocre to severely negative), then the system is almost certainly not going to hold up over time.
Again, I know that I've not answered your question. I've simply tried to show what procedure would be required to start answering it.
Alex