I've written my own autotrading system which shows positive expectancy on testing and I am relatively new at trading it. A couple of days ago I had it supposedly sitting on standby so as not to take any trades and during the day while at work the system placed a long trade at the very top of the day. As I was leaving work I received a phone call so instead of driving, sat and chatted for 30 minutes making me even later getting home. When I walked in the door I was down $1500. Man was I shitted to put it mildly. There was a small glitch in my formula which I hadn't spotted in spite of all my backtesting. Conclussion: Several, but one which seems to come back at me often is that errors in trading are a large reason for many of our losses. Funny how errors often cause a loss and not a win. Anyhow, the loss really shook me up, I tried revenge trading during the night, all night in an attempt to regain my loss. Result;another smaller loss and totally buggered at work the next day. Went to bed at 6:30PM last night to recoup my energy. Just thought to share, I hope to learn from this and remain positive. Ps: Don't tell me about revenge trading, I know I was an idiot but no pain, no gain (in the lesson department of the brain)
The system design is great imo. I have a piece of the formula which is an emergency kill switch 'go flat' It takes me out of a trade but what I wasn't aware of it it had a hole in the formula which allowed a trade in. Took a minute to find and a minute to recode and fix, but 'oh the pain'. I had the formula running during the day with kill switch on, alas....
welcome to software. this might help http://www.murphys-laws.com/ dont ass-u-me. turn off anything you dont need
Neither....just lack of robust testing. Also, having a "go flat" kill switch is not enough. The switch has got to also cancel all open orders, and stop all future new entries.
I'm not a system trader but it would seem simple enough to rig up an email connection to alert you of trades unless you are doing hundreds a day.
You don't mention a "forward test" ... if you are not doing these, add them to your process before you trade "live" ... and run them for long enough to ensure you can build a decent picture of how the strat behaves against real data ...
+1. So true, and even my implementation of this is half-assed. I rely on my broker's risk management functions too much. This thread should inspire me to really clean up some of these issues. (One frustration I have as an individual trader is that I have to watch the trading and then go program when I have the energy to do it. To test new features (even safety features) is a chore. To keep a separate instance of my trading platform with all the same features, I have to shell out yet another platform fee + another round of subscriptions. So features get rolled out much more slowly than I'd care for, although I may just need to bite the bullet and pay out the fees. And most brokers just don't have decent off-hours testing facilities.)