This seems contradictory to me. If backtesting cannot integrate descretion how does it elimate systems that do not work when they require some descretion?
Ill try to explain. On a historical chart the bar moves above the moving average and goes a couple of points against you then reverses and closes just above the bar. Looking backwards the close was profitable but in live trading you dont know that it is going to reverse and exit the trade. If you try to set the exit as it crosses you eliminate all the profitable trades when it reverses back below the ma and closes below it and continues the trend. This was my experience. Perhaps there is a way to cover every possibility but I couldnt find it. With a bit of patience waiting for those obvious trending situations the system works fine but so will almost anything else.
Wrong . . . good thing your investment was only 2 cents. I trade manually and am consistently profitable. I have an automated version of what I do manually and it is profitable as well. Anything based on discretion or subjectivity will always have a problem being profitable but not all automated systems are built on such a wobbly and unstable foundation. You need to narrow the stroke of your statement and widen your vision.
Actually, that's a good way to study, learn, study, learn, study, learn and then study and learn some more ... eventually you should arrive at being consistently profitable if: (i) you keep applying yourself and you find what works for you, or (ii) you figure-out that you can't figure it out on your own, but are able to put enough pieces together by working at it diligently with other traders But I also see what you're saying here Prof; ... and IMHO that's because you're a trader who knows how to put together an automated system, not a programmer who is trying to put together a trading system. Good trading, JJ
I don't know what you are saying. If your strategy is to take the signal on a bar close, then your back test should do the same, as BigBubba suggested. If your strategy is to take the signal intra-bar, then again, your back test should do the same. So, what is the problem?
I want to thank you guys for making me rehash this. I went back to recreate the system to see if I could illustrate what I meant but in the process I found out something that I did not see the first time around. Now Im all excited again. I want to see the results of this before I say any more. Its probably another fake out but you never know.
Ok, I tested this as it is dead simple to do. I tested it on the DOW stocks from Jan 1 2001 - Jan 1, 2006. I didn't test 2006 because if it was marginally profitable, i would have modified it and used the last year as an out-of-sample forward test. Anyway, my testing showed 36% of trades to be profitable with a win/loss ratio of 1.53: 1 and a profit factor of 0.86. I used a fixed position of 100 shares. The win percent isn't a shock as this is typical of trend-following systems and this is really an ultra-short term turtle strategy. There's nothing wrong with simple strategies though - look at the good results Art Collins has shown in his journal.
.... which means of course that the REVERSE strategy might be a good basis for a system. I tested the original strategy on a stop-and -reverse basis. Food for thought: if an idea fails miserably, turn it around and it might work.