EMA 25-50 Trading System (Detailed) | Page 104 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

10 Ema and 200 Ema cross system 6 replies

EMA crossover (ema 14 & 34) 4 replies

Please help code an EMA Crossover 12ema x 24 ema cross 1 reply

EMA's and MA's confiming small time correctoin ema's crossing 1 reply

EMA Close Crossing EMA High/Low Help Please! 2 replies

Attachments: EMA 25-50 Trading System (Detailed)
Exit Attachments

EMA 25-50 Trading System (Detailed)

  • #2,061
  • Aug 6, 2008 3:31am Aug 6, 2008 3:31am
aajduk,
I thought in PMs we could send attachements... since we can't, here it is.

Some considerations:
If you want to look only at open above, and close below EMA for short (and reverse for long), it has to be on the previous candle, because of the "close" - orelse, the Indicator would repaint.
The entry will be given at the open of the next candle.
Since you maintain the other signals (RSI & MACD), it gives very few entries.

About the pip counter:
I'm not sure of the extend - and I remember that I tested it with ATR=true... don't know if with fixed values it counts right. But I wanted to evaluate the ATR, so I stopped there.
But there's an easy way to evaluate it, please do the following:
Locate in your chart the hour of the last good signal;
Do that counting yourself;
Then set the start time of the counter to the previous hour of that entry signal (use the hour from your MT), and compare the pip result.
Attached File(s)
File Type: mq4 EMA_25_Indicator_aajduk.mq4   29 KB | 344 downloads
  • #2,062
  • Aug 6, 2008 4:19am Aug 6, 2008 4:19am
aajduk,
I forgot to add your change of the RSI values.
Don't forget to do it yourself, to have it as you wanted it.
  • #2,063
  • Aug 6, 2008 4:47am Aug 6, 2008 4:47am
please i will like to know the difference between the EA with file extention .ex4 and .mq4

And is beta3 the latest of the EA's?
  • #2,064
  • Aug 6, 2008 6:34am Aug 6, 2008 6:34am
anyway we can still resolve the Stochastic Problem with this:

if an hour closes like this:
MACD (12,24) > 0
RSI >50
Stochastic H4 -> long position

then, second check, buy if:
New candle opens above EMA 24

but Stochastic check has to be at the "previous" hour and once the hour has ended the stochastic has to be "locked" at the level it was when the hour closed.

can you code this?
  • #2,065
  • Aug 6, 2008 1:53pm Aug 6, 2008 1:53pm
Corre,
The Stochastic is coded as you say since the beginning

Between the 2 stochastic I came across with, there was a big difference:

When set to 4H in the 1H timeframe, for example if we are in the last fourth hour, one varies the last 4 hours value (candles), and the other locks that value hour by hour - that's the one I used, looking at it's past hour value (i+1).

So I don't know how is this possible - I have to see it happen to understand it. For that I moved to 1m timeframe, with 5m stochastic.

Just to be sure, you say that when stochastic is crossing, if it keeps crossing up and down, you get the arrows repainted? An arrow keeps appearing and disappearing?
  • #2,066
  • Aug 6, 2008 2:41pm Aug 6, 2008 2:41pm
at 20.00 gmt on EURUSD, the 5th august we had a short-loss. then H4 moved and the whole indicator recognized a short at 18.00 that before we hadn't, this is because at 18.00 stoch was LONG but after some hours it returned short and the segnal line ended the period coming exactly over the base line. (yes, very strange period).
so the 20.00 loss became a 18.00 gain tp1. VERY VERY VERY WRONG.

anyway the system moves the stochastic back to four hours it is not true that it remains fixed, you can see it everytime the 2 lines are very near. the lines keep crossing and they keep moving for the last (at maximum obviously) 4 hours.
this is the problem of the "lines", if a line connect 2 points and these 2 points keep moving the lines will move.

it don't want to make it too complicated but it almost impossible to keep the data simply locked as they were at the end of the hour unless coding deeply the MTFStoch indicator alone.
  • #2,067
  • Aug 6, 2008 4:10pm Aug 6, 2008 4:10pm
One last question:
Between these changes you have seen, did you switch between timeframes in your chart (I mean did you see the first sign, then changed timeframe in the chart, and then return to the 1H timeframe and the arrow had changed place)?

I ask this because this is a limitation of using higher timeframe data in a lower timeframe chart (unless in our case you only look at i+4). Follow me on this:

When you load the Indicator in a chart, it does all the past calculations... but there is only data from stochastic every 4 hours, so there's 4 1H candles based on the same stochastic value;
But if you leave the chart open and running, every 1 hour (in hour case) it locks the 4H stochastic value at that moment, and so gives signals based on that - of course, if you change the timeframe (or close MT and re-open it), it recalculates all past data, using 1 value of stochastic for each 4 1H candles.

