Hey I've got a few comments and questions while on the topic of backtesting. First off I want to say that in the begining backtesting as a useful tool made sense to me. As I started to try and understand the science behind market fluctuations I eventually came to the conclusion that back testing is not viable. I will say though that I've never met anyone who has proved to me that it does or doesn't work. From a probabilistic point of view I read that it is not unlikely for a person to use indicators, mathmatical algorithms and statistics to find a way that backtests profitably. But the illusion here is that after so much time of adjusting parameters and values it is entirely possible to "custom fit" your system to past market data. This sort of "fit" is coincidence and has no correlation with new data. Have I tried this? No. Secondly I was wondering if anyone has studied chaos theory and fractal geometry. To me this is very interesting. Aka sensitive dependence on initial conditions. A small signal can amplify into a totally unpredictable large output. I have yet to relate this to human emotions and market fluctuations but it seems as though as with the weather and its unpredictability the market may operate in a similar way. But unpredictability just doesn't seem like the right word. I know what a piece of wood looks like before looking at it even though the pattern of grain is unique in every piece. I know what market fluctuations look like and are going to look like even though I can't "predict" what will happen tomorrow. So what is it that I know? It seems that there is a non-linearity between back testing and market fluctuations. I guess I am trying to say that back testing is a known, predictive, defined pattern that repeats where market fluctuations do none of the above. The market may keep a certain "similarity" at all times but its never the same. How can they work in harmony? I hope I am making some sort of sense. This is where I am at now and trying to make sense of it all. -Ray
Well first off I do not know who that is. Second, I don't believe that the way to profitable trading is to see who made the most returns and try to mimic them. Buffet, James Simmons, George Soros, Jim Rogers... I do not think they ever envied someone, bought all of their books and became who they are. If their is anything to learn from successful traders I would think it would be how they cultivate and utilize their own personal qualities. Its their drive, inquisitive minds and enthusiasm in understanding trading, business, math, science and what ever else is applicable that got them where they are. Lastly, the word "backtest" is kind of vague. If someone deems backtesting viable, unless they map it out for us, how can anyone have the slightest idea of which kind of tools they use and how they are using them? If someone claims that backtesting works and they are profitable I would definitley rule out the possibility that they messed around with a few indicators till it matched up with past data. I would have to think that they obviously know something and are using something that no one has ever thought of.
I did a lot of backtesting over a few years... eventually I did come up with things that were way more suitable for a computer to do than a human. I did discover mathematical realities that led to promising stuff. A couple of different realities led me to set that aside and start learning to screen trade. One is that my wife is about ready to do a murder suicide if I don't get some income... Finding things that are not really random is the key. Most of the stuff that comes with most software packages is entirely random. The problem is that you can optimize it to appear to not be random. You can busy yourself for a long time doing that. People say to test it on one sample and retest it on another... let us say it's really random but you tweak things so you have a winning strategy on the first sample. If it's random then there is a 50 /50 chance it will work on the second sample... so you see these threads where the guy says "this works really good in the sample data and not quite as good in the out of sample data, I'm taking it live tomorrow".. then they fiddle with it for a week or two while it systematically destroys their account... When I started screen trading I tried to manually trade the tool set I developed for automation but that was clunky and very difficult for a human to do with the available trading stuff to work with, besides it wasn't that much fun, so slowly I morphed it into stuff easily doable by a human. I'm terrible at implementing all my ideas in realtime so far but it's looking more promising every day, becoming sort of easy but I make major screw ups sometimes... I learn a lot some days though.. after I recapped todays work after hours I realized that I could document every scenario that I might encounter all day and document what action to take... this was the first day I could think of doing that really after about nine weeks on the screens... Would I automate that trend following routine? Not unless the machine centric ideas I have don't prove out... would I do all my development with a backtester? No, it's too sterile of an environment, I got more development done in realtime the last few weeks than the typical year of back testing... So there are two approaches to automation, one would be the expert system where you just program the machine to do what an expert would do, the other is to dig in and find stuff that a machine can do better and find something that is provably not random. You should prove the non randomness before backtesting really, because you can't prove non randomness with backtesting.
That would work out to 32.6% annualized per year. Or assuming 250 trading days a year, 0.13% per day. If a person only made one point per day in the ES, with no compounding or additional contracts, based on $2K starting margin, it works out to 625% return in one year. $50 per point X 250 = $12,500. ((Ending account value/Starting account value)-1)*100 = ROR% (($14,500/$2000)-1)*100 = 625% Over 18 years that's an 11,250% return, non-compounded. At some point you might decide to add a contract or two to boost returns. Either do that or, if you can manage 2 points a day, you're a market wizard by Eckhardt standards. That doesn't account for taxes. Fun with numbers.
Trader666- Are you sure that you are not ION re-incarnated? You make Mr. E look like Homer. http://www.pitt.edu/~kis23/ION.pdf Read the text a few times to get the real meaning of what Socrates is saying-if you need some help then please ask-for that is what we are here for. When you "see" the answer-you will then realize how futile your efforts are-same goes for any average trader who thinks that backtesting is of any "real" use in trading. The General
mynd66- Like Eight-you are embarking on the wrong journey. What you need to do is forget about all that you have "thought" you have learned-for it is all rubbish. I am here to help those who want some guidance-but I do not entertain those who are not willing to see what is what. Ask what you may-but I will tell you now-before I even start-that you will not like the answers-but that is a good sign-for when you hear what you want to hear-that is a warning sign. Fractal Geometry is for classrooms-not for trading the markets. The General
You don't know who William Eckhardt is? That explains it then... you're posting here without even reading the prior posts. And you don't know what you don't know. You wrote before that you "eventually came to the conclusion that back testing is not viable" apparently at least in part because you'd "read" about the optimization trap. Which is a beginner's trap. So you're coming to conclusions about things you don't understand. My point was, if I knew almost nothing about a subject and then someone like Eckhardt endorsed it, that would give me pause.
Very few programmers actually know how to programme correctly-right now I am looking at B&S programmes kicking in and out-and most traders don't even know why it happens. Inspiration can be very dangerous-but only if you let it be. http://pm1.bu.edu/~tt/neat/geller.pdf The General
Because YOUR efforts have been futile doesn't mean mine or anyone else's are. I'm sure you can't pitch a 100 mph fastball either. And I doubt you can come up with an intelligent thought on trading. So far you've said backtesting doesn't work and you think Jack Hershey knows what he's talking about First, this is not what backtesting is. Second, this implies that charts of past data are irrelevant which is the height of stupidity.