Central Banks & Big Players | Page 283 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

How do you know all the positions held by big players? 25 replies

ECN && STP && Scalping && Hedging - Broker Review - Oct 2009 24 replies

Central Banks and Big Players: Slight change in code reqd 7 replies

Japanese housewives are wrecking the forecasts of big players! 25 replies

Big players and Forex 4 replies

Attachments: Central Banks & Big Players
Exit Attachments

Central Banks & Big Players

  • #5,641
  • Dec 7, 2011 3:03pm Dec 7, 2011 3:03pm
Quoting jrodatus
No, it can't work in the MT4 strategy tester. One of the essentials of Tony's method is to refund weekly losses. AFAIK, this cannot be simulated in strategy tester. To paraphrase Tony in post #5430, refunding losses is crucial to making an eventual profit on the account: if you reduce your account to $100 even an 8888 pip run would not refund all losses, let alone make a profit. In other words, by the time you get your first win in the MT4 tester, your account balance will have already been so reduced by all the losses, that the win (bcs it is determined...
Josh,

I'm not concerned with refunding the account for testing purposes. I am trying to see it more simply than that for now. Here is how I intend to run calculations, if I can ever get it to work!

A trade taken can have only one of 3 outcomes:

1. BE (same as 0SL)
2. -2SL
3. Win

Definitions:
1. BE (0SL): Trade is opened, price moves no more than -1.9 pips before hitting at least +2 pips, which allows us to move SL to BE. Price then moves no more than +19.9 pips before reversing to hit SL.

2. -2SL: Trade is opened, price moves no more than +1.9 pips before reversing to hit -2SL, which means we did NOT have a chance to move SL to BE

3. Win: Trade is opened, moves no more than -1.9 pips before hitting at least +20 pips.

Then I want to look at this question to optimize TP:

For all Winning trades, how many reached +20, +25, +50, +75, +140, etc.?

I think the question of refunding losses is more appropriate to address as part of MM and TP optimization.

Now I'm thinking something else: Birt's web page is all about how to convert tick data to use in MT4. I have much more confidence in getting that to work than NinjaTrader. So... if I could do that, would it be possible to write an EA that runs based on those trade definitions? It would have to run on historical data, too. What do you think?

AF
Luck occurs when preparation meets with opportunity.
  • #5,642
  • Dec 7, 2011 4:13pm Dec 7, 2011 4:13pm
Quoting AlienFrog
The next major hurdle to overcome is that I have to convert these csv file to NinjaTrader import format. On that "Birt's EA Review" site there is a link to a script to convert Dukascopy data to NinjaTrader import format (if I am reading it right). But how do I run a .php script on the .csv file?
Hey Alien, do you still have the original .bin files from Dukascopy? That php Birt mentioned actually converts bin->ninja, not csv->ninja. If you got your csv using DukasCopier, then it might've deleted the original bins after processing.

The section on Birt's page entitled "Windows download & convert to CSV how-to" shows you step-by-step how to download the bin files and convert them to CSV all using the PHP scripts. To convert BINs to Ninja's format instead of CSV, just run the process_dukascopy_ninja.php instead of the process_dukascopy_data.php. Again, the instructions for running PHPs on the BIN files are in that section of the article.

Question, did you get your CSV file directly from the Dukascopy history website, or rather using DukasCopier or the PHP scripts?? In my experience, both of the latter generate only 4-digit tick data, with an occasional 5-digit that is always '5' . That throws off the tests- IMO, we might as well be using MetaQuotes data. But I know for a fact that the Dukascopy history center (the flash app) accesses the same .bin files as its data source- same as the scripts/DukasCopier. I can only figure that Birt's decoding algorithm is incomplete...

~Josh

P.S. sorry- didn't mean the programming speak to...uh... alienate you there
Those who can kneel before God, can stand before anyone.
  • #5,643
  • Dec 7, 2011 4:44pm Dec 7, 2011 4:44pm
Quoting AlienFrog
Josh, I'm not concerned with refunding the account for testing purposes. I am trying to see it more simply than that for now. ...
Cool, that makes sense! So basically you would look at the statistical report, and ignore the account balance graph. Got it. Just make sure you set your initial balance to a few billion or something to avoid margin call before the test finishes.

