NFP System - Backtest data provided | Page 5 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

How to Correctly BackTest, Collect Data, and Analyze A System 13 replies

How much historical data needed to backtest day trading system 0 replies

Visual backtest result different from non-visual backtest 0 replies

data needed to backtest system 1 reply

How to backtest system with past data? 8 replies

Attachments: NFP System - Backtest data provided
Exit Attachments

NFP System - Backtest data provided

  • #81
  • Apr 2, 2007 8:08am Apr 2, 2007 8:08am
I read the whole thread twice but didnt bother to read the text file in first post.... Thats stupid of me i guess..

Thanks
Quoting hksweeper
Had you download the file in the first message on the first page and read the rules?
Thanks and Regards Zamir M Iqbal
  • #82
  • Apr 2, 2007 8:15am Apr 2, 2007 8:15am
what about profit target

If i get it correctly when price drops below MA 336 then you close it?

And it is a bit tricky to follow it all the time (char is 30min), I mean EA should do it better then me if I cant watch the screen and charts 24H a day.
So i guess it would be good thing to even write EA just for closing (if not the whole thing)
  • #83
  • Apr 2, 2007 11:18am Apr 2, 2007 11:18am
Hello all,

This system looks very promising. Like MrMikal, I've been looking at some similar math, but more from an arbitrage position. So, I was excited to take a look.

I have, perhaps, a silly question. It seems like there should only be ONE index number. In other words, when one does the calculation, by combining the 7 pairs, you end up with a single index number. If that's true, does anyone know why I might be getting different index numbers using the indicator depending on the chart I have up? I have all 7 pairs up, each with the sentinel index indicator on it (SMA version). Yet, my current index value for GBPUSD is 2.1782 but AUDUSD it's 1.8137. All other pairs have unique values.

Thanks.

stockwet
  • #84
  • Apr 2, 2007 12:47pm Apr 2, 2007 12:47pm
Sockwet,

That's a potentially big problem there...if it gives different values, then odvoiusly there is something wrong with the indicator. You're right, all 7 values should come up with one straightforward Sentinal number. I'd hate to be sending you guys mixed signals (har har har), so I'll look at it today and see what the problem might be.

Ahil, there is a free service called AlertFX.com (or used to be free anyways, I haven't checked as of late) that will let you know -- via cell phone txt message or email -- when a price has crossed a SMA that you provide. Also, it watches specific numbers and lets you know if you've hit those (good for stops and limits, so you don't even have to be behind your computer to know what you just made ). Give that a try...but your right, the cross could come while you're sleeping, so something automated isn't really neccesarily all that bad of an idea, at least on the exit portion.

zamiriqbal, download the file on the first page, and read it The rules file lets you the backbone to everything you'll need; the rest of it has been written here by your trading brothers.

-Kev
Savage Pip Fiend
  • #85
  • Apr 2, 2007 1:16pm Apr 2, 2007 1:16pm
hiksweeper,

Good eye..almost All of the following is looking at the GBP/USD pair. With that in mind...

Last month would have been good for this system...breakeven on initial trade (up 50 pips, or 45 if you count spread), then you have your first re-entry at 3/16/2007 at 13:30, because the price on 3/2/2007 at 13:30 (2 weeks prior..just so happens to be the previous' 2 weeks high as well) is below the close of the candle on 3/16/2007 at 13:30. Notice we didn't take the trade when the candles broke the high. They broke it a few times, but the 13:30 candle was the first to close above.

That trade would have been good for 130 pips aprox. (close on other side of 336 SMA and exit on 3/28/2007 at 17:30). So for last month, we would have had about 170 after spread. There is one more re-entry at 4/2/2007 at 09:00. This closes above the high of the last 2 weeks, established on 3/22/2007 at 10:30. As of now, we've hit our 50 pip breakeven point, so tack another 45 onto the total pips, and what do you know? We get our average...215 pips a month That is, of course, purely coincidental lol. I probably would have not taken that last trade. The Sentinal number, if you bother charting it outside of the MetaTrader platform, has created some unique patterns. Its having a hard time pushing above this current level of resistance, but it just punched above it when the last trade was triggered. That's too close for me. Also, you'll notice that on 3/26, the Sentinal number punched below the SMA, and again (even lower this time) mid day on 3/29, and the patterns lead me to believe we're consolidating and going to be picking a direction. A major move that late in the month? I wouldn't have bet on it, so I would have kept my 170 for the month.

