halcyonn Posted May 31, 2010 Report Share Posted May 31, 2010 okay iwjw,many thanx,my skills at backtesting are no good,i would wait till we reach backtest and trade stage. thanx for the patience and your kind replies. Quote Link to comment Share on other sites More sharing options...
⭐ atto Posted May 31, 2010 Report Share Posted May 31, 2010 If I understand how the tester works, I'm pretty sure I did it correctly. Saved a .log of the signals from the Lindencourt indicator (the one you modified to do the log), and then ran a backtest with that. Put my indicator on, and it did not match up. Any thoughts about the fill issue I mentioned above? I think you're grabbing a completely unrealistic fill. What would be better is entering on the open of the NEXT bar, after the signal. Quote Link to comment Share on other sites More sharing options...
forexi Posted May 31, 2010 Author Report Share Posted May 31, 2010 Ok well I have done a backtest with this on open prices only and still get crazy profits ! Running it on Demo at Jade also will keep you up to date ! Quote Link to comment Share on other sites More sharing options...
iwjw Posted May 31, 2010 Report Share Posted May 31, 2010 (edited) I'm just running forexi's EA in combination with atto's indi in "real mode" Got a triple within 7 days testing May 2010 and then the account crashed. I hope that I'll find the mistake trying to test this system offline ...what looks too good to be true.... Edited May 31, 2010 by iwjw Quote Link to comment Share on other sites More sharing options...
⭐ atto Posted May 31, 2010 Report Share Posted May 31, 2010 Okay, this is what I mean. When you use the indicator, it records the time of the bar that the signal was generated on. So, if the signal was on at the close of the bar, it gets recorded (that's how MetaTrader backfills the signal, using close data only). However, the EA tester loads the times, and places the entry at the OPEN of the bar. This is a problem, because since this is a trend system, the open is almost always a better fill than the close. To illustrate my point, I changed the EA to initiate the order at the open of the next bar (which is generally within a pip of the close of the previous). Here are the two equity curves on Eur/Usd from mid January this year: http://i.imgur.com/OSupK.png Biiiiig difference (which is completely expected). So no, the EA is not as good as the backtest indicates. Yes, it looks like it still might be profitable, but we still have work to do. Quote Link to comment Share on other sites More sharing options...
forexi Posted June 1, 2010 Author Report Share Posted June 1, 2010 Ok here is what I picked up trading it Live :) yes yes I know but i am having fun :) I just got a SELL signal 2010.06.01 00:45;SELL The EA opened 4 trades all in profit and i closed them manually but hey it worked - woohoo ! $ 39.00 in the pocket. Quote Link to comment Share on other sites More sharing options...
forexi Posted June 1, 2010 Author Report Share Posted June 1, 2010 Ok here is what I picked up trading it Live :) yes yes I know but i am having fun :) I just got a SELL signal 2010.06.01 00:45;SELL The EA opened 4 trades all in profit and i closed them manually but hey it worked - woohoo ! $ 39.00 in the pocket. Quote Link to comment Share on other sites More sharing options...
saragosa Posted June 1, 2010 Report Share Posted June 1, 2010 Ok here is what I picked up trading it Live :) yes yes I know but i am having fun :) I just got a SELL signal 2010.06.01 00:45;SELL The EA opened 4 trades all in profit and i closed them manually but hey it worked - woohoo ! $ 39.00 in the pocket. can u explain what which ea did u use? what about setting thanks Quote Link to comment Share on other sites More sharing options...
iwjw Posted June 1, 2010 Report Share Posted June 1, 2010 ok guys, I'm stopping this "offline" backtesting here because it has nothing to do with reality. The signals generated after attaching the indi differ from those generated during live feed. Backtesting in real mode is a real pain...so I stopped this also...no chance to run the Optimizer for at least 2010 One thing left: if some1 has an account on a VPS Server it would be nice if he/she recorded the signals of the indicator for a month or so. Same procedure as for the backtest. You only have to set StartDate to the date when the recording begins. Then I could compare the life recording with the offline data for that period. Maybe I can find something Thanks in advance Quote Link to comment Share on other sites More sharing options...
