My EA Gift to Traders | Page 31 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Screen shot indicator free gift 0 replies

DECEMA - Xmas gift! 6 replies

Vegas 1HR GIFT: Big opportunity coming! 5 replies

Attachments: My EA Gift to Traders
Exit Attachments

My EA Gift to Traders

  • #601
  • Edited 4:09pm Feb 18, 2015 3:58pm | Edited 4:09pm
Quoting drofwarc
Here's the problem I spotted. I'm using simple candle reversal trade logic to make legitimate trades easy to spot. Short trade: if ((C1>O1)&&(C<O)) Long trade: if ((C1<O1)&&(C>O)) I hope this is clear for you. {image} {file}
I would need to see the full code block including how you handled the variable ShiftV which is integral to the code.
below is an example. You would also need to disable all the other patterns in the CalcCandleSticks subroutine.

bool candlePatternBearishEngulfing(int ShiftV)
{
double O = iOpen(Symbol(),ReqTimeFrame,ShiftV);
double O1 = iOpen(Symbol(),ReqTimeFrame,ShiftV+1);
double C = iClose(Symbol(),ReqTimeFrame,ShiftV);
double C1 = iClose(Symbol(),ReqTimeFrame,ShiftV+1);
if ((C1>O1)&&(O>C)&&(O>=C1)&&(O1>=C)&&((O-C)>(C1-O1))) return(true);
return(false);
}

One other point I noticed, is that the info panel shows UseTimeOfDay as enabled but the set file shows it as disabled. If it is to be enabled, you must now have the endhour as 23
  • #602
  • Feb 18, 2015 4:05pm Feb 18, 2015 4:05pm
Quoting Dave_C
{quote} I would need to see the full code block including how you handled the variable ShiftV which is integral to the code. below is an example bool candlePatternBearishEngulfing(int ShiftV) { double O = iOpen(Symbol(),ReqTimeFrame,ShiftV); double O1 = iOpen(Symbol(),ReqTimeFrame,ShiftV+1); double C = iClose(Symbol(),ReqTimeFrame,ShiftV); double C1 = iClose(Symbol(),ReqTimeFrame,ShiftV+1); if ((C1>O1)&&(O>C)&&(O>=C1)&&(O1>=C)&&((O-C)>(C1-O1))) return(true); return(false); }
Here you go. All I've done is modify your trade logic slightly to make trades easy to spot.

bool candlePatternBearishEngulfing(int ShiftV)
{
double O = iOpen(Symbol(),ReqTimeFrame,ShiftV);
double O1 = iOpen(Symbol(),ReqTimeFrame,ShiftV+1);
double C = iClose(Symbol(),ReqTimeFrame,ShiftV);
double C1 = iClose(Symbol(),ReqTimeFrame,ShiftV+1);
if ((C1>O1)&&(C<O))
//if ((C1>O1)&&(O>C)&&(O>=C1)&&(O1>=C)&&((O-C)>(C1-O1)))

return(true);
return(false);
}
//+------------------------------------------------------------------+
bool candlePatternBullishEngulfing(int ShiftV)
{
double O = iOpen(Symbol(),ReqTimeFrame,ShiftV);
double O1 = iOpen(Symbol(),ReqTimeFrame,ShiftV+1);
double C = iClose(Symbol(),ReqTimeFrame,ShiftV);
double C1 = iClose(Symbol(),ReqTimeFrame,ShiftV+1);
if ((C1<O1)&&(C>O))
//if ((O1>C1)&&(C>O)&&(C>=O1)&&(C1>=O)&&((C-O)>(O1-C1)))
return(true);
return(false);
}
  • #603
  • Feb 18, 2015 4:13pm Feb 18, 2015 4:13pm
Quoting drofwarc
{quote} Here you go. All I've done is modify your trade logic slightly to make trades easy to spot. bool candlePatternBearishEngulfing(int ShiftV) { double O = iOpen(Symbol(),ReqTimeFrame,ShiftV); double O1 = iOpen(Symbol(),ReqTimeFrame,ShiftV+1); double C = iClose(Symbol(),ReqTimeFrame,ShiftV); double C1 = iClose(Symbol(),ReqTimeFrame,ShiftV+1); if ((C1>O1)&&(C<O)) //if ((C1>O1)&&(O>C)&&(O>=C1)&&(O1>=C)&&((O-C)>(C1-O1))) return(true); return(false); } //+------------------------------------------------------------------+ bool candlePatternBullishEngulfing(int...
Did you disable all the other patterns in CalcCandleSticks subroutine? bold and underlined to illustrate
Like this