Anyways, there's my quick rundown of last month

-Kev
Savage Pip Fiend
  • #86
  • Apr 2, 2007 1:43pm Apr 2, 2007 1:43pm
Hello

I still canot figure it out how, NFP data or numbers are relative to this method / system.... what does NFP has to do with the system??

Any help on above will be appreciated??


Quoting Kevin_E
hiksweeper,

Good eye..almost All of the following is looking at the GBP/USD pair. With that in mind...

Last month would have been good for this system...breakeven on initial trade (up 50 pips, or 45 if you count spread), then you have your first re-entry at 3/16/2007 at 13:30, because the price on 3/2/2007 at 13:30 (2 weeks prior..just so happens to be the previous' 2 weeks high as well) is below the close of the candle on 3/16/2007 at 13:30. Notice we didn't take the trade when the candles broke the high. They broke it a few times, but the 13:30 candle was the first to close above.

That trade would have been good for 130 pips aprox. (close on other side of 336 SMA and exit on 3/28/2007 at 17:30). So for last month, we would have had about 170 after spread. There is one more re-entry at 4/2/2007 at 09:00. This closes above the high of the last 2 weeks, established on 3/22/2007 at 10:30. As of now, we've hit our 50 pip breakeven point, so tack another 45 onto the total pips, and what do you know? We get our average...215 pips a month That is, of course, purely coincidental lol. I probably would have not taken that last trade. The Sentinal number, if you bother charting it outside of the MetaTrader platform, has created some unique patterns. Its having a hard time pushing above this current level of resistance, but it just punched above it when the last trade was triggered. That's too close for me. Also, you'll notice that on 3/26, the Sentinal number punched below the SMA, and again (even lower this time) mid day on 3/29, and the patterns lead me to believe we're consolidating and going to be picking a direction. A major move that late in the month? I wouldn't have bet on it, so I would have kept my 170 for the month.

Anyways, there's my quick rundown of last month

-Kev
Savage Pip Fiend
Thanks and Regards Zamir M Iqbal
  • #87
  • Apr 2, 2007 1:48pm Apr 2, 2007 1:48pm
are there any backtesting results of last few months available? Because IIRC your document doesn't cover the last time.
  • #88
  • Edited 3:39pm Apr 2, 2007 1:48pm | Edited 3:39pm
OK. I think I found the issue. Maybe one of the developers can comment.

It looks like there's an issue with:

Inserted Code
Index[i] = Index[i+1] + x
If I remove that, and simply use:

Inserted Code
Index[i] = 100 + x
everything works OK. For some reason, that line is grayed out. The reason I use 100 is because Kevin said to use a base that you add to the sentinal values.

However, I have to say that I like the signals for the previous indicator. I'm just not sure they are accurate.
  • #89
  • Apr 2, 2007 1:49pm Apr 2, 2007 1:49pm
Did you read the text file that was included with Kevin's zip file on the first post?

Quoting zamiriqbal
Hello

I still canot figure it out how, NFP data or numbers are relative to this method / system.... what does NFP has to do with the system??

Any help on above will be appreciated??
  • #90
  • Apr 2, 2007 2:05pm Apr 2, 2007 2:05pm
I should also clarify that you shouldn't use either of these files until the original developers comment. The reason is 1) I may be doing something entirely incorrect and 2) I don't know if my modification will work in different modes - I've only looked at in using the pips mode.

Quoting stockwet
OK. I think I found the issue. Maybe one of the developers can comment. I created a test file, which I have attached.

It looks like there's an issue with:

Inserted Code
Index[i] = Index[i+1] + x

If I remove that, and simply use:

Inserted Code
Index[i] = 100 + x

everything works OK. For some reason, that line is grayed out. The reason I use 100 is because Kevin said to use a base that you add to the sentinal values.

However, I have to say that I like the signals for the previous indicator. I'm just not sure they are accurate.

