From "Trading as a Business", by Charlie Wright (http://www.elitetrader.com/tr/index.cfm?s=17), I read that he uses Tradestation and EasyLanguage to backtest his strategies. However, I'm with Interactive Broker, and I don't think I can switch to Tradestation that easily since I live in Canada. I wonder whether there are any ways (as cheap as possible) for me to backtest my trading strategies ? Anyone with IB who does backtesting, please give me some hints.
Yea, i have thought about excel, it's doable and I will simply collect my own data and buy some. But if EasyLanguages seem to make everything easier, I will see whether I can purchase something like that and capable of reading data on excel sheet.
You will probably find that backtesting is not a viable means for finding anything out. The reason why is fairly straightforward to determine. If you are backtesting something and you feel you have grasp the thing you are backtesting, then you have failed to really grasp what (the thing) is going on. Here is an example: I prepared a mechanical method for trading DJXX (long before the emini's were invented). The CRT display takes any person through the trading and suggests the prices and stops for long or short trades and signals the moments to complete the actions. It is simply a signal screen that looks like a RR block control display. The description is a set of diagrams with links and nodes. Each link is named and the nodes are "containers" of functions that can be coded with a few pages of coding each. 1 page is the screen display with a reference to the page the signal comes from. Two pages make up the general 7 sections of the logic. The remaining 30 pages of nodes and their links take the details down to a level where the nodes are "content" describable for a programmer to code. There are six levels of sentiment, one of which is always on display; five levels of market pace are depicted as well. Two pages of links and nodes are required for the sentiment and it is the same for the market pace. These are displayed just as a side issue; their main purpose is to provide "gating" signals to other parts of the system, primarily. These are only "secondary" gating sets, the more complicated aspects require primary more complex gating. Operating the system manually makes a multiple of the market's daily range, daily. Coincidentally, I reviewed these pages with a programmer a few days ago since I wanted to update it to include the eminis and their relationships to the cash markets. At one point while looking at the sentiment in which I used three levels of regression (the "pin wheel"), I asked the programmer jokingly, if he thought anyone could backtest my trading if I printed the results of using the system. His response was what I am trying to get across to you. If a person grasps how the markets work and is able to reduce it to a mechanical process and use that process very successfully, then by the nature of that process of making money, it cannot be backtested. there is another thread going on where a person is asking about how to be sure a system is going to work for a fund to help the fund trade. In that thread it is spoken, mistakenly, that a system for a fund could expect to work as well as a system used by a trader (i.e., it would have a 90% chance of failure as do traders). Others in the thread commiserated and provided commentary on what to do to preclude failure. All of this was couched in testing parlance. Roughly speaking, it was all a joke by well intentioned folks. There is a dilemma regarding backtesting. It goes like I have said above. If you can successfully test what you have comeup with as a trading something or other, then the something or other is not really a viable trading approach that takes out of the market the potential that is there. Fortunately, I am going to make my mechanical systems public to others. This will introduce a rather large reservoir of coding that has not been seen before in the financial industry. Its effect as it is used more and more widely will be to smooth the markets if it ever gets to a sufficient level of usage and it will smother a vast array of edges now in use(up to about 80 that I am aware of). What you may read into this is that there is a lot of information and bandwidth in market data. (See Braithwaite). If you follow the "big Bangers", you already know that a lot of determinations are made about celestial time by regarding the signal levels in the microwave range that have been bouncing around for billions of years. The markets are the same, regarding the confluence of trading methods and their relative influence according to their strength (size) and signatures. Recently, I went through some preliminary discussion on the recognition of range breakouts that others were mistaking for the usual and continual sophomoric plays seen on level II. None of these people have a remote idea of how filtering works in the financial industry. I've noticed over the last 15 years that when my prints are backtested, there is always a range of results that are spread over more than one order of magnitude. the tests are invariably on one side of the reality and they never come remotely close to what is actually going on. This astounding consesus of being all over the ball park can only be corrected in one way. By using the mechanical version of the manual method and making a lot of money at the same time. By making it public, I feel that its transference can also be tested. I know these tests will not be valid since they will only be done by the A people and not the B people.
Jack, I've read the LevII posts you are referring to and found them very enlightening, thank you. What do you mean here by "filtering"? I'd appreciate if you could give an explanation as it is a broad term to google and Larry Harris doesn't have it in his index. Thanks
Youâll better google to find a program for backtesting. Uisng such software program you can create, backtest and fine-tune your systems and then just send orders manually to IB or use the software broker API (is such is available).