Stormin_Norman Posted June 1, 2010 Report Share Posted June 1, 2010 the strange solution of the log files sounds like it is allowing the EA to 'see' into the future. i do not mind letting it test over an extended time on one of my mt4 instances. i am confused however at the different versions of the EA. could someone please point me to the most accurate version? the one which is slow at backtesting in the usual way? i will then have a look at it and add my suggestions. iwjw 1 Quote "It is inconceivable that anyone will divulge a truly effective get-rich scheme for the price of a book." Victor Niederhoffer (1943–), US hedge fund manager and statistician Link to comment Share on other sites More sharing options...
iwjw Posted June 1, 2010 Report Share Posted June 1, 2010 (edited) Hi Stormin_Norman, the offline backtest is not kinda looking into the future per se. The EA only knows the actual signal if Time[0]==timestamp from file. So this part should be ok I think. Would be nice if you could record the signals...thanks I forgot: the EA currently in "use" is the one used for backtesting...if attached to a chart it only reacts on the live feed. I would change the default MaxOrders from 5 to 1 because it doesn't make sense to open 5 orders on 1 signal with all of them having the same SL/TP/TS Edited June 1, 2010 by iwjw Quote Link to comment Share on other sites More sharing options...
mograst Posted June 1, 2010 Report Share Posted June 1, 2010 (edited) Can somone please posta all Neede files and EA again the way we can work together? There are the folliwng files missing in this thread to let the EA run: - LC-RSI.ex4 - FIP_Line.ex4 - LC-CCI.ex4 thanks Edited June 1, 2010 by mograst Quote Link to comment Share on other sites More sharing options...
Stormin_Norman Posted June 1, 2010 Report Share Posted June 1, 2010 Hi Stormin_Norman, the offline backtest is not kinda looking into the future per se. The EA only knows the actual signal if Time[0]==timestamp from file. So this part should be ok I think. Would be nice if you could record the signals...thanks I forgot: the EA currently in "use" is the one used for backtesting...if attached to a chart it only reacts on the live feed. I would change the default MaxOrders from 5 to 1 because it doesn't make sense to open 5 orders on 1 signal with all of them having the same SL/TP/TS i am sorry to say i am confused :) which EA of those listed (post number) can be used on a regular backtest? Quote "It is inconceivable that anyone will divulge a truly effective get-rich scheme for the price of a book." Victor Niederhoffer (1943–), US hedge fund manager and statistician Link to comment Share on other sites More sharing options...
iwjw Posted June 1, 2010 Report Share Posted June 1, 2010 I just uploaded the ea together with the indicators. It's a modified version of forexi's ea. Don't know whether it has bugs in it. What I saw in my tests is that there are trades on candles without a signal of the indicator. Didn't want to run it in visual mode. I've tested M15 and H1 for May but both were not profitable http://www.4shared.com/file/l1EUrO5_/LC_EA.html taipan 1 Quote Link to comment Share on other sites More sharing options...
jackdan Posted June 1, 2010 Report Share Posted June 1, 2010 I just uploaded the ea together with the indicators. It's a modified version of forexi's ea. Don't know whether it has bugs in it. is forexi ea without bugs? why did you modify this one? This thread is a little confusing, because fo what i read, forexi one is profitable, no? Quote Link to comment Share on other sites More sharing options...
iwjw Posted June 1, 2010 Report Share Posted June 1, 2010 I've just attached the LCSignal indicator to a H4 chart eurusd...at a first glance it could be profitable only to trade one H4 candle that opens with a signal and close the trade after 4 hours latest no matter what. There are not that many trade opportunities with H4 but maybe if traded with several pairs... Quote Link to comment Share on other sites More sharing options...
iwjw Posted June 1, 2010 Report Share Posted June 1, 2010 It's confusing because of the backtest thingy. It's nearly impossible to run BT over a longer period...so nobody knows whether the EA is profitable or not and which settings have to be used. Therefore I decided to make a special version for backtests that seemed to be huge profitable but in reality it isn't. The modification did I for my own testings. The signallogic now comes from an external indicator and not from the EA itself. I needed this to test my own indicator that I introduced at the beginning of the thread not knowing that an other trader already had coded an indi for LC Quote Link to comment Share on other sites More sharing options...