What this does is having faster signals when the chart is running live, so we may have some that won't be there when you reload the chart (and vice-versa).

But there's one abnormal thing I can noticed (I'm testing and writing at the same time), and that may be what you saw (best noticed in 1M timeframe, in moments with low volatility):
When in 1m we only get 1 tick, our i value of stochastic is at a certain place;
when a new minute (candle) comes, the previous stochastic value (now called i+1) changes to the openning value of the present i candle and then stays there (will not change anymore).
It doesn't happen everytime (that's why I haven't seen it before)... I'm not sure but it may be only when it concerns crossing, or changes in direction of lines.

The only way to avoid it would be coding stochastic at i+2, instead of i+1... it may be a little later and change a bit performance (still faster than waiting for 4H to close), but will avoid those repainting in live mode.
But it won't avoid some possible differences when reloading the chart (only possible to avoid by coding it to i+4);
But those I consider of minor significance and not important to the overall general performance.

What do you think?
  • #2,068
  • Aug 6, 2008 6:12pm Aug 6, 2008 6:12pm
I never mentionned it before (because I thought it was obvious), but there's another limitation to the coding of the indicator - I don't know if you noticed it:

If you have a entry signal, and the present candle moves in one direction but at some point reverses completely to the other, it's possible for example to first see a black cross (SL signal), but then this signal disappears and later turns into a TP1 signal (and then stays that way).

And when you load a chart, in such a candle it's impossible to know which came first. The candle may reach in both sides of it SL and TP1 values, but the calculation will stop at the condition reached first in it's code (when calculating past data), indicating that signal - but if it would had been live, maybe the other signal had came later and then repaint.

Those are:
coding limitations;
working with 4 values (open, close, high, low) limitations.
  • #2,069
  • Aug 6, 2008 6:19pm Aug 6, 2008 6:19pm
Quoting rover666
i never change anything from this beta EA, i download and install and it running well..
If it's a demo account, can you tell us which server (service provider) are you using?

Because in the defaults parameters, it's set to allow fractional lots (mini/micro lots), but my demo account only seems to work with normal lots, so the defaults don't work. Yours does.
And I would like to try it with smaller lots (more realistic to my money investment possibilities).

Thanks.
  • #2,070
  • Edited 6:53pm Aug 6, 2008 6:35pm | Edited 6:53pm
Quoting tonirai
One last question:
Between these changes you have seen, did you switch between timeframes in your chart (I mean did you see the first sign, then changed timeframe in the chart, and then return to the 1H timeframe and the arrow had changed place)?

I ask this because this is a limitation of using higher timeframe data in a lower timeframe chart (unless in our case you only look at i+4). Follow me on this:

When you load the Indicator in a chart, it does all the past calculations... but there is only data from stochastic every 4 hours, so there's 4 1H candles based on the same stochastic value;
But if you leave the chart open and running, every 1 hour (in hour case) it locks the 4H stochastic value at that moment, and so gives signals based on that - of course, if you change the timeframe (or close MT and re-open it), it recalculates all past data, using 1 value of stochastic for each 4 1H candles.
may be but ehi... this means there can be 500 wrong pip-counting... we have to think something that CAN'T repaint

Quoting tonirai
The only way to avoid it would be coding stochastic at i+2, instead of i+1... it may be a little later and change a bit performance (still faster than waiting for 4H to close), but will avoid those repainting in live mode.
But it won't avoid some possible differences when reloading the chart (only possible to avoid by coding it to i+4);
But those I consider of minor significance and not important to the overall general performance.

What do you think?
I think that we have to avoid the charts that can be not exact with their "history"
ALL of this i+1 project is due to the problem of miscounting with the "base" indicator, do you remember? we can't bring the same tumor to this new indicator.

But I don't understand what you meant with that i+2 thing

edit: aaa you mean that we go back to the last ALREADY CLOSED stochastic h4? this can be cool.

let's use the closest stochastich4 that has already completly closed, for the first 2 hours it is almost the same the problem is with the third and the fourth hour perhaps...

mmm I don't know let's test this if it's possible. perhaps we can add another indicator faster for the last 2 hours.

Anyway let's test this with "the last Stoch H4" already closed, can we do it?
  • #2,071
  • Aug 6, 2008 6:52pm Aug 6, 2008 6:52pm
Quoting Corre
ALL of this i+1 project is due to the problem of miscounting with the "base" indicator, do you remember? we can't bring the same tumor to this new indicator.
Yes, but like I said above, it will never be 100% accurate;
At the very least, we'll always have the problem/possibility of one candle reaching both SL & TP1 values, but indicating only one of them, and we can't know which came first... the software can't know how that candle evolved.

