SampleMACD!?!? Would you get a load of this!?

[Batchboy]

Originally Posted by rsmereka

JB,

I forward tested this EA (modified so it would only consider it's own trades) for about 3-4 weeks on EURUSD M1 and another instance on GBPUSD H4.

The EURUSD instance, at first, made good profit but it ended up giving back way more. The GBPUSD instance traded way less frequently and was much more stable (in terms of keeping profit) but I dumped it because it would only do one or less trade a week.

Rick

The

Rick,
You doing this on periodicity = 1min EURUSD ??? The parameters I arrived at are in top post and show on that back-test url. Is it IBFX platform? If still troubles I don't know what to say, it as you see on both live$ and back-test urls is "Verrrr-king".
jb

[Batchboy]

oh, forgot, I use a live$ account platform, so back-data on a demo I haven't done optimization or tests on, only live$ data account platform, could most likely make a difference. But it only takes $10 sent to IBFX to open a mini size account, thus if you don't already have open a live$ platform you could start one for back-testing.
jb

[rsmereka]

Originally Posted by Batchboy

Rick,
You doing this on periodicity = 1min EURUSD ??? The parameters I arrived at are in top post and show on that back-test url. Is it IBFX platform? If still troubles I don't know what to say, it as you see on both live$ and back-test urls is "Verrrr-king".
jb

Jb,

Ok. I'll try another test based on your back-test settings.

Rick

[Batchboy]

Originally Posted by rsmereka

Jb,

Ok. I'll try another test based on your back-test settings.

Rick

If you don't have live$ platform data (just demo) then you most likely need to do optimization around thOSE ballpark figues to arrive at what's needed for demo running.
I just had 1ka do his SoftWyer-Lib version and even though Sample-MACD had one 11 pip loser the new version had +6 pip gainer which is 46th in a row winner for it. There is something new & different TA wise on periodicity=1min that 1 bar apart MACD impulse sensing tells of the eventual future move no matter a third to a half the trades have momentary drawdowns of 50 to 250 pips ..............it'll sooner or later will end-up in positive gainer territory, never seen anything like it. Ordinarily seeing something like this in the past we'd all say "hogwash, some kind of over curve fitting", but it walks forward ditto!
jb

[rsmereka]

Originally Posted by Batchboy

If you don't have live$ platform data (just demo) then you most likely need to do optimization around thOSE ballpark figues to arrive at what's needed for demo running.
I just had 1ka do his SoftWyer-Lib version and even though Sample-MACD had one 11 pip loser the new version had +6 pip gainer which is 46th in a row winner for it. There is something new & different TA wise on periodicity=1min that 1 bar apart MACD impulse sensing tells of the eventual future move no matter a third to a half the trades have momentary drawdowns of 50 to 250 pips ..............it'll sooner or later will end-up in positive gainer territory, never seen anything like it. Ordinarily seeing something like this in the past we'd all say "hogwash, some kind of over curve fitting", but it walks forward ditto!
jb

Hi Jb,

I ran the strat tester which came up with about $2,300 profit between Nov 22, 2008 and April 11, 2009. It took almost 24 hours to run I guess because I was stepping most of the external parameters. After I got the result, I realized that the dates were not accurate and I am now in the process of re-running. Meanwhile, I changed the EA parameters to the ones which generated the highest profit. When this run finishes (MT4 currently says that there are almost 192 hours left but I am certain that these estimated numbers are wrong).

Be careful of over-optimization. The more you optimize in a tight date range (1-3 months), the more the solution will be curve-fitted to the current market. I usually prefer optimizating using at least six months of data. The problem with this is that given the number of ticks, an optimization run can take a l.o.n.g time.

I am optimizing my modified version of the original MACD EA. The original version looked at all trades in the account and had no magic number. I gather that your TripleHeaderMACD is also a modified version.

Rick

[Batchboy]

hi Rick,
Is that profit off of initial equity of a Thou? For about 5 months? What kind of settings did your optimization end-up-with for that back-test result? On 1min periodicity, right? How far off from my settings are yours for that period? Why not just optimize from thart November to current date? If it will take a lot of time narrow the ranges around what values you came up with so far to reduce optimization time. Also inform of your dd, mine are worst case 20%, often the usual ones 1/3rd of the time are 5-10%.