Quoting AlienFrog
Trade is opened, price moves no more than -1.9 pips before hitting at least +2 pips
Actually, because of spread, the trade will open at something like -0.9 pips loss (depending on the spread) and then must move no more than -1.1 pips (whatever would close the 2 pip SL) before hitting at least +2 pips.

Quoting AlienFrog
Birt's web page is all about how to convert tick data to use in MT4. I have much more confidence in getting that to work than NinjaTrader. So... if I could do that, would it be possible to write an EA that runs based on those trade definitions? It would have to run on historical data, too. What do you think?
Assuming the correction above, yeah I can send you my MT4 EA configured to trade the way you defined.

~Josh
Those who can kneel before God, can stand before anyone.
  • #5,644
  • Dec 7, 2011 5:22pm Dec 7, 2011 5:22pm
Quoting jrodatus
Hey Alien, do you still have the original .bin files from Dukascopy? That php Birt mentioned actually converts bin->ninja, not csv->ninja. If you got your csv using DukasCopier, then it might've deleted the original bins after processing.
Actually I used Birt's "easy way" and opened a Dukascopy demo account. It downloads as much data as you want directly to csv. I do not see where I can download to .bin files from within NT. I'll try Birt's scripts.

Quoting jrodatus
P.S. sorry- didn't mean the programming speak to...uh... alienate you there
Hey, it's all good!
Luck occurs when preparation meets with opportunity.
  • #5,645
  • Dec 7, 2011 5:27pm Dec 7, 2011 5:27pm
Quoting jrodatus

Question, did you get your CSV file directly from the Dukascopy history website, or rather using DukasCopier or the PHP scripts?? In my experience, both of the latter generate only 4-digit tick data, with an occasional 5-digit that is always '5' . That throws off the tests- IMO, we might as well be using MetaQuotes data.
Josh,

This is not true, maybe it happened to earlier years, say 2007. But in 2011 you definitely see true 5-digit tick data. For example:

2011.01.05 00:00:00.654,1.33181,1.33192,1650000,1690000
2011.01.05 00:00:01.156,1.33185,1.33193,1990000,1500000

About 3 months ago I downloaded a few pairs and computed median bid-ask spread and I remember for EURUSD it's 0.9, so a lot of the bid/ask rates should be different from four digits or four digits plus a 5.

I haven't touched those bin files for a long time until two weeks ago when I did notice that Dukascopy' history data feed gave volumes in millions (for the above timestamps it would be 1.65, 1.69, 1.99 and 1.5, respectively). Also the column order is different from that of DukasCopier. So be careful not to mix data from these two sources.

Best regards,

screen

Edit: If all 2007 data are either four digits or four+5 then we probably shouldn't back-test year 2007!
  • #5,646
  • Dec 7, 2011 5:36pm Dec 7, 2011 5:36pm
Quoting jrodatus
Just make sure you set your initial balance to a few billion or something...
Ok, np.

Quoting jrodatus
Actually, because of spread, the trade will open at something like -0.9 pips loss (depending on the spread) and then must move no more than -1.1 pips (whatever would close the 2 pip SL) before hitting at least +2 pips.
Thanks, I forgot about the spread. You are right.

Quoting jrodatus
Assuming the correction above, yeah I can send you my MT4 EA configured to trade the way you defined.
This would be awesome! I guess I would still need the .bin files. That's ok, I'll get them.

