1. The robot was developed with backtesting only through mid-2007 because the 20 DMF sensibility evolved drastically after the uptick rule changed in the summer of 2007. Optimal buy and sell signals rules are much different before and after that event. A consequence of this is also that the robots will have to be suspended if there is any new change in uptick rule and it will take time before a new 20 DMF model could become reliable.
There is only satisfaction with the robots performance. They have perfectly fulfilled their role since being online, in a choppy environment that did not offer any long term trend to follow yet. Not only was capital well protected compared to buy & hold, but they also did much better than the alternative of staying in cash. Actual performance can only be measured from early March 2011 because all prior trades were still part of the backtesting timeframe. And we cannot take into account the trades from the beta versions.
Pascal will provide a more detailed performance analysis soon, but here are my own records of the robots:
IWM Robot : Compounded Return since 3/21: +4.45% vs -1.67% IWM Buy & Hold.
Type Entry date exit date Entry price Exit price Profit Reason
Buy 3/21/2011 4/11/2011 80.92 83.23 2.85% MF
Short 4/13/2011 4/20/2011 82.77 83.83 -1.28% MF
Short 4/26/2011 5/10/2011 84.71 85.29 -0.68% SL
Short 5/12/2011 5/27/2011 84.81 83.44 1.62% SL
Buy 6/10/2011 78.39 80.04 1.94%
Compounded Return 4.45%
GDX Robot: Compounded Return Since 3/9: +10.14% vs – 11.84% GDX Buy & Hold
Type Entry date exit date Entry price Exit price Profit Reason
Shorted 3/9/2011 3/10/2011 59.69 56.83 4.79% MF
Bought 3/11/2011 4/1/2011 56.25 59.77 6.27% MF
Bought 4/6/2011 5/3/2011 62.49 58.71 -6.05% SL
Bought 5/10/2011 5/10/2011 57.24 56.76 -0.83% MF
Shorted 5/11/2011 5/11/2011 56.66 54.91 3.08% MF
Bought 5/12/2011 6/6/2011 54.31 54.82 0.94% MF
Shorted 6/7/2011 6/8/2011 55.28 53.43 3.35% MF
Bought 6/23/2011 53.26 -1.28%
Compounded Return: +10.14%
2. We hope first that the robot will provide excellent long term risk-adjusted performance. Increased popularity should become exceptional to really create feedback effects and we are far from it with our current audience. We won’t hesitate to charge $2,500 per robot next year if this is the case! But more seriously, we prefer to expect some periods of disappointment that are statistically unavoidable. The robots were first created for our own trading discipline needs and we are sharing them with pleasure to our group of friends and followers without an aggressive marketing plan in mind. We believe a good product sells on its own if it meets the needs of their users, and so far it is meeting ours. The robots subscription fees will honestly be marginal compared to our trading P/L bottom line. If it becomes a management burden or encounters too many pervert feedback effects, we can always limit the number of subscribers giving priority to the “early birds”.
3. One can never know for sure. The 20 DMF component of the robots is publicly available, but it is indeed crucial that we keep the risk management system undisclosed and proprietary. That’s why we don’t provide backtest results on the true robot signals, but only on the system before risk management. Reverse engineering is impossible with data publicly available. Also, our intention is to continually improve the algorithms over time and to switch ASAP to a real-time system instead of EOD to stay ahead of any reverse engineers.
Billy