Profitable, tested mechanical trading systems THIS THREAD IS CLOSED AND REQUEST TO DELETE IT HAS BEEN SENT I would like to start a thread that is dedicated to testable, mechanical trading systems that really work. With tested results for each system. The definition of a "mechanical system" in this thread is that the system must not have any "discretionary" element, ie. it must be possible to program the system as a code and it must be positive tested on about 20 years of data. 1990+ is good because that's the time computers started to get really involved in trading and many of the existing methods changed their behavior for worse. Results must be posted to all major pairs (EURUSD, USDJPY, GBPUSD, USDCHF) and possible all the minor or "commodity" pairs as well (AUD/USD, NZD/USD, USD/CAD). That covers most of the volume of the market. Other pairs tend to be just aggregated moves of these pairs ie, technical analysis works worse with them. Begin amount for testing is $10,000. For position sizing we'll use risk of 3% based on initial capital (10,000 usd) - In other words if trade hits initial SL, the loss must be maximum of 3% of initial deposit. If the position size would be calculated to current account balance, the results would be better on profitable systems. However the results would always weight the later data to be more important than the begin data. I open this thread mainly to gain knowledge about different methods that work so it would be possible to evaluate what kind of systems work on longer term, ie. are they trend systems, countertrend systems. If you have a system that you would like to contribute, please post it. Note that the system must be mechanically testable per abpove rules. If you can't code the system (can't program etc) that you know to be a profitable system into a testable program by yourself, feel free to contact me. I'll do the programming (I can't promise any schedules how fast I can do it) and post the tested results + the system to this thread. I start the thread with one actually profitable system. It's my personal modification of the turtle system, in fact it's even simpler than turtle original turtle system was. Note that all the daily bar tests below are made with a very crude control points method because I don't have tick data for all the currencies from 1990+. If someone wants to help with the testing by pointing to a better trading platform than mt4 (I'm looking into rightedge at the moment but the problem is forex data), I would be very thankful. Warning disclaimer: Backtesting does not necessarily reflect future results. You must understand risks and whys before you can trust any backtesting results. Also, I don't promise that my tests do not contain bugs, they might very well do. Don't trust my results - Do the tests yourself. Last edited by mikkom, Feb 9, 2009 3:22pm |
#2 ![]() | |||
| |||
![]() The following are testing results for a simple donchian breakout system that opens position when price penetrates high/low of last 4 weeks. Trades are opened every day when no orders already exist with stop orders at high/low of last 4 weekly bars (longs are opened at exact high of last 4 weeks, shorts exact low of last 4 weeks) and always uses sl of last weeks high/low. SL is trailed on high/low of last week and never moved down (on longs) or up (shorts). ie, On each monday, sl is moved to low of last week (longs) if low of last week is higher than previous SL. In same way, SL is moved down (shorts) to last weeks high if last weeks high is lower than current SL. When one position is opened, no more positions will be opened until the open position is closed. Results from 1990-1-1 to 2008-12-1 (edited to use geometrical MM, results use 3% risk of initial deposit for each trade). Results under each currency in form of profit/profit factor/max dd% Majors EURUSD 8637/1.44/12.29% GBPUSD 8121/1.41/25.16% USDCHF 3815/1.19/16.74% USDJPY 8313/1.63/9.61% Minors AUDUSD 363/1.02/40.34% NZDUSD 6310/1.28/33.39% USDCAD 4694/1.21/22.80% As you see, this system works well with more liquid pairs which seems to be the usual with most of the mechanical systems. On the test period, this system would have made (with fixed lot of 0.1) profit of $40253 (not including initial deposit of $10,000) Below is visual stimulation from EURUSD, this is a screenshot from the tester screen. Last edited by mikkom, Dec 6, 2008 7:02pm |
#3 ![]() | |||
| |||
![]() Here are testing results for other system called FirstStrike Plus. The system is created by a veteran trader who is also a member of this forum, TheRealThing. The full system can be downloaded from his blog at http://infiniteyieldforex.blogspot.com. the trades are again made with a risk on 3% for each trade. (profit/pf/max DD) EURUSD 197527/1.68/14.04% GBPUSD 72395/1.24/33.5% USDCHF 91979/1.34/27.91% USDJPY 27730/1.11/57.43% minors: AUDUSD 28527/1.09/77.36% NZDUSD -9825.92/0.5/98.83% USDCAD -7371/0.91/98.3% This method drains all of the account on few commodity pairs but works very well on most of the pairs. In fact the total result was very surprisingly high. Total profit for the period was $400,962 or in words: insane. The result might be so well because control points testing method does not work very reliably with tighter sl's OR the method might just work this well. I have to admit that the method worked so well that I had to try it on longer periods and on USDCHF from 1980-1990 it almost drained the account and on GBPUSD on the same range, max DD was 93%. On sidenote, the trading environment changed about at that time so these might be too old results to count in (or then not). The 4/1 week donchian method mentioned in last post worked multiple times better before 1990 than it worked after 1990. Image from the tester below Last edited by mikkom, Dec 6, 2008 9:22pm |