Do you know you can set the spread in MT4 while offline? Save the symbols.set file in Crogram FilesOANDA - MetaTraderhistory<accountname> folder at a time when the spread is whatever you want it to be. say 1.2 pips. Then open MT4 and disable auto login. Close MT4, copy the saved symbol.set file back to the <accountname> folder - yes, overwrite the existing one - and open MT4. Until you log back in to your account the spread will stay at 1.2. When you log back in the symbols.set file gets updated and overwritten with live info.
Luck occurs when preparation meets with opportunity.
  • #5,647
  • Dec 7, 2011 6:43pm Dec 7, 2011 6:43pm
Quoting screenname
This is not true, maybe it happened to earlier years, say 2007. But in 2011 you definitely see true 5-digit tick data.
I stand corrected, sir! Since 2011 isn't completely available (still need rest of December), I was only utilizing data thru 2010. And my data is 4 digits+5 all the way up through November 2010. Can you confirm this with me?

Quoting AlienFrog
Actually I used Birt's "easy way" and opened a Dukascopy demo account. It downloads as much data as you want directly to csv.
Doh! And I find this out just after I finish deciphering DukaScopy's own .bin parser on their site. Ah well, it was educational anyway

Quoting AlienFrog
Do you know you can set the spread in MT4 while offline? Save the symbols.set file in Crogram FilesOANDA - MetaTraderhistory<accountname> folder at a time when the spread is whatever you want it to be. say 1.2 pips. Then open MT4 and disable auto login. Close MT4, copy the saved symbol.set file back to the <accountname> folder - yes, overwrite the existing one - and open MT4. Until you log back in to your account the spread will stay at 1.2. When you log back in the symbols.set file gets updated and overwritten with live info.
Good to know, thanks. Actually, that solves another reason my EA has problems in the MT4 strategy tester, since it won't trade high spreads.

~Josh
Those who can kneel before God, can stand before anyone.
  • #5,648
  • Dec 8, 2011 7:56am Dec 8, 2011 7:56am
Quoting jrodatus
Never mind, my programming was wrong!

EDIT: Now I can confirm your values. However, Van Tharp evidently said that sample sizes > 100 inflate the SQN so that it does not match the scale (as it would for # of trades < 100). Some say he recommended that you just put 10 in the numerator (i.e. sqrt of 100) for large samples to avoid this. That would lower the SQN value for our tests. But I say arbitrarily capping inputs sounds a little sketchy

~Josh
Probably that's correct, if you just have a history of 100 or a little more trades, but if you want to calculate for 10000 trades for example, then I guess you should average it from an infinite number of combination out-of-sample sets of 100 trades to be more accurate, because you will never know which 100 trades are better to pick if you just wanted to use N=100. And for sure for 10 trades would be insignificant.
Actually I did a little more digging about this: SQN was borrowed from the statistical t-score and it suppose to be chosen any number of trades, then looked-up in a critical values table for that number which will give the level of confidence that your mean (E) is different that 0 or baseline E for the same SD. If the level of confidence stays the same after many trades, then we can trust it even more that our E doesn't come from a normal distribution with E=0, basically. Usually it drops, unless the system is very good.
r u ready?
  • #5,649
  • Dec 8, 2011 8:08am Dec 8, 2011 8:08am
Quoting jrodatus
Brilliant. Good news here

To compensate for the 4-digit chopping error, I set the movement threshold to 2 pips (instead of 1). Then I thought, why not set the trailing SL to 3 pips instead of 2 ? Couldn't hurt to try!....
Josh,
btw I see you used trailing SL for these tests? I thought we move to BE after 5 pips, no trailing, at least that's how I test.
Please someone correct me if I'm wrong: enter in the direction of the first (5 sec) candle after the open - exactly when this candle closes (at 8:00:05 AM EST). SL = 2. Move SL to BE after 5 pips profit.
r u ready?
  • #5,650
  • Dec 8, 2011 11:45am Dec 8, 2011 11:45am
Ok, I am temporarily abandoning NinjaTrader because I found a gigantic section on Birt's page about converting CSV data to MT4 format. You just use the JForex2FXT script. Oh, so easy? Ha! Right! Anyway...

After FINALLY figuring it out and getting it to work (my scars will heal someday) I ran the script in MT4, a 3.2gb EURUSD.fxt file was created, a bunch of .hst files for all TFs were created, I moved them to their respective folders according to Birt's instructions, and I started MT4.

