The strategy behind all this is based on a successful discretionary technique I learned a couple years ago. I didn't want to use it at first because so many other ideas were more interesting to me, but they have all been disqualified, over time, in live discretionary trading. I'm ready to go live now. My current employer will be making that as difficult as possible, unfortunately. No personal computers or even cell phones allowed inside the plant. Check out my thread in career trader "trend following + range trading" for the first iteration of the automated system, SPY being the first victim. That system is up over 10% since I posted that thread. I paper traded it at my last job and watched it execute exactly per the simulation. The main risks that worry me are connection and data issues, the need for an equity curve signal to switch the strats on / off based on performance, overnight moves are always potentially dangerous, etc. And of course, the risk of me interfering with the programs out of doubt. But the strat itself is the best thing I've ever tried and the only thing that has made money for me consistently when discretionary trading on a couple different time frames, many different stocks and commodity etf's. I also watched one of my early subscription services demonstrate it, to my dismay, pulling in win rates of 70-90% with roughly 3-R, as I got spanked continually with fundamental (emotional) trading. It is based on two indicators. And each strat has between six and nine degrees of freedom, depending if you count your risk and trailing stops. I've also seen a slight variation of the strat presented in Trading Systems by Jaeckle and Tomasini, where they walk you through (pun intended) the entire design and optimization process, including out of sample tests, monte carlo and parameter sensitivity analysis. It's not a new idea.
Instead of you selecting what to show, maybe it would be better to have someone suggest a ticker. Out of 7,000 stocks, some will perform well with your method. Why don't you try your method on all DJ30 stocks and tell us the results. Let us see if you can beat a monkey...
Actually this is a good suggestion to try on all DJ30 stocks and see the overall performance. The scientific way to test it would be to construct portfolio of 20/30/40 stocks randomly choosen from S&P500 stocks and see out of 100 such portfolios constructed, how many perform well. If you can get 75 or more portfolios with PF > 1.5 (taking care of execution and slippage cost), it will be a very good system to trade going forward. I mean to say that system will pass the scientific test that it applies to many stocks and is not curvefit to specific stocks. Ofcourse, feel free to adjust parameters of the system to individual stocks, as individual stocks behavior would be a bit different from each other. Btw, can you even do such a thing as above in Portfolio Maestro, or is it too hard ?
I actually did that over a year ago - tested the strat on all Dow stocks going back 10+ years. Not all names made money, but most did. And overall it came out way ahead. This was before any optimization.
How naive are the professional fund managers with 10 year track records who use the same approach? http://www.tradingblox.com/forum/viewtopic.php?t=7033
it is always too little fitting which is a huge problem, not too much fitting. for example: holding a trade for a 900 tick loss, instead of 43 tick loss, because "50 tick stoploss is just a curve fitting!" or trading natural gas together with German Bunds on 2 day bars, instead of 15 min bars for NG and 60min bars for FGBL, but no, it would be fitting to the curve. Perfect, I pray to the Lord, the more traders with this kind of thinking, the better for me And please, do not ever use tradestation. better do mean reversion or relative value models in matlab
I'm pretty sure it can. It takes your data from the main console and runs a back test on whatever you give it. So if you have an FX data feed on the 9.1 platform, Maestro should be able to do its thing just the same. I think FX pairs are a great application for this type of trading because they are less correlated than individual stocks.