The debugging continues. As with any computer program, every time it seems you have it fixed, another bug arises. We thought we had it 100% ready last night, ran it live on all 13 pairs, and it had problems closing out trades. The entries were executed perfectly but the exit orders were not always accepted by the broker's server. We believe we have identified the problem, made the necessary changes, and now will see how it performs tonight.
The autotrader performs perfectly in backtesting. Because it does not require live server communication it is now completely free of errors when backtesting. So, we have begun the work of backtesting the last 6 months to get a feel for how accurate the system is across multiple months and market types. We already have identified a concern during ranging months that could hurt its ability to be consistent. I have an idea that I believe will allow it to make money even during those periods. We have created an extension to the autotrader to test this theory and if it proves to be worthwhile, I will unveil it to the group.
Because of the massive processing power it is taking to run the autotrader on 13 pairs across multiple timeframes 24/7, we are experiencing some drag on order processing and entries. We are weighing our options and considering placing the autotrader on a high power server located on the same network as the broker we will be trading through. It is either that, or network multiple computers together to utilize the processing power of the computers collectively.