What I've accomplished so far is to import reliable historical data into MT4. Josh how do I now run your EA on historical data, if not in the History Tester? I tried that and it looked like it started and then didn't do anything. Thanks!
Luck occurs when preparation meets with opportunity.
  • #5,651
  • Edited 6:49pm Dec 8, 2011 6:35pm | Edited 6:49pm
Quoting rtrs111
Josh,
btw I see you used trailing SL for these tests? I thought we move to BE after 5 pips, no trailing, at least that's how I test.
Please someone correct me if I'm wrong: enter in the direction of the first (5 sec) candle after the open - exactly when this candle closes (at 8:00:05 AM EST). SL = 2. Move SL to BE after 5 pips profit.
It's pretty safe to assume that Josh sticks to the method defined by its original creator- as close as possible. The basis for my 2-pip trailing stop is the example Tony gave in the official rules on the first page. Also, somewhere in the thread here, Tony defined the "direction of the first candle" as its first movement of 1.0 pip. He also said that if the first candle fails to move one pip, then look at the next one, and so on. He also wrote that he tries to move SL to BE as soon as it hits +2 profit. All of this is effectively what my code does.

Of course, you're welcome to modify my code to do something different...
Those who can kneel before God, can stand before anyone.
  • #5,652
  • Dec 8, 2011 6:47pm Dec 8, 2011 6:47pm
Quoting AlienFrog
Ok, I am temporarily abandoning NinjaTrader because I found a gigantic section on Birt's page about converting CSV data to MT4 format. You just use the JForex2FXT script. Oh, so easy? Ha! Right! Anyway...

After FINALLY figuring it out and getting it to work (my scars will heal someday) I ran the script in MT4, a 3.2gb EURUSD.fxt file was created, a bunch of .hst files for all TFs were created, I moved them to their respective folders according to Birt's instructions, and I started MT4.

What I've accomplished so far is to import reliable historical...
Wow, good work Alien! Hopefully what I'm about to say isn't bad news... you do need history files for the actual tick data, not just timeframes. Remember, the strategy relies on split-second price action momentum.

Probably set the GMToffset option to 0, because that's what Dukascopy's is. Make sure you select "Every tick" in the strategy tester. Then click the "Journal" tab at the bottom of the window and hit start. The EA should report what it's doing in the Journal- let me know what it says. If it's complaining about the spread, then you could use the trick you gave me earlier to set the spread. I might not get back to you till tomorrow evening, cause of work.

~Josh
Those who can kneel before God, can stand before anyone.
  • #5,653
  • Dec 9, 2011 2:04am Dec 9, 2011 2:04am
a little late london entry
Attached Image (click to enlarge)
Click to Enlarge Name: ee4.gif Size: 7 KB
Invest in alarm clocks
  • #5,654
  • Dec 9, 2011 2:18am Dec 9, 2011 2:18am
Do you really earn money with that strategy ? I imagine trading this style is very annoying and psychological painful.
If you add proper MM than even worse. How Do you handle your draw downs ? I am sure you are not profitable to trade let'S say 1 Lot every day to London open depending on setting the TP.
Too much trades by chance / randomness . You cannot optimize the system by prior Highs/lows if market rallies.

I'm watching the system for half a year now, interesting idea but I think it's not working. Backtested it one time (for a year). Chance Risk was 1.2 on average.

The big advantage by this system is the possibility to use full margin and if a win occurs (say 80pips) than 1.2 Faktor may be suitable to make a living from it.

Tony, what's your impression ?
  • #5,655
  • Dec 9, 2011 2:21am Dec 9, 2011 2:21am
20 mins into London open
Attached Image (click to enlarge)
Click to Enlarge Name: ee5.jpg Size: 52 KB
Invest in alarm clocks
  • #5,656
  • Dec 9, 2011 2:25am Dec 9, 2011 2:25am
I made money with the system and bought a lot of fancy things with it in the last year. No draw down. I reimburse every loss, every week. The only way I lost money is by using oanda as a casino, which I don't do anymore.


