Hi, thanks for the reply. To answer your questions: 1) Trading with Zero Edge = Simulated trades in a synthetic time series that is totally random, ie: where there is no possible edge. All trades should therefore have a zero expectation. 2) Trading with a Positive Edge = Simulated trades in a synthetic time series that has an upward bias, thus all (long) trades have a positive edge / expectation. 3) Trading with a Negative Edge = Simulated trades in a synthetic time series that has a downward bias/drift, thus all (long) trades have a negative edge / expectation. If you need more info, just let me know and I'll try and elaborate further if I can.
Hi, I have something of a problem with the implied premise of these tests. That is, the "edge" of the entry is independent of the exit. I cannot see how that can ever be true. Therefore, exits matter - alot. Stops do little other than helping to delimit risk. I would also add that a contrived scaling out strategy (or any other) makes no sense if not based on the characteristics of the trade distribution. Note that i am not advocating one thing or the other. So to me, while the OP's effort is to be commended I do wonder whether he may in fact simply be confirming his own beliefs/prejudices. Thx D
@jazzguysoca I had to ponder this for a minute - but I get it - never thought of testing a system this way.... ...Damn, you all are good!
Thanks for the reply, ucf. I actually have that issue sitting here on my desk and have read the article. Its actually one of their better articles, IMHO. If memory serves, they tested a random entry with various indicator-based algorithmic stops, but found that a ATR-based trailing stop outperformed. Its interesting that they did not try a time-stop as a base line (I think they did try a random exit, though). For some reason, they used sequential entries, spaced by random time periods rather than just choosing entry points at random in a more scattered Monte Carlo fashion. I think their sequential sampling method may add an unintended bias to their results. But their results were certainly interesting. My tests were much simpler as I was just attempting to see how simple exits that use no prior history would perform on a relative basis.
That's actually a very valid point. The tests definitely assume that exits are 100% decoupled from entries. Plus, the exit strategies tested were purposely selected to be very simple and not based in any way on prior price history, before or after the entry. But I'm not sure this completely invalidates the results. There's a million ways to scale out - I just chose one of the more common ones to see how it did. However, I suspect that no matter how you sliced it, any scaling out technique would behave the same way (ie: trading absolute returns for lower volatility and smoother equity curves), at least in the simplified synthetic markets used in the tests. I have to agree with Buy1Sell2: You can definitely make money scaling out, but you'll make more money on an absolute basis dumping your whole load at maturity. However, you'll pay for it with more volatility in your equity curves. There's no free lunch.
Thanks for the reply and comments gusta. (I'm not sure how good I am today - I'm currently getting spanked. Will someone please remind the Nas100 to remember to mean revert after the Fed meeting?) Anyway, assuming you're just getting started backtesting systems, here's some book recommendations for you: Aronson's "Evidence-Based Technical Analysis" Fogel's "How to Solve It: Modern Heuristics" Faith's "Way of the Turtle" Brabazon's "Biologically Inspired Algorithms for Financial Modeling" Chen's "Computational Intelligence in Economics and Finance" The first one will save you a shit load of money; the others are a good source for optimization techniques and trading ideas.
@jazzguysoca Actually I have a lot of experience with back testing - Wrote my own desk top trading app, back tester, optimizer,yada,yada...(took three years - but it was worth it). I never thought about using completely synthetic time series - I like that! As far as the books you recommend, I will definitely consider reading them. Thanks Again.
Well, I guess the meeting was productive. What happened, exactly? Is Ben out there in his chopper dropping bundles of newly printed hundies over the masses? (sorry, I'm talking to myself again - I'll stop now)
Sorry, my bad. But very cool - I've made a similar journey (I've been at it for ten years though - I'm a slow learner). Sounds like you already know your stuff, but if you haven't already, check out "Biologically Inspired Algorithms". I came across it about five years ago and it really blew my mind. Unlike most trading books, its very short and to the point (very high info/bs ratio).
@jazzguysoca Tens Years Huh... I have 40 years developing apps - the experience does help. I can give you some pointers, possibly some code examples as well - to help if you need it. gusokeiff@gmail.com "Biologically Inspired Algorithms" Another interesting book I will look into.