Simbasystem-gbpusd - page 46
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Malcik Test
Hi all,
I finally seem to have solved the technical problems. The reason the EA didn't work with other than factory csv is that I forgot to put an empty line at the end of the csv.
I have first 90% quality backtest for you - it's JMA 13 phase 0 on H4. Everything is working fine, I randomly checked some of its entries and they are correct. Still I'm not really sure about it though - in my backtest, it basically wiped out the 50 grant account in 5 months which doesn't look real given that it was profitable for Simba at 70%-80% quality.
If someone else gets 90% quality, please do this backtest too so that we can compare before I move on with backtesting. Attached is the full report with all the trades and with equity curve. This report contains my EA setting as well -- take a look at those too. Second is the csv file.Hi Malcik,
Your csv settings are ok,and your testing parameters are "almost" the same as mine,I have tp&sl at 1000 you have them at 999..also,I have been able to arrive to 87.5% modelling quality and my results are also very negative,I am starting to fed up with this data issue,both for SATL and JMA 13...I suggest that you test on 90% quality the 4h tf,while Itest on D1 tf,checking visually that the rules are functional and that the ohlc are right by random testing days(I have .csv daily data going back to 1991,that I buy and it is absolutely clean..),so,even if the daily modelling quality is 30%,if the open,high,low ,close are ok it shouldn`t affect the reliabilty of our results..
Regards
Simba
My Results 4hjma13 Phase 0
Strategy Tester Report
MultixpMAEAV1[1].0.0
Symbol GBPUSD (Great British Pound vs US Dollar)
Period 4 Hours (H4) 2004.07.09 00:00 - 2007.02.09 00:00 (2004.06.17 - 2007.02.10)
Model Every tick (based on all available least timeframes with fractal interpolation of every tick)
Parameters xpmafile="xpma.csv"; Order_Setting="---------- Order Setting"; OneEntryPerBar=false; ReverseTradesOnly=false; StopLoss=1000; TakeProfit=1000; BreakEvenAfterPips=40; Slippage=5; MaxOpenTrade=1; MinPriceDistance=5; TrailingStop_Setting="---------- Trail Management Settings"; TrailingStopType=1; TrailingStop=0; Moving_Average_Setting="---------- MA close Settings"; CloseMA=0; RSI_Settings="---------- RSI Filter Settings"; UseRSIfilter=false; RSIPeriod=21; RSIPrice=1; RSIBUYthreshold=50; RSISELLthreshold=50; CCI_Settings="---------- CCI Filter Settings"; UseCCIfilter=false; CCIPeriod=21; CCIPrice=1; CCIBUYthreshold=50; CCISELLthreshold=-50; Volume_Settings="---------- Volume Filter Settings"; UseVolumeFilter=false; VolumeRange=1; VolumeThreshold=7; Time_Parameters="---------- EA Active Time"; UseHourTrade=false; StartHour=0; EndHour=11; MM_Parameters="---------- Money Management"; Lots=5; MM=false; AccountIsMicro=false; DecreaseFactor=0; Risk=10; Alert_Setting="---------- Alert Setting"; EnableAlert=true; EmailAlert=false; SoundFilename="alert.wav"; Testing_Parameters="---------- Back Test Parameter"; PrintControl=true; Show_Settings=true;
Bars in test 4130 Ticks modelled 3965702 Modelling quality 87.82%
Initial deposit 50000.00
Total net profit -41804.59 Gross profit 176680.66 Gross loss -218485.25
Profit factor 0.81 Expected payoff -278.70
Absolute drawdown 41804.59 Maximal drawdown 41804.59 (83.61%) Relative drawdown 83.61% (41804.59)
Total trades 150 Short positions (won %) 75 (52.00%) Long positions (won %) 75 (25.33%)
Profit trades (% of total) 58 (38.67%) Loss trades (% of total) 92 (61.33%)
Largest profit trade 20420.02 loss trade -11199.98
Average profit trade 3046.22 loss trade -2374.84
Maximum consecutive wins (profit in money) 9 (26140.15) consecutive losses (loss in money) 14 (-19984.91)
Maximal consecutive profit (count of wins) 26140.15 (9) consecutive loss (count of losses) -24515.02 (7)
Average consecutive wins 2 consecutive losses 3
Simba
Hi Simba and all,
Here are results for H4 JMA phase 0 periods 34, 45, and 204. Looks like the account never makes it to Christmas 2004...