Also, I included my mod which allows you to draw the indicator as a histogram.
  • #91
  • Apr 2, 2007 2:06pm Apr 2, 2007 2:06pm
Dear Stockwet

I did, but couldnt figure out the NFP part, how is it relevant to the system??

If u cud explain ...thx

Quoting stockwet
Did you read the text file that was included with Kevin's zip file on the first post?
Thanks and Regards Zamir M Iqbal
  • #92
  • Apr 2, 2007 2:40pm Apr 2, 2007 2:40pm
Here you go - from the txt file:

Quote
1. Draw box (horizontal line on high and low) between 8:30 on NFP Friday and midnight the following Monday.

The theory behind the system is that the NFP provides the primary direction for the month, for the USD. Therefore, the box represents the primary boundary you look at for breakouts. The other indicators are confirming indicators.


Quoting zamiriqbal
Dear Stockwet

I did, but couldnt figure out the NFP part, how is it relevant to the system??

If u cud explain ...thx
  • #93
  • Apr 2, 2007 2:45pm Apr 2, 2007 2:45pm
OK. I will have to amend my statement. In truth, reason number 1 in my earlier post seems to be the correct - I am doing something wrong. I re-read the original definition of the sentinal number and realized that you need to add the new value to the previous value. It's a chained or cumulative value, which is why the statement:

Inserted Code
Index[i] = 100 + x

is incorrect. So, if

Inserted Code
Index[i] = Index[i+1] + x

is correct, does anyone know why it would produce inconsistent results between different graphs? I'm back to where I started on testing.


Quoting stockwet
I should also clarify that you shouldn't use either of these files until the original developers comment. The reason is 1) I may be doing something entirely incorrect and 2) I don't know if my modification will work in different modes - I've only looked at in using the pips mode.
  • #94
  • Apr 2, 2007 3:06pm Apr 2, 2007 3:06pm
  •  iya
  • | Joined Oct 2006 | Status: always protect yourself | 129 Posts
Stockwet, you're right, it's an accumulating indicator. So the reason for different values is the different start date of each chart. It shouldn't affect the relative motion though, and could be corrected with the BIAS value.
Missing history data is another problem. I'm not sure about the best way to handle it, but I'll try to make a version that is a little more robust.

About problems with backtesting/EA, maybe this is the reason:
Quote
At the testing, price data of the same symbol but of another timeframe are modelled exactly, with the exception of volumes. Volumes of another timeframe are not modelled. Price data of other symbols are not modelled, either. In all cases, the amount of bars in time series is modelled exactly.
  • #95
  • Apr 2, 2007 3:11pm Apr 2, 2007 3:11pm
The whole primace behind the system was to catch a move of the NFP. The NFP, in Rob's eyes, is a great market catalyst capable of slowing down trends, reversing a trend, making a pair consolidate...We don't need to know the NFP numbers. We just need to know that price will be effected by the NFP, and try to ride the wave in the correct direction.

Taking trades on a breakout of 8:30 NFP Friday forward to midnight on Monday guarentees that a move in a big nature will most likely be the result of the NFP report, since we aren't paying any attention to data that came out before the 8:30 candle, erego before the report. So if it breaks the box and you take your trade, chances are that the movement's catalyst was the NFP report....hence....trading the NFP report Its not all number based, its all price action. You're looking at purely fundamental trading, this is a 20% fundamental 80% technical system probabl.y Hope that answers your question.

And as for changing the code to what you did sockwet, sorry bro, that won't work. See, you're in what we coders call a FOR loop. That code is executed, looks like to me, once for every bar. They pretty much start at the beginning, calculate the sentinal number for each bar, and decrement the number of bars until it is 0. Your code pretty much is nothing but Index[i] = x; that he had up there before, but with a 100 offset, so you'll get the same chart with different numbers. His code basically calculates the Sentinal number by taking the previous calculated sentinal number and adding your average to it to get the current one. Logistically, its not the same EXACT way I did it before, but in theory should be the same chart, simply with different numbers. So, when the 336 SMA is layed over it to get our signal, you should get the same 336 SMA as well relative to your sentinal number, giving you the same signals.