My new own independent version its up to 47 winners now, nooooo losers!:
http://206.71.180.236/index/Strategy...MACD6-1-09.htm

It "seems" to be behaving as it's own fundamental TA style, a certain large impact of MACD determined by difference of what is was 1 bar ago and what it is current bar, on 1min that have to be kinda large eh? Then the other Qual is an MA crossing, this combo "seems" to "predict" if this is a crowd of beting folks that then "follow the crowd" and 1/3rd the time go for it in same direction then unload it a little later, thus 1/3rd the time there is fair to medium drawdowns, the other 2/3rds of the time its a fast dip with a retracement pronto, so those are no-dd movements. In either case it seems to "predict" this behavior.

Anxious to hear your longer period results Rick.

jb

[rsmereka]

Originally Posted by Batchboy

hi Rick,
Is that profit off of initial equity of a Thou? For about 5 months? What kind of settings did your optimization end-up-with for that back-test result? On 1min periodicity, right? How far off from my settings are yours for that period? Why not just optimize from thart November to current date? If it will take a lot of time narrow the ranges around what values you came up with so far to reduce optimization time. Also inform of your dd, mine are worst case 20%, often the usual ones 1/3rd of the time are 5-10%.

My new own independent version its up to 47 winners now, nooooo losers!:
http://206.71.180.236/index/Strategy...MACD6-1-09.htm

It "seems" to be behaving as it's own fundamental TA style, a certain large impact of MACD determined by difference of what is was 1 bar ago and what it is current bar, on 1min that have to be kinda large eh? Then the other Qual is an MA crossing, this combo "seems" to "predict" if this is a crowd of beting folks that then "follow the crowd" and 1/3rd the time go for it in same direction then unload it a little later, thus 1/3rd the time there is fair to medium drawdowns, the other 2/3rds of the time its a fast dip with a retracement pronto, so those are no-dd movements. In either case it seems to "predict" this behavior.

Anxious to hear your longer period results Rick.

jb

Hi Jb,

Results were for an initial deposit of $5,000.00 About 50% return in 5 months. I am re-running with the same initial deposit one mini (0.1) fixed lots (my modified version has no MM). Using dates Dec 1, 2008, Jun 1, 2009. Still running. Been running for 7.5 hours so far. MT4 says 141 hours to go I think it lies...at least...I hope that it is lying like those wonderful Windows progress bars and estimated time to completion.

Rick

[Batchboy]

Originally Posted by rsmereka

Hi Jb,

Results were for an initial deposit of $5,000.00 About 50% return in 5 months. I am re-running with the same initial deposit one mini (0.1) fixed lots (my modified version has no MM). Using dates Dec 1, 2008, Jun 1, 2009. Still running. Been running for 7.5 hours so far. MT4 says 141 hours to go I think it lies...at least...I hope that it is lying like those wonderful Windows progress bars and estimated time to completion.

Rick

yes, reason it does like that is under testing tab you'll see "genetic algorithm", which means much of the time it randomly jumps around on parameter checks and then zooms in on something in good ballpark, thus suddenly shortening the time if that method were not checked. Unless I have very few parameters to go thru I usually have that checked, otherwise, yes,
LOL
jb

[Batchboy]

Rick,
You gave me the inspiration to do another tighter range optimization on the new TripleHeaderMACD. I wanted to see if moving things around the originally arrived settings would fluctuate performance much, I conjecture that if in these 4-mo+1wk back-1min-data there were to be a result that is very few then such results would be for a very tight range that shows-up any sort of decent performance then this would indeed be overly-curve-fit, if however there were many choices to pick of profit and not much losers to none and an acceptable range of drawdown then this would for now prove-out not overly-curve-fit.

And the optimization was run! The result is the latter, fortunately, I had this "feeling" by way it continued to perform live$ from day-1 in earliest posts and how it didn't need further optimizations even though first optimizations were crude results! I needed to do it this way as it without luck that I have no more back-data on 1min EURUSD than I have, very much depends on luck as to how much it seems IBFX will avail.