void CalcCandleSticks(int ShiftVal)
{
GetLastClose();
CurrPrice = iOpen(Symbol(),ReqTimeFrame,0);
CandleSticksBuySignal = 0;
CandleSticksSellSignal = 0;
if(candlePatternBullishEngulfing(ShiftVal)== true) CandleSticksBuySignal = 1;
//if(candlePatternDoji(ShiftVal)== true) CandleSticksBuySignal = 1;
//if(candlePatternHammer(ShiftVal)== true) CandleSticksBuySignal = 1;
//if(candlePatternPiercingLine(ShiftVal)== true) CandleSticksBuySignal = 1;
if(candlePatternBearishEngulfing(ShiftVal)== true) CandleSticksSellSignal = 1;
//if(candlePatternShootingStar(ShiftVal)== true) CandleSticksSellSignal = 1;
//if(candlePatternDarkCloudCover(ShiftVal)== true) CandleSticksSellSignal = 1;
}
  • #604
  • Feb 18, 2015 4:19pm Feb 18, 2015 4:19pm
Quoting Dave_C
{quote} If you're referring to the number of trades, I had previously commented out a number of the patterns because I am concerned about the maths for the patterns (remember it is borrowed code). I have made repeated requests for candle stick definitions but nobody has been forthcoming. 15 min charts are not suitable for candlestick trading.
I'm personally not big on trading candlestick patterns as I've never managed to detect a profitable edge using them. My focus at the moment is trading breakouts. In my view, a breakout is a breakout. Of course, everyone likes to see an engulfing bar, but I find it doesn't matter too much what a candle looks like once it's closed above/below resistance/support. That said, now that I'm getting set up to do more rigorous research than I've done in the past, I'm going to have a look at them. However, it may take some weeks before I get there. At present, I don't have trade logic for the different candlestick patterns to share with you. When I do, I'll pass it along.
  • #605
  • Feb 18, 2015 4:26pm Feb 18, 2015 4:26pm
Quoting Dave_C
{quote} Did you disable all the other patterns in CalcCandleSticks subroutine? bold and underlined to illustrate Like this void CalcCandleSticks(int ShiftVal) { GetLastClose(); CurrPrice = iOpen(Symbol(),ReqTimeFrame,0); CandleSticksBuySignal = 0; CandleSticksSellSignal = 0; if(candlePatternBullishEngulfing(ShiftVal)== true) CandleSticksBuySignal = 1; //if(candlePatternDoji(ShiftVal)== true) CandleSticksBuySignal = 1; //if(candlePatternHammer(ShiftVal)== true) CandleSticksBuySignal = 1; //if(candlePatternPiercingLine(ShiftVal)== true) CandleSticksBuySignal...
Dang! That's always the first thing I do when I install a new version of the EA, but forgot this time. I've just disabled the other subroutines and it works fine on H1.

Apologies for putting that on your plate.

I still have to test the other timeframes that you altered and get back to you. Shouldn't take me too long.
  • #606
  • Feb 18, 2015 4:53pm Feb 18, 2015 4:53pm
Okay, I reviewed the EA on the 30, 15 and 5 minute timeframes as per #post 574 and it seems to be working correctly.
  • #607
  • Edited 12:37am Feb 19, 2015 12:18am | Edited 12:37am
I have completely rewritten the CandleStick Pattern Recognition formulas and added extra patterns.

The patterns recognized are

Bullish Engulfing
Bullish Piercing Line
Bullish Kicker
Bullish Harami
Bullish Abandoned Baby
Bullish Morning Star Doji
Bullish Inverted Hammer
Bullish Three White Soldiers
Bearish Engulfing
Bearish Dark Cloud
Bearish Kicker
Bearish Harami
Bearish Abandoned Baby
Bearish Evening Star Doji
Bearish Shooting Star
Bearish Three Black Crows

Also added is a Day of Week filter (in addition to Time of Day), Gold and Silver as currency pairs and MinPipGain display on panel on top left of chart.
If you are using .set files that involve Time of Day filters, you will need to click on "Reset" and redo your inputs.

Download at Post #1

Have added an extra strategy in Trade Explorer - RSI 60min. (view from my profile)
  • #608
  • Edited 11:58am Feb 19, 2015 10:59am | Edited 11:58am
Thank you!

Hi David, I just finished testing the new version. Thanks again for the new day-of-week filter and candlestick patterns. However, the EA is now acting strangely. For one thing, I'm finding that it works correctly when I run a test on the EU. But when I switch the symbol to some of the other pairs, such as the USDCAD and EURNZD, it doesn't work. This occurs on both the every tick and open prices methods

Also, something's going on with the spread. When I test the EU with the spread at current through 20 points, it works. But if I raise the spread above 20, it stops working.