As for the real difference between the numbers, and why they are different, somebody else is going to have to comment on that. I know C/C++, not MQL4 yet

-Kev
  • #96
  • Apr 2, 2007 3:18pm Apr 2, 2007 3:18pm
Yeah, I think I've looked at your code 5 different times now, and I can't see anything that would cause things to be thrown off.

The only concern I have about your statement relates to this last test I ran. I hand calculated the sentinal value. In other words, I created a var that calculated the raw sentinal value for the current bar. Then, I created a var that calculated the raw sentinal value for the previous bar. Then, created a third var that added them together and then plotted it. I still came up with a consistency across all the different charts. On the other hand, it still doesn't quite feel right.

So, I tried something else. I made the number of bars to count a consistent number - 600. That did the trick.

So, my suggestion would be to have a max number of bars to count, which would ensure that every chart has a consistent indicator.

Summary: I would suggest adding 1) the ability to view as a histogram and 2) the option to count a specific number of bars (must be greater than 336, though.)

Quoting iya
Stockwet, you're right, it's an accumulating indicator. So the reason for different values is the different start date of each chart. It shouldn't affect the relative motion though, and could be corrected with the BIAS value.
Missing history data is another problem. I'm not sure about the best way to handle it, but I'll try to make a version that is a little more robust.
  • #97
  • Apr 2, 2007 3:27pm Apr 2, 2007 3:27pm
just my two penneth,

I am no programmer, but what about chosing some specific date in the past and simply inputting the combined values of the seven pairs at that date as the base sentinel value.

One could then reinput this base value with regards to each consequent bars unique average movement away from said base.



surely this would create a single value across the spectrum without limiting bar numbers, which would always change the sentinel's base.
quote of the day: . It's like deja-vu all over again
  • #98
  • Edited Apr 11, 2007 8:22pm Apr 2, 2007 3:37pm | Edited Apr 11, 2007 8:22pm
Well, after thoroughly confusing everyone with my posts, I might as well post the mods I suggested. This is the same code as the original EA (SMA version), with 2 additions.

1) You can choose to view the Sentinal values as a histogram instead of a line.

2) You can select the number of bars back to go. Remember, you need at least 336 bars in order to draw the SMA. So, pick a big number. By default, it is set to 0, which means all bars in the chart will be used and you will have the same effect as previously - that is, different charts will have varying Sentinal numbers. However, as iya pointed out, the different numbers should not impact the crossings and the overall curve of the dataset.

Regards and sorry for the confusion.

stockwet

PS: Great work on this EA, guys. This code is EXTREMELY clean and easy to read.
  • #99
  • Apr 2, 2007 3:49pm Apr 2, 2007 3:49pm
Hi Kev,

I'm actually quite proficient with MQL and found the code quite readable. The issue wasn't that I was in a FOR loop. Rather, I inaccurately evaluated your definition of how to calculate the value. I missed that the Sentinal value is a cumulative value. That's why iya's statement makes sense - the starting points on each chart are different, which throw off the cumulative result of the numbers.

Also, as I know you're new to MQ4. Any code within the "start" function of MQ4 executes every tick, rather than every bar. So, the value is actually being recalculated on every tick movement that comes in. But, since we're looking at BAR level data, you don't see any changes. As you get into MQ4, you'll realize that that is an important point - especially when you program an alert and wonder why the darned thing throws an alert every time a tick comes in.

Anyway, kudos on the system. It looks fantastic and I can't wait to back and forward test it. You've done a fabulous job.

stockwet

Quoting Kevin_E
And as for changing the code to what you did sockwet, sorry bro, that won't work. See, you're in what we coders call a FOR loop. That code is executed, looks like to me, once for every bar. They pretty much start at the beginning, calculate the sentinal number for each bar, and decrement the number of bars until it is 0. Your code pretty much is nothing but Index[i] = x; that he had up there before, but with a 100 offset, so you'll get the same chart with different numbers.
  • #100
  • Apr 2, 2007 7:46pm Apr 2, 2007 7:46pm
Hi:
Thanks for posting modified version of indicator.
For some reason, I can't not make indicator to show on my chart.
Am I missing something obvious?
Does anyone having same problem? I didn't change any setting. Just left it as is.

Quoting stockwet


PS: Great work on this EA, guys. This code is EXTREMELY clean and easy to read.
1 trader viewing now