#4 ![]() | |||
| |||
![]() Here are the results for "One night stand" method from TheRealOne's site at http://www.infiniteyield.com/ At this point I have to warmly thank TheRealOne, he is one of the only people who I know that shared an actual, mechanical, working trading systems. This one is a bit trickier to test with geometrical MM because no SL is used. Therefore I have used a geometrical position sizing of 3% based on fictional sl of open +- atr of 2 periods. This might be correct or incorrect way to calculate max risk but it's the best way I can think of at the moment. Note that while position size is calculated this way, no SL is used in trades as said in the rules. Results: EURUSD 4556/1.46/9.53% GBPUSD 2353/1.27/25.53% USDCHF 2732/1.3/10.71% USDJPY 3378/1.58/6.48% minors AUDUSD 385/1.06/18.55% NZDUSD -3326/0.7/43.28 USDCAD -1902/0.82/36.13% As usually, this method works much worse on minor pairs. In fact the profit would have been much better on just major pairs (profit for minor pairs is $-4843) Total profit for the period: $8176 Last edited by mikkom, Dec 7, 2008 5:06pm |
#5 ![]() | |||
| |||
![]() I got pm about the results posted here. I'll answer the questions here to clarify things for other people too. All results are in dollars, not pips. There is a very good reason for this, tighter sl should mean more risk/reward and results in pips don't tell about r/r as much as do results in actual dollars where trade size is weighted by good predefined rules. With pips, it's also impossible to know what the exact risk is because you still have one element (position sizing) missing from the system. For example by weighting certain trades by using bigger lot size for them, you would get totally different results from the above. Even using fixed lot would mean skew the results because risk for each trade would be different - for tighter stoploss orders trade risk calculated by account balance would be much less than for those with larger stop. All the results are for the full test period for each currency from 1990-1-1 to 2008-12-1. One interesting point to note is that ALL of the above systems open positions by stop orders - that is something I already guessed but it's interesting to see if anyone can post a profitable system that doesn't use stop orders - in other words does not buy strength / sell weakness. It's also interesting to note that not one of the above systems uses "indicators" of any kind. (If high of X bars isn't considered an indicator of course..) I'm not going to post any more systems after these - I'm waiting for contributions from other members. Feel free to post here or contact by PM if you know of a system should be included here - it doesn't have to be your system, a link to a system definition is as good as a system that you have made yourself. I want to see if anyone else is willing to give away their secrets about working systems because I know how rare and valuable this information is. Someone might be puzzled why I decided to post [some of] my findings here, it's for two reasons. First of all, I think it's very important to see what kind of methods really work based on longer term testing and to promote longer term testing. Some people think that a system is profitable when they have traded with it for a year with one pair. It really doesn't tell you anything. Second and most importantly (yes I have selfish reasons too), I want to hear from other people who know of the other pieces of puzzle that I'm missing myself. Last edited by mikkom, Dec 7, 2008 8:34pm |
#6 ![]() | ||||
| ||||
![]() Hi Mike.... Not sure why you want to close this thread .... you put a lot of work into it and I just stumbled onto it myself..... I have been watching and communicating with Joel for about a year, he truly is a great guy and what he has given to us is very effective if we follow the rules, but that is true with any workable system... you must be consistent. anyway I hope you are doing well and feel free to contact me anytime. Regards, johnedoe __________________ Same Whore .... Different Dress |
#7 ![]() | |||
| |||
![]() Quote:
I got banned for a month after working for a day to post the above posts because for some reason moderators looked at the posts and decided for some reason that I don't know that this thread is spam. After I sent a clarification via contact form (I couldn't even post PM to moderators) that they should look at the links and not randomly ban people who post something I got my posting rights back. I might do a similar thing in the future somewhere else (maybe I'll put up a website - posting graphs and so on is easier that way) but for now I'll close this thread. |
#8 ![]() | |||
| |||
![]() Quote:
You posted a link to Joel's commercial web site instead of to his free blog site. It is th blog site which has the details of the FirstStrike and OneNight Stand systems. Ian |
#9 ![]() | |||
| |||
![]() Quote:
Also I'm not sure why that page is commercial page because there is not any purchase link or anything being sold there - sure it looks like one of those "we'll make you rich, pay $999" but it really is just a text document that doesn't sell anything (or then I just didnät find it). |
联系客服