As I write this, it occurs to me that these two issues are related... it seems pairs with spreads wider than 2 pips (20 points) have stopped working.

drofwarc
  • #609
  • Feb 19, 2015 1:01pm Feb 19, 2015 1:01pm
Quoting drofwarc
Thank you! Hi David, I just finished testing the new version. Thanks again for the new day-of-week filter and candlestick patterns. However, the EA is now acting strangely. For one thing, I'm finding that it works correctly when I run a test on the EU. But when I switch the symbol to some of the other pairs, such as the USDCAD and EURNZD, it doesn't work. This occurs on both the every tick and open prices methods Also, something's going on with the spread. When I test the EU with the spread at current through 20 points, it works. But if I...
You need to change the MaxSpread setting.
Attached Image
  • #610
  • Feb 19, 2015 1:16pm Feb 19, 2015 1:16pm
Quoting Dave_C
{quote} You need to change the MaxSpread setting. {image}
Thank you.
  • #611
  • Feb 20, 2015 6:23am Feb 20, 2015 6:23am
If you have been frustrated by the inability to get successful back tests by combining strategies and filters, I have just discovered a major bug in the EA.

Before I release it, I would appreciate a few testers to give me some feedback.

As an example, enclosed are 2 set files and an updated EA which highlights the value of being able to correctly mix strategies and filters. With the EA as it stands at Post #1, no successful results would result in strategy set 2 back test with optimization. With the revised EA, strategy set 2 displays a completely different picture and demonstrates the value of the combination. In this case the first back test was for candlesticks only and the second back test was for candlesticks but with the confirmation of an RSI cross within the previous 5 bars.

Both optimised back tests were EURUSD 2014-01-01 to 2015-02-20.
The first set (Candlesticks only) produced a profit of $32,170
The second set (Candlesticks plus RSI confirmation within last 5 bars) produced a profit of $268,467
Attached Image(s) (click to enlarge)
Click to Enlarge Name: Candlesticks Only.jpg Size: 164 KB
Click to Enlarge Name: Candlesticks plus RSI confirmation within last 5 bars.jpg Size: 153 KB
Attached File(s)
File Type: zip Traders Gift Enhanced.zip   153 KB | 298 downloads
  • #612
  • Feb 20, 2015 9:47am Feb 20, 2015 9:47am
Hi, i have been playing around with testing a bit, i also tested your last two sets with the new enhanced ea, but i get nowhere near your results, and the graphs are just totally different, i put the same date, H1 period, every thick mode, eur usd, traders gift enhanced, and loaded both sets through expert properties, I might just have put the setttings wrong, im not sure

Anyway, I also have a question about optimizing, id be happy to post some results if i get it to work right, i want to check all possible combinations of short and long ema both from 10, stepping up by 10 until 50, i should get many combinations, expect for the ones with the same periods, but i get just a few

thanks
  • #613
  • Feb 20, 2015 10:55am Feb 20, 2015 10:55am
Quoting Dave_C
If you have been frustrated by the inability to get successful back tests by combining strategies and filters, I have just discovered a major bug in the EA. Before I release it, I would appreciate a few testers to give me some feedback. As an example, enclosed are 2 set files and an updated EA which highlights the value of being able to correctly mix strategies and filters. With the EA as it stands at Post #1, no successful results would result in strategy set 2 back test with optimization. With the revised EA, strategy set 2 displays a completely...
Same for me as david026 first screenshot is results that i got.

But i made some modifications and got better results in the second screenshot set file attached .

As you can see Dave as soon as you started working for increasing win ratio (fine tuning entries ) results are getting better.
I really don't know why you have ignored my previous post according to scaling as soon as you wanted to dig in that direction.

Also i got only one error for the whole period. third screenshot.
Attached Image(s) (click to enlarge)
Click to Enlarge Name: Screenshot1.png Size: 147 KB
Click to Enlarge Name: Screenshot2.png Size: 146 KB
Click to Enlarge Name: Screenshot3.png Size: 161 KB
Attached File(s)
File Type: txt TG_EUR_60_Jan2014_Feb2015_EmaEvent_Candlesticks_RSI5Bars_SLMultiplier_TWMultiplier_BreakEvenFilt.txt   13 KB | 270 downloads
  • #614
  • Edited 3:38pm Feb 20, 2015 2:43pm | Edited 3:38pm
Quoting david026
Hi, i have been playing around with testing a bit, i also tested your last two sets with the new enhanced ea, but i get nowhere near your results, and the graphs are just totally different, i put the same date, H1 period, every thick mode, eur usd, traders gift enhanced, and loaded both sets through expert properties, I might just have put the setttings wrong, im not sure Anyway, I also have a question about optimizing, id be happy to post some results if i get it to work right, i want to check all possible combinations of short and long ema both...
I suspect the reason that your results are completely different is because you are using the very tick mode. The EA is explicitly designed to use the model -
"Open prices only (fastest method to analyse the bar just completed, use only for EA's that explicitly control bar opening)"

Whilst you are getting more accurate results, the EA does not look at any activities that occur within the current bar affecting SL calculations which every tick back testing does in fact do. However to do comprehensive optimization on tick data would take weeks, hence the design of the EA.