Quoting Corre
But I don't understand what you meant with that i+2 thing
I mean that what I saw was a repainting in the stochastic in some cases (not always), possibly related to crossing moment, or changing direction's moment, but only in i+1 candle. So coding it to i+2 should solve that, even though changing performance (because it will be slower to change signal).

One thing I've been wanting to ask you, about the change in the MACD_main>0:
Did you count the pips for that particular only change?
Versus the MACD_main>MACD_signal, or versus the "no MACD" version?
Results?
  • #2,072
  • Edited 7:25pm Aug 6, 2008 7:04pm | Edited 7:25pm
Quoting tonirai
Yes, but like I said above, it will never be 100% accurate;
At the very least, we'll always have the problem/possibility of one candle reaching both SL & TP1 values, but indicating only one of them, and we can't know which came first... the software can't know how that candle evolved.
yes but this happens quite rarely.. to have both SL and TP1 in the same candle.
Anyway to be more prudent let's count them as loss (is it technicly possible to code it?)

An idea for the coding may be this:
If we're long first consider the Minimun, then consider the Max, then consider the Close
If we're short, first consider the Max then the Min at last the close.

So if SL is touched is surely would also be in the worst part (the max for the short, the min for the long)

Else we can make a great job (but this sounds impossible) to consider the M1 timeframe in the count. M1 knows where the prices have gone during the hour so if we look at it, or we have the system looking at it we can have a correct pip count and correct segnals

but anyway the min/max in first consideration thing is already enough and would be far easier to be coded.

Quoting tonirai
I mean that what I saw was a repainting in the stochastic in some cases (not always), possibly related to crossing moment, or changing direction's moment, but only in i+1 candle. So coding it to i+2 should solve that, even though changing performance (because it will be slower to change signal).
yes but what do you mean with i+2? we return behind of another hour? ok, perfect but this would be cool just for the second hour. the third hour has to be i+3 and the fourth i+4, hasn't it? or there's some way to have locked down the "2 hours ago Stoch H4 even if still running" ?

edit: yes, i checked it right now. I+2 would be enough, Stoch levels are fixed every hour.
edit2: no, they are not fixed. if we change tf/close MT4 they repaint again... the only values that remains are the open and the close of stochastic, so we have to use the closest that has ALREADY closed.
for the first hour is just +1, for the second it becomes +2 as you said, but for the 3rd and the 4th it must become +3,+4. indeed.

Quoting tonirai
One thing I've been wanting to ask you, about the change in the MACD_main>0:
Did you count the pips for that particular only change?
Versus the MACD_main>MACD_signal, or versus the "no MACD" version?
Results?
the best until now is macd_main >0 but i just counted and compared with Eurusd. deeper counting are in line to be done.
  • #2,073
  • Aug 6, 2008 7:05pm Aug 6, 2008 7:05pm
Recently found this thread and have read it through and hopefully have caught the jist of how it works (although it gets easily confusing I have to admit).
I do have the beta3 ea running but I set the 'use macd' setting to false, and am only using the stoch and rsi as I believe was the last consensus here.
Hopefully that is correct.
it opened it's first string of trades today on the eurusd at 20:00 FXDD time.
Of course, the pair is stuck in glue so not much has happened, but hopefully that will change after tomorrows news.
Just thought I'd share that I am testing the EA and am more than happy to share results or charts or however I can help.
  • #2,074
  • Aug 6, 2008 7:15pm Aug 6, 2008 7:15pm
Quoting tazmaan
Recently found this thread and have read it through and hopefully have caught the jist of how it works (although it gets easily confusing I have to admit).
I do have the beta3 ea running but I set the 'use macd' setting to false, and am only using the stoch and rsi as I believe was the last consensus here.
Hopefully that is correct.
it opened it's first string of trades today on the eurusd at 20:00 FXDD time.
Of course, the pair is stuck in glue so not much has happened, but hopefully that will change after tomorrows news.
Just thought I'd share that I am testing the EA and am more than happy to share results or charts or however I can help.
ea has some bug in line to be fixed, I don't recommand to use it with a live account
  • #2,075
  • Aug 6, 2008 7:28pm Aug 6, 2008 7:28pm
Quoting Corre
yes but this happens quite rarely.. to have both SL and TP1 in the same candle.
Anyway to be more prudent let's count them as loss (is it technicly possible to code it?)
Just by looking to the code, I would say yes, it's just a question of changing the order, putting the SL condition's check before the TP conditions.

