Murray lernt fast here: http://www.elitetrader.com/vb/showthread.php?threadid=225189 that ET don't want nuttin' for free. A sponsor, a respected poster, and a successful entrepreneur wants to teach ET sumpin'? No way! We awreddy know it all! Arrogant fuckstick to think he can teach us something! So while Murray wounds his licks and gets up his courage again, I will imagine what he intended to say. To wit: Lissen up, dickheads! You think backtesting is easy? That your average ET assclown can do it? Well suivez-moi and lessee. You got a strategy idea? Who the fuck doesn't? What you don't got is a way to test it. You need a data divider. You need a chart disservice with built-in backtesting. You need a coding language. And you need a fucking brain, which mosta youse ain't got. (Ass-side: Murray provides all of the above, but I am perverse and use EasySignal for the following eggsamples.) Forget the strategy idea. First you have to build the code to represent that idea and to set up the built-in backtesting to test it. Here is what that looks like in EasySignal (ape-ologies to ES users who REALLY know what they are doing). We are going to test the lame-brained idea that entering on a retrace to the open is the road road to riches and bitches. //retrace to open system structure //variable declarations //function declarations //set up chart parameters //backtesting code { //determine the time //initialize variables //establish trade day validity //test for time window { //trade loop { //set up trade entry conditions //entry code //stop loss code //profit stop code //end of day exit code //profit calculation and "helper" code } } } (Lest you think I am winging it as I go, the code is all done and I am just building it up in baby steps that show what decisions you have to make in backtesting.) Flame away! Intelligent questions will be answered politely. Stupid questions will be derided mercilessly.
To give you an idea where we are going, the code to implement the simple hypothesis test "does price move tradably after retracing to the open?" requires 130 lines of fairly compact code in EasySignal's Formula Script (efs). How long does it take to learn to knock out code like that quickly every day so you can test a couple of brain farts every day? Don't ask. I imagine that the majority who consider it decide "I'll stick to intuitive trading."
Well, most likely it is a program you already had, but I'm sure if there are paramters involved it can produce profitable results. However, I doubt it will be robust, by definition any statistically signifcant strategy having at least 100 trades, and positive overall expectancy from pessimistic projections. What is your time period and how much data and what type of data do you plan to use?
Yes, the basic code has been working for years. I just plugged in tests for retrace to the open. You are correct, it doesn't work. I'm not stupid enough to post a working system. "Statistically significant?" I assert that there is absolutely no way to make any such determination in trading. Why 100 trades? (ET gets to decide how many in the example) Why not 100 OPPORTUNITIES to trade? What are pessimistic projections? (I may have to code that)
An explanation is in order. I picked retrace to the open for the hypothesis test because it illustrates a weakness in hypothesis formation. We tend to over-focus on dramatic events: "Oh fuck! Look at that! I knew I should have bought the open and now I have another chance!" The dramatic blinds us to the mundane. Also I picked it because it ALMOST works. "Almost works" tempts you to waste time devising ever more bizarre conditions where it MIGHT work.
Backtesting is a concept of the 1980s. It is a process full of pitfalls. It is ill-conceived at its root. I believe the first program that allowed traders to backtest somewhat efficiently was System Writer Plus by Omega Research, nowadays known as Tradestation Technologies. Also, around the same time people used desktops with DOS and 8086 processors. Anyone who is backtesting lives 30 years in the past. Technology has progressed and no wise trader wastes time to backtest anything, other than to develop final code for automating a strategy. Nowadays people use machine generation to find systems. I have explained all these in past posts. Some though like to live in the past. You cannot compete with a guy that can efficiently employ machine generation and develop hundreds of high probability strategies every single day. Eventually he will find some kind of an edge and take your money. http://www.elitetrader.com/vb/showthread.php?s=&postid=2968698#post2968698
Guess I'll just crawl in a hole and die, then. But before I do, let me suggest that humans code the strategy generators. And they are IMO unlikely to be looking for the absurd, the ridiculous. Also let me add that there are not hundreds of strategies to be found. So I guess those strategy generators found everything there is to find the first day running. The strategy generator in my head runs slow, but it is capable of things that a machine cannot yet do, like asking "WTF?"
You are overestimating your capabilities. It's like saying because people designed an airplace, you can do better flying. The strategy generators depend on input data and they have their own problems, like data snooping, fitting, etc. that should be dealt with but as more data are available, day after day, the potential increases. It is simply humanly impossible to run all the ideas you tried in the past on a backtester to see if significance changed due to new historical data. Hiowever, you are free to believe what you want. I just tried to warn you that backtesting idea after idea is like buying lottery tickets and waiting for the drawing. As we speak, I have three fast computers crunching numbers using one program that generates simple strategies. There would be no way to do this job even in a hundred years manually.
Perhaps I should concede something so we can get past it: yes, I am an arrogant prick full of overweening pride who is not nearly as smart as he thinks he is. And yes, I am using a slow old laptop running decade-old softaware on a thirty-year-old paradigm. That conceded, let us define our respective playing fields. I trade one index instrument. I watch it all day on a one-second chart. I do not identify strategies using the computer hardware. I identify them using my wetware computer. After years of this, I have seen it all. And gradually tested most of it. In my sandbox, there ain't many turds to sift out. The same big dogs come and crap in it every day.