Just my experiences so far..
Posted: Wed Jan 05, 2005 10:30 am
Up to November of last year, I have been developing systems and back testing by hand using historical charts. Not the easiest of tasks, although certainly pretty laborious! The charts I referred to were both end-of-day and intraday (in order to check if positions were stopped out and had to be reentered the same day).
In November, I wanted to see if I could develop any worthwhile day trading strategies. Since I knew this would involve far more data than eod data, and I didn't really want to screw up my eyes looking at intraday charts, I thought I would enlist the power of computers (welcome to the 21st century!) I wasn't too keen on learning how to program, so I thought I might be able to use Excel for testing since I am pretty familiar with it. I did manage to 'code' a system up just by using functions and logic rather than VB script. However, this took absolutely ages to do and also I found it was very restricting. I can change parameters and see what the results would be but to create another system would probably mean another couple of weeks to 'code' in to Excel.
So I have now (a little begrudgingly-I don't really want to learn how to program, but may have to) started to look at commercial back testing platforms. I have now downloaded the trial version of Amibroker and have managed, within a day, to code my existing systems into it! The trial version cannot deal with intraday data, but I am assured by the developer that the Professional version will be able to do this and so I am seriously considering purchasing that.
I am also considering Veritrader 2.0 since apparently I will be able to code my own intraday systems in. One thing I noticed about the current version and previous versions of Veritrader is that it doesn't allow intraday data. I thought this odd since one of the main features of Veritrader is to test the Turtle system. If I recall correctly, exits using the alternate 'whipsaw' method are engaged at 1/2N away from the entry. Since N, in terms of market movement, refers to (approximately) the movement in a day, wouldn't it follow that on the day that a breakout occurred, a position may be entered and then exited (at 1/2N) and then reentered and exited a number of times? If only end of day data was being used, this would be impossible to detect and test . I would be grateful for comments on this (apologies in advance if addressed elsewhere).
In November, I wanted to see if I could develop any worthwhile day trading strategies. Since I knew this would involve far more data than eod data, and I didn't really want to screw up my eyes looking at intraday charts, I thought I would enlist the power of computers (welcome to the 21st century!) I wasn't too keen on learning how to program, so I thought I might be able to use Excel for testing since I am pretty familiar with it. I did manage to 'code' a system up just by using functions and logic rather than VB script. However, this took absolutely ages to do and also I found it was very restricting. I can change parameters and see what the results would be but to create another system would probably mean another couple of weeks to 'code' in to Excel.
So I have now (a little begrudgingly-I don't really want to learn how to program, but may have to) started to look at commercial back testing platforms. I have now downloaded the trial version of Amibroker and have managed, within a day, to code my existing systems into it! The trial version cannot deal with intraday data, but I am assured by the developer that the Professional version will be able to do this and so I am seriously considering purchasing that.
I am also considering Veritrader 2.0 since apparently I will be able to code my own intraday systems in. One thing I noticed about the current version and previous versions of Veritrader is that it doesn't allow intraday data. I thought this odd since one of the main features of Veritrader is to test the Turtle system. If I recall correctly, exits using the alternate 'whipsaw' method are engaged at 1/2N away from the entry. Since N, in terms of market movement, refers to (approximately) the movement in a day, wouldn't it follow that on the day that a breakout occurred, a position may be entered and then exited (at 1/2N) and then reentered and exited a number of times? If only end of day data was being used, this would be impossible to detect and test . I would be grateful for comments on this (apologies in advance if addressed elsewhere).