Quoting Corre
Else we can make a great job (but this sounds impossible) to consider the M1 timeframe in the count. M1 knows where the prices have gone during the hour so if we look at it, or we have the system looking at them we can have a correct pip count.
Don't go there... it brings it to an "impossible to code" level
And for those things, we need buffers, and we are already using 7!

Quoting Corre
yes but what do you mean with i+2? we return behind of another hour? ok, perfect but this would be cool just for the second hour. the third hour has to be i+3 and the fourth i+4, hasn't it? or there's some way to have locked down the "2 hours ago Stoch H4 even if still running" ?

edit: yes, i checked it right now. I+2 would be enough, Stoch levels are fixed every hour.
Good you figure it out... I didn't know how to put it in another words


I think what's important is that we have to limit the most (software limited) the miscounts/repaintings, without hurting too much the Indicator's performance;
Then it's got to be forward tested (by manual trading, or EA), to give the final green light to it.
  • #2,076
  • Aug 6, 2008 7:35pm Aug 6, 2008 7:35pm
Quoting tonirai
I think what's important is that we have to limit the most (software limited) the miscounts/repaintings, without hurting too much the Indicator's performance;
Then it's got to be forward tested (by manual trading, or EA), to give the final green light to it.
check the edit2 if we close and reopen the platform the only data remaining are open and close. we can't use single hour (+2) closes we have to use the h4 closes... and we have to use the nearest

the only problem with this would be in the last 2 hours... for the first and the second doesn't change a lot, but for the fourth there's some delay
  • #2,077
  • Aug 6, 2008 7:41pm Aug 6, 2008 7:41pm
Quoting tazmaan
Recently found this thread and have read it through and hopefully have caught the jist of how it works (although it gets easily confusing I have to admit).
I do have the beta3 ea running but I set the 'use macd' setting to false, and am only using the stoch and rsi as I believe was the last consensus here.
Hopefully that is correct.
it opened it's first string of trades today on the eurusd at 20:00 FXDD time.
Of course, the pair is stuck in glue so not much has happened, but hopefully that will change after tomorrows news.
Just thought I'd share that I am testing the EA and am more than happy to share results or charts or however I can help.
tazmaan,

First of all, thanks for sharing.
Like Corre said, I hope you're doing it in demo for now.
"
But I don't know if the EA has bugs (as he says), or if it's just problems concerning the parameters, depending if you use a service provider that allows micro lots.

But it's important that everyone who's testing it, comes to share the results (good or bad).
One thing I would do is to trade several pairs, so that a serie of possible bad results can be offseted by good ones in other pair.

The reason why it gets confusing (I think) is that we come here to discuss "minor" issues, configurating issues, improvements, etc.
We do it here instead of by PM or e-mail, to keep the thread alive and open for discussion.
If we would do it the other way, coming here just to post final results, I think it wouldn't be so interesting... though less confusing
  • #2,078
  • Aug 6, 2008 7:41pm Aug 6, 2008 7:41pm
to get the point of the situation:

we need to code:

1) the system that first consider the max for the short and the min for the long.
or at least a system that if there are 2 segnals in the same candle and 1 of them is a SL (even with profit) it closes the position.

2) fix the stoch thing. I+1,+2,+3,+4 is probably the only way. and honestly just returning back manually it doesn't seem to change the performance at all...

after these 2 modifications the system CAN'T have rapainting or miscalculations, and then we can start playing with macd, Ema, RSI, etc. etc. etc. and overall we can have a correct count of the performance.
  • #2,079
  • Aug 6, 2008 8:25pm Aug 6, 2008 8:25pm
Quoting Corre
to get the point of the situation:

we need to code:

1) the system that first consider the max for the short and the min for the long.
or at least a system that if there are 2 segnals in the same candle and 1 of them is a SL (even with profit) it closes the position.

2) fix the stoch thing. I+1,+2,+3,+4 is probably the only way. and honestly just returning back manually it doesn't seem to change the performance at all...

after these 2 modifications the system CAN'T have rapainting or miscalculations, and then we can start playing with macd, Ema, RSI, etc. etc. etc. and overall we can have a correct count of the performance.
1) The bold option seems to easiest to change, without having to change the all indicator.

2)Only way would be to use i+4 but I'm scheptical about that... It won't have repainting issues, but I think that in live mode the previous gives better signals.
  • #2,080
  • Aug 6, 2008 8:42pm Aug 6, 2008 8:42pm
Here is the change in the SL and SL_after_profit management.

If you load it on your current EURUSD, you'll see it in action at 01/08:
a short trade reaching TP2 becomes a losing one, because that candle reaches both levels (this is seen in my config of RSI at 60/40, and no MACD).
Attached File(s)
File Type: mq4 EMA_25_Indicator_Configurable.mq4   40 KB | 316 downloads
10 traders viewing now