The journal said there were 440 significant passes during optimization out of the approx 1200 available sets, very excellent sign. When I clicked above profit column and counted down from top profit row there were 162 above around $800 profit and below 30% drawdown , just as it got to number 163 there was a 47% dd.
The start, step & stop values for each parameter was:
TP 65, 5 80
TS 10, 5, 30
macd-fast 11, 1, 13
macd-slow 25, 1 27
macd-signal 8, 1, 10
ma-trend 11, 1, 13
macd-open 8, 1 12
macd-close 5, 1, 7
SL was fixed at 300

The top profit result w/o exceeding 20% dd was $1468 with 16.7% dd. Settings are:
TP =65, TS=25, macd-fast=11, macd-slow=25, macd-signal=9, ma-trend =12,
macd-open=10, macd-close=7. This is what I now set mine and Bert's puny accounts to to continue the walk forward. The above back-test performance url shows this result as I pasted the new results over previous ones in the server.
jb

[rsmereka]

Originally Posted by Batchboy

Rick,
You gave me the inspiration to do another tighter range optimization on the new TripleHeaderMACD. I wanted to see if moving things around the originally arrived settings would fluctuate performance much, I conjecture that if in these 4-mo+1wk back-1min-data there were to be a result that is very few then such results would be for a very tight range that shows-up any sort of decent performance then this would indeed be overly-curve-fit, if however there were many choices to pick of profit and not much losers to none and an acceptable range of drawdown then this would for now prove-out not overly-curve-fit.

And the optimization was run! The result is the latter, fortunately, I had this "feeling" by way it continued to perform live$ from day-1 in earliest posts and how it didn't need further optimizations even though first optimizations were crude results! I needed to do it this way as it without luck that I have no more back-data on 1min EURUSD than I have, very much depends on luck as to how much it seems IBFX will avail.

The journal said there were 440 significant passes during optimization out of the approx 1200 available sets, very excellent sign. When I clicked above profit column and counted down from top profit row there were 162 above around $800 profit and below 30% drawdown , just as it got to number 163 there was a 47% dd.
The start, step & stop values for each parameter was:
TP 65, 5 80
TS 10, 5, 30
macd-fast 11, 1, 13
macd-slow 25, 1 27
macd-signal 8, 1, 10
ma-trend 11, 1, 13
macd-open 8, 1 12
macd-close 5, 1, 7
SL was fixed at 300

The top profit result w/o exceeding 20% dd was $1468 with 16.7% dd. Settings are:
TP =65, TS=25, macd-fast=11, macd-slow=25, macd-signal=9, ma-trend =12,
macd-open=10, macd-close=7. This is what I now set mine and Bert's puny accounts to to continue the walk forward. The above back-test performance url shows this result as I pasted the new results over previous ones in the server.
jb

Jb,

I did not realize that your SL was fixed at 300. I don't know what mine is set to. I just realized that there is no SL in the external parameters. I just check the source code and there is no stop loss other than a trailing stop. Freeky

My strat test is still running. MT4 now says 60 hours and 44 minutes to go. I have my EA settings set to last run and I just had a big loss 107.5 pips. Account is down $104.10. 23 trades so far 12 wins, 11 losses. I can't tell you what my optimization settings are right now unless I stop the strat test (which I don't want to do). Wait one...I have a set file that I created for the strat test. Here it is:
TakeProfit=700.00000000
TakeProfit,F=1
TakeProfit,1=50.00000000
TakeProfit,2=10.00000000
TakeProfit,3=1000.00000000
Lots=0.10000000
Lots,F=0
Lots,1=0.10000000
Lots,2=0.00000000
Lots,3=0.00000000
TrailingStop=200.00000000
TrailingStop,F=1
TrailingStop,1=30.00000000
TrailingStop,2=10.00000000
TrailingStop,3=700.00000000
MACDOpenLevel=10.00000000
MACDOpenLevel,F=1
MACDOpenLevel,1=2.00000000
MACDOpenLevel,2=1.00000000
MACDOpenLevel,3=20.00000000
MACDCloseLevel=6.00000000
MACDCloseLevel,F=1
MACDCloseLevel,1=2.00000000
MACDCloseLevel,2=1.00000000
MACDCloseLevel,3=20.00000000
MATrendPeriod=12.00000000
MATrendPeriod,F=1
MATrendPeriod,1=2.00000000
MATrendPeriod,2=1.00000000
MATrendPeriod,3=24.00000000

I have my TP stepped from 5 to 100 in 1 pip steps. Lots is fixed. TS is stepped from 3 to 70 in one pip steps. OpenLevel is stepped from 2 to 20 in 1.0 steps. CloseLevel is stepped from 2 to 20 in 1.0 steps. TrendPeriod is stepped from 2 to 24 in 1.0 steps.

Keep in mind that my Alpari UK demo account is 5 decimals.

Tell me why you decided to fix the SL at 300 pips?

Rick