Also I have enclosed (as .txt) the .set file to achieve the EMA for "i want to check all possible combinations of short and long ema both from 10, stepping up by 10 until 50, i should get many combinations, expect for the ones with the same periods, but i get just a few".

EMA Cross by itself is not a good strategy.

If you are using tick data back testing, make sure the chart time frame is referenced in the ReqTimeFrame (4th input from the end)
Attached Image (click to enlarge)
Click to Enlarge Name: EUR_1_EMA_Nov14_Feb15_Optimization.jpg Size: 171 KB
Attached File(s)
File Type: txt EUR_1_EMA_Nov14_Feb15.txt   13 KB | 183 downloads
  • #615
  • Feb 20, 2015 2:59pm Feb 20, 2015 2:59pm
Quoting alex8585
{quote} Same for me as david026 first screenshot is results that i got. But i made some modifications and got better results in the second screenshot set file attached . As you can see Dave as soon as you started working for increasing win ratio (fine tuning entries ) results are getting better. I really don't know why you have ignored my previous post according to scaling as soon as you wanted to dig in that direction. Also i got only one error for the whole period. third screenshot. {image} {file} {image} {image}
Alex, it wasn't a case if ignoring your post. The considerations went far beyond the scope of the EA in that you talk about opening additional orders and the EA was never designed to handle more than 1 trade at a time. The consideration regarding scaling was an idea to scale out of orders to protect profit.

At this stage I'm focused on addressing concerns relating to the EA as it stands but I would not preclude addressing any idea exclusively.
  • #616
  • Feb 20, 2015 3:33pm Feb 20, 2015 3:33pm
Quoting Dave_C
{quote} I suspect the reason that your results are completely different is because you are using the very tick mode. The EA is explicitly designed to use the model - "Open prices only (fastest method to analyse the bar just completed, use only for EA's that explicitly control bar opening)" Whilst you may feel you are getting more accurate results, in fact, you are getting less accurate results because the EA does not look at any activities that occur within the current bar which may affect SL calculations which every tick back testing does in fact...
Hi David,

I don't follow your meaning here. Are you saying the EA will produce more accurate results in every tick mode or open prices?

Thanks,

drofwarc
  • #617
  • Edited 3:53pm Feb 20, 2015 3:34pm | Edited 3:53pm
Quoting alex8585
{quote} Same for me as david026 first screenshot is results that i got. But i made some modifications and got better results in the second screenshot set file attached . }
That is exactly the .set which demonstrates what we have all been looking for. I ran this .set through actual tick data as you can see with the 99.9% accuracy in the report.

I have enclosed Alex's set file here. Rename to .set. I'm going to start a Trade Explorer next week with this strategy and post on this thread because it looks very impressive. The EA is now at Post #1
Attached Image(s) (click to enlarge)
Click to Enlarge Name: 2015-02-21 07_29_02-Metatrader 4 with Tickstory Launcher (FXT Overwrite Prevention, Fixed spread.jpg Size: 117 KB
Click to Enlarge Name: 2015-02-21 07_29_28-Tester.jpg Size: 92 KB
Attached File(s)
File Type: txt TG_EUR_60_Jan2014_Feb2015_EmaEvent_Candlesticks_RSI5Bars_SLMultiplier_TWMultiplier_BreakEvenFilt.txt   13 KB | 208 downloads
  • #618
  • Feb 20, 2015 3:40pm Feb 20, 2015 3:40pm
Quoting Dave_C
{quote} That is exactly the .set which demonstrates what we have all been looking for. I ran this .set through actual tick data as you can see with the 99.9% accuracy in the report. {image} {image}
Nice! I have an optimization running at the moment on my testing platform (it has the Tickstory data), but I'll run the test later myself.

(On a related note, I'm going to have to start paying more attention to candle patterns.)
  • #619
  • Edited 3:57pm Feb 20, 2015 3:44pm | Edited 3:57pm
Quoting drofwarc
{quote} Hi David, I don't follow your meaning here. Are you saying the EA will produce more accurate results in every tick mode or open prices? Thanks, drofwarc
I think we were both posting at the same time. If you look at Post #614 you can see that I have edited the text.

The point is as you infer, tick data back testing is more accurate. It's just that optimization using tick data is almost impossible except over really short periods. With that in mind, I wrote the EA to use the "Open price" model. Post #617 above is an illustration of a confirmation of the strategy using tick data.
  • #620
  • Feb 20, 2015 3:45pm Feb 20, 2015 3:45pm
Quoting drofwarc
{quote} Nice! I have an optimization running at the moment on my testing platform (it has the Tickstory data), but I'll run the test later myself. (On a related note, I'm going to have to start paying more attention to candle patterns.)
That data is through Tickstory.
5 traders viewing now