Simba, yes we should check it. I did a couple of random checks at JMA204 and I have to say I'm not sure how the EA handles the situation when the MA is flat which occurs very often...
If you have clean D1 data: I used to work part-time as a programmer -- maybe I could try to write a simple algorythm that would "trade" your csv files. That way we wouldn't have to deal with such Strategy Tester blackbox. It would be a tester for JMA specifically, I don't feel like creating my own MQL4 and my own Strategy Tester
I could program one for JMA, SATL, and HULL. It would take some time though.
C.E.O., you are familiar with the multiplexpMAEA -- what are your suggestions? How did you arrive at the results and insights you posted?
Thanks everyone
I am not sure what you guys are trying to achieve as I havent read the whole thread, but i will tell you my testing procedure and my opinion. Pick your pair and timeframe. Make sure your data is good, and are getting 90% which I believe you are. Well assume a JMA on a 4h chart. run at least a year to start. Set up the EA for a 5 period, no SL/TP. Run it and record the results in profit in pips and % of winning trades. If you want to do it right, you would have started with step period 1, and phase 0. Increase step period to 2 and run again, then 3, then 4, then 5, then,6. Now set Phase at 20 and repeat thru all phases, then 40 and repeat, then 60, and repeat, and so on. Now for one period which is 5, you would have to do 30 backtests. You then start at a period of 10, do the same, then 15, then 20.......... After a while you can weed out some of the variables as I did. That is how I came to the conclusion that a step period of 3-4, and phase of 100 is most profitable. Those numbers, no matter the period, those step period and phase consistently were the better. At that point you could assume that will always be the case and eliminate some of the backtests.
At this point you should have a spreadsheet with a type of MA, a timeframe, and a pair. You should have at least 20 periods on there with the best additional settings for each. You will see what I mean after this is done about there being better and worse periods as I said in my earlier post. At that point you could increase your backtesting to 2 or 3 years. For example in your spreadsheet you see that somewhere aroung a 45-55 period is most profitable. Start with 45, then 46, then 47 and so on until you find the most profitable in your range. If you wanted to continue on as I did, then you would start with stoploss in steps of 5 from 25-30 on up to whatever. then a breakeven, then a trail. If you REALLY want to find the best fitting MA, you could end up with hundreds of tests for one pair and timeframe. you could also use say a 100 pip SL to start the testing, but stick with the same one till the end.
I see you are on a 4hr timeframe. I feel like you are going to need a faster period MA for this. I did alot of work on 1hr and less. Example. On a 15m chart, how many pips will 4 bars really move? 100 or so in extreme conditions? Now how many pips could 4 bars move on the 4h? it could be hundreds easy. You are going to need pretty fast reactions on there for a single MA. Either that or some kind of minimal stoploss. Just for shits and gring, try a 150 SL and see how things go. I saw some massive losses on your backtests. Heres another tip. Round everything off to multiples of 10 for testing. your looking for maximum pips, not money. trying to figure out what kind of loss was suffered from 5 lots with 50000 is much harder that starting with 10000 and 1 lot. Or better yet 0.1 lot. every dollar is a pip. You can easily see at the end of a test you made 2834 pips instead of dividing bu 50, then subtracting 50000.
Ill post more tonight, but im at work sucking up their money right now.
Another comment if you are planning to do backtesting.
The results will vary depending on which currency pairs one uses.
I believe the worst results would be using GBP/USD. Some reason this pair is too volitile. Will be keeping away from this pair in any future trading situation.
The best to do testing with would be using EUR/USD, then USD/CHF, then AUD/USD, and USD/YEN.
Those are my 2 cents.
Ahh, but what if you came up with a profitable MA on the cable? test to survive in worst condition, profit in best. Really, each pair needs to be tested individually. they each have their own characteristics. You may come up with one that does pretty well on all. But even all this testing only gives a probability it will work, not a guarantee.
New Entry 5m
Hi All,
Just a few minutes ago we had a new sell entry in the 1.9468/75 area..please see pic
Simba
I went in this one at 1.9471.
We'll see how it goes.