Quoting JackintheBox
Do you really earn money with that strategy ? I imagine trading this style is very annoying and psychological painful.
If you add proper MM than even worse. How Do you handle your draw downs ? I am sure you are not profitable to trade let'S say 1 Lot every day to London open depending on setting the TP.
Too much trades by chance / randomness . You cannot optimize the system by prior Highs/lows if market rallies.

I'm watching the system for half a year now, interesting idea but I think it's not working. Backtested it one time (for a year)....
Invest in alarm clocks
  • #5,657
  • Dec 9, 2011 4:13am Dec 9, 2011 4:13am
Thanks for reply Tony !

Ok, I see, you reimburse your losses but have you looked at the performance without reimbursing ?

I mean if you reimburse 15000 Dollars a year and loose 25000 dollars by the system and you earn 20000 dollars by the system you have (15000-25000+20000) 10000 dollars "net profit" which is not correct cause you have reimbursed 15000...->so a loss of 5000 $, am I right ?

So if you have a "profit faktor" of 1.7 for last year may be without reimbursing it's just 1.2, do you now what I mean ?

greets Jack

I wish you luck, that market will not turn now !
  • #5,658
  • Dec 9, 2011 4:16am Dec 9, 2011 4:16am
Quoting Tony112
I made money with the system and bought a lot of fancy things with it in the last year. No draw down. I reimburse every loss, every week. The only way I lost money is by using oanda as a casino, which I don't do anymore.
Hi Tony!
It is very good that you have success with your method.
Did you change the stop loss? It looks if it would be a little bigger than 2 pips plus spread?
What stop do you use now?

Regards Erkro
  • #5,659
  • Dec 9, 2011 5:16am Dec 9, 2011 5:16am
It hit +134.2 pips 5 mins ago. Going up now....
Luck occurs when preparation meets with opportunity.
  • #5,660
  • Dec 9, 2011 6:04am Dec 9, 2011 6:04am
Quoting jrodatus
Wow, good work Alien! Hopefully what I'm about to say isn't bad news... you do need history files for the actual tick data, not just timeframes. Remember, the strategy relies on split-second price action momentum.

Probably set the GMToffset option to 0, because that's what Dukascopy's is. Make sure you select "Every tick" in the strategy tester. Then click the "Journal" tab at the bottom of the window and hit start. The EA should report what it's doing in the Journal- let me know what it says. If it's complaining about the spread, then...
It worked!!!! It actually ran!!! I am dancing circles around my desk!

I still have some things to work out that will take time though.

I got Journal results for trades over a 1 year run. That appears to be working great.

I don't know how to confirm that it is using the test data I downloaded - maybe delete it, then try to run the EA again and see if it fails?

I assume it is using tick data from within the TF files. When I ran the JForex2FXT script it created a big .FXT file, and a bunch of .HST files like EURUSD1.HST, EURUSD5.HST, EURUSD30.HST, etc. These appear to be time frame files, and I assume they are generated from the downloaded tick data.

It is overwriting my FXT file. Birt says this is a problem and lists a fix, which I will try later. I don't know why that's a problem.

EDIT: Also, here is a pic of the tester Report page. It shows red for the mismatched chart errors, and 25% tick modeling quality. Is this a problem? I don't know what it really means but it looks bad.

EDIT2: The Results page only shows negatives or zeros on the Profit column, although the Report page shows that there were some winning trades. Journal page also shows a lot of the following lines:

"2011.12.09 06:58:10 2011.12.02 13:00 Tony112 EURUSD,M1: Unable to trade: openTrade() reported error 131"

I will have to solve these issues to have confidence that the tests are accurate. But this was a MAJOR milestone this morning!

Josh, thanks for your help so far! We'll get this to work and make some pips!

AF
Attached Image (click to enlarge)
Click to Enlarge Name: pic1.jpg Size: 77 KB
Luck occurs when preparation meets with opportunity.
7 traders viewing now