hitescape Posted June 1, 2010 Report Share Posted June 1, 2010 (edited) ...To illustrate my point, I changed the EA to initiate the order at the open of the next bar (which is generally within a pip of the close of the previous). Here are the two equity curves on Eur/Usd from mid January this year: http://i.imgur.com/OSupK.png... Why aren't the balance and equity lines the same? Is the EA opening more than one position and leaving it open? I understand the scaling out with partial closings but I didn't expect to see such a big difference in equity. Or am I just reading it incorrectly? Also, I looked at the forexi EA and would classify it as an "Alpha" version. I would not trade live with it without more testing and some modifications. Edited June 1, 2010 by hitescape Quote Link to comment Share on other sites More sharing options...
iwjw Posted June 1, 2010 Report Share Posted June 1, 2010 I just found the reason why a backtest takes so long We are only interested in a signal at the beginning of a new candle. So why call the indicator every tick? I'm just playing around with the modified version...so if you want to do yourself a favour don't BT with the version I uploaded earlier...it's really a pain Quote Link to comment Share on other sites More sharing options...
Stormin_Norman Posted June 1, 2010 Report Share Posted June 1, 2010 i am waiting to be told i can backtest a version. i can dedicate CPU time to it easily. i just need a reliable one to test. Quote "It is inconceivable that anyone will divulge a truly effective get-rich scheme for the price of a book." Victor Niederhoffer (1943–), US hedge fund manager and statistician Link to comment Share on other sites More sharing options...
⭐ atto Posted June 1, 2010 Report Share Posted June 1, 2010 hitescape, that was the EA that was posted here, I did not develop it. I was simply showing how the tester didn't work. I am moving this strategy to Matlab to do further testing and analysis. The current EA is horribly inefficient. Additionally, there are other elements to this strategy that are not in any indicator or EA here. Refer to the original system for those. hitescape 1 Quote Link to comment Share on other sites More sharing options...
iwjw Posted June 1, 2010 Report Share Posted June 1, 2010 I've just uploaded the modified EA version for faster backtests. Included is a .set file for eurusd H1 timeframe. Showing "good" results on M30 and M15 also. I've choosen H1 because I ran an optimization for 2010 on SL/TP/TS. Looks like results are better without trailing stop. According to the original LC rules the used indicator doesn't take into account that ema(7)/ema(21)/ema(84) should be used for M15 only...if memory serves me right their is a table with periods for higher timeframes than M15 in the manual. Sorry for the confusion with the offline backtesting but you can't imagine how I felt when first I saw these terrific results. The EA can now be backtested in regular mode without having to wait for hours to complete. Btw. I've deleted the old stuff except my selfmade Indicator for LC http://www.4shared.com/file/XG1hR3F2/LC_EA_V21.html halcyonn, hitescape, vince-ac and 6 others 9 Quote Link to comment Share on other sites More sharing options...
Stormin_Norman Posted June 1, 2010 Report Share Posted June 1, 2010 http://www.4shared.com/file/XFctJsZL/Lindencourt.htm original system which this EA is based on. ----> http://www.lindencourt.net/ remember- -- if you find this EA useful you should pay the 99 quid to the author. good work which profits us should be rewarded. we are about testing and finding good systems here, not taking good work for nothing. synergy, ⭐ capsmart, suryadi and 4 others 7 Quote "It is inconceivable that anyone will divulge a truly effective get-rich scheme for the price of a book." Victor Niederhoffer (1943–), US hedge fund manager and statistician Link to comment Share on other sites More sharing options...
iwjw Posted June 1, 2010 Report Share Posted June 1, 2010 best settings I've found for g/u are TP40/SL50 on H1 and M30. It's nice that e/u and g/u deliver profits between 30% and 50% within 5 month. But I do not quite like the shape of the equity curve. There are periods with up to 5 consecutive S/L hits. Maybe some1 can analyse the market condition during such a period. I'm still getting trades without having a signal arrow on that specific bar. don't know why Quote Link to comment Share on other sites More sharing options...
hitescape Posted June 1, 2010 Report Share Posted June 1, 2010 (edited) ... deleted ... Edited June 2, 2010 by hitescape Noy applicable to this thread. Quote Link to comment Share on other sites More sharing options...
Recommended Posts
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.