Ceo
I am not sure what you guys are trying to achieve as I havent read the whole thread, but i will tell you my testing procedure and my opinion. Pick your pair and timeframe. Make sure your data is good, and are getting 90% which I believe you are. Well assume a JMA on a 4h chart. run at least a year to start. Set up the EA for a 5 period, no SL/TP. Run it and record the results in profit in pips and % of winning trades. If you want to do it right, you would have started with step period 1, and phase 0. Increase step period to 2 and run again, then 3, then 4, then 5, then,6. Now set Phase at 20 and repeat thru all phases, then 40 and repeat, then 60, and repeat, and so on. Now for one period which is 5, you would have to do 30 backtests. You then start at a period of 10, do the same, then 15, then 20.......... After a while you can weed out some of the variables as I did. That is how I came to the conclusion that a step period of 3-4, and phase of 100 is most profitable. Those numbers, no matter the period, those step period and phase consistently were the better. At that point you could assume that will always be the case and eliminate some of the backtests.
At this point you should have a spreadsheet with a type of MA, a timeframe, and a pair. You should have at least 20 periods on there with the best additional settings for each. You will see what I mean after this is done about there being better and worse periods as I said in my earlier post. At that point you could increase your backtesting to 2 or 3 years. For example in your spreadsheet you see that somewhere aroung a 45-55 period is most profitable. Start with 45, then 46, then 47 and so on until you find the most profitable in your range. If you wanted to continue on as I did, then you would start with stoploss in steps of 5 from 25-30 on up to whatever. then a breakeven, then a trail. If you REALLY want to find the best fitting MA, you could end up with hundreds of tests for one pair and timeframe. you could also use say a 100 pip SL to start the testing, but stick with the same one till the end.
I see you are on a 4hr timeframe. I feel like you are going to need a faster period MA for this. I did alot of work on 1hr and less. Example. On a 15m chart, how many pips will 4 bars really move? 100 or so in extreme conditions? Now how many pips could 4 bars move on the 4h? it could be hundreds easy. You are going to need pretty fast reactions on there for a single MA. Either that or some kind of minimal stoploss. Just for shits and gring, try a 150 SL and see how things go. I saw some massive losses on your backtests. Heres another tip. Round everything off to multiples of 10 for testing. your looking for maximum pips, not money. trying to figure out what kind of loss was suffered from 5 lots with 50000 is much harder that starting with 10000 and 1 lot. Or better yet 0.1 lot. every dollar is a pip. You can easily see at the end of a test you made 2834 pips instead of dividing bu 50, then subtracting 50000.
Ill post more tonight, but im at work sucking up their money right now.Hi C.E.O ,
Thanks for clarifying the matter.
Basically our system buy dips in an uptrend and sells rallies in a downtrend
What you propose is basically what we had in mind,with a conceptual difference regarding the aim of the ma/dig filter motivated by our philosophy
We will just set the TP and Sl at extremely high levels(keep with me they will come back later..
)
Our aim is not to find the ma/digital filter(and related tps ,sl,etc) that gives the best continuous entries and exits..For the rest the idea is basically extensive testing as you mention..
We just want to find the consistent one that keeps us in the trend with a positive edge-even with big drawdowns(we need the "theoretical drawdown" since this is where we will really enter,not at the change of slope in the MA,but at the most adverse possible point,while still in the trend,with a volatility based stop that covers our back just in case the retracement is really a change of trend),since we will take care of dds by a different additional filter,now we are using %R-but we want to test several settings,alternatives and combinations...
We will start by testing the lengths of the ma that fit with the cyclic model of the currency pair(usually 35,45,50..or in between..that is why Hull ma 45 or 50 works so well on most pairs)..and then test systematically the variations in length,step,phase(case of JMA),etc
So,regarding ma/digfilter we are looking for a trend filter..and we need it either in 4h or D1 tf..since 1H doesn`t consistently gives enough room for the retracements we are looking for..and we will test 1h too,just in case it is possible to find a consistent trend filter in that tf,which will be a blessing
The reasons why I contacted you were,first to thank you for this excellent testing MAs EA and second to try to evaluate the possibility of incorporating and testing additional indicators,not for trend filtering,but for "retracement catching ",and then we will incorporate all the necessary SL and TP testing of the system..within a previously defined trend
Thanks again for your kind comments and suggestions
Simba
ssotiro
I went in this one at 1.9471. We'll see how it goes.
Hi ssotiro ,
well you sold better than me... I am in @ 1.9468..;)..well done
Simba