Old Dog with New Tricks | Page 14 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Old Dog's Taming of the Beast? 34 replies

Old Dog Fib Trading - EA Development 18 replies

Old Dog Fib Trading - EURUSD 399 replies

5 Tricks of Trend Trading, by Rob Booker 25 replies

How To Work Around A Market Maker's Tricks 0 replies

Attachments: Old Dog with New Tricks
Exit Attachments

Old Dog with New Tricks

  • #261
  • Jun 15, 2010 1:07pm Jun 15, 2010 1:07pm
Ron,


Here is how i understand the strategy based on the the EA implementation:

- every day at midnigt (00:00), the EA "re-optimizes" itself;

- then it uses the resulting "trading style" (breakout or counter trend) for the entire day, until next midnight;

- "re-optimization" takes place over the past "BarStart" bars, and does this:

- each day at midnight, a first high/low range is determined over the "InitialRange" bars preceeding midnight; yellow vertical lines are drawn to show this initial range period;
- then, after midnight, this high/low range widens when price makes a higher high or a lower low;
- when price makes a higher high or a lower low, we test the outcome of a breakout trade and a counter-trend trade that would be sent at that time and price, and update win/loss counters for both styles;
- once we have covered the full "BarStart" period, we compare the win/loss results for breakout style and counter-trend style, and we select the best for the rest of the day, provided that the result ratio meets the expected inputs (WinPercent for instance).


There is an exception to the "midnight re-optimization rule": it optimizes immediately when you start the EA (it will not wait until midnight to estimate the first trading style).


If my interpretation is correct, then i have the following remarks/questions:


- the hi/lo range widens all day long and is reset only at midnight the next day;
meaning that the more we get into the day, the less we will be able to determine if market is moving to ranging type (and therefore take counter-trend trades rather than breakout trades ?)
If price starts retracing, then it may take very long before it breaks the other side.
Having a way to "refocus" the hi/lo range at regular intervals over the same day (instead of only once a day at midnight) would benefit, no ?
Perhaps a "ReOptimizeInterval" input could be interesting ?
Based on a duration rather than a number of bars to make it independent on the timeframe.
At the moment, we could say it's hard-coded at 24-hours interval.


- there is no way to tell the EA what trading hours we want to use, which would also be taken into account for the optimization process;


- on MONDAYS we have an issue: if i set the BarStart to be less than 2 days, say 300 bars on M1 chart (optimizing will take place only over the last 5 hours), then we don't get any trades on monday because the bars examined will be saturday and sunday bars, ending at 00:00 on monday morning.
At least this is what i observed in (visual) backtest.
If we set BarStart to be more than 2 days, then we still go over useless saturday and sunday bars.
I think we should exclude saturday and sunday from the examined bars.
(and from the bars used to determine the outcome of test trades also).


- "midnight" is estimated in local time.
Meaning that we will not have the same EA behaviour all over the planet, due to the hard-coded midnight trigger.


I hope i'm not completely mistaken here...


Sq.
  • #262
  • Jun 15, 2010 2:22pm Jun 15, 2010 2:22pm
Quoting squalou
Ron,


Here is how i understand the strategy based on the the EA implementation:

- every day at midnigt (00:00), the EA "re-optimizes" itself;
Should be midnight broker time. Basically, while the EA is waiting for the initial range to form, it reoptimizes itself.

Quoting squalou

- then it uses the resulting "trading style" (breakout or counter trend) for the entire day, until next midnight;
Yes

Quoting squalou
- "re-optimization" takes place over the past "BarStart" bars, and does this:
It's more a backtest than an optimization, but yes.

Quoting squalou
- each day at midnight, a first high/low range is determined over the "InitialRange" bars preceeding midnight; yellow vertical lines are drawn to show this initial range period;
No. InitialRange bars are the number of bars after midnight (day start).

Quoting squalou
- then, after midnight, this high/low range widens when price makes a higher high or a lower low;
Yes

Quoting squalou
- when price makes a higher high or a lower low, we test the outcome of a breakout trade and a counter-trend trade that would be sent at that time and price, and update win/loss counters for both styles;
Quoting squalou
- once we have covered the full "BarStart" period, we compare the win/loss results for breakout style and counter-trend style, and we select the best for the rest of the day, provided that the result ratio meets the expected inputs (WinPercent for instance).


There is an exception to the "midnight re-optimization rule": it optimizes immediately when you start the EA (it will not wait until midnight to estimate the first trading style).
Correct.

Quoting squalou
If my interpretation is correct, then i have the following remarks/questions:


- the hi/lo range widens all day long and is reset only at midnight the next day;
meaning that the more we get into the day, the less we will be able to determine if market is moving to ranging type (and therefore take counter-trend trades rather than breakout trades ?)
If price starts retracing, then it may take very long before it breaks the other side.
Having a way to "refocus" the hi/lo range at regular intervals over the same day (instead of only once a day at...
The theory behind the system the EA is trading is as follows:

"Whenever a new daily high or daily low is made, the price does one of two things:

 

  1. Break out
  2. Retrace

The purpose of this system is to decide whether or not a daily high is more likely to breakout than to retrace."

I never meant to code this for the purpose you're outlining, but this EA can certainly be modified to do that.

Quoting squalou

- there is no way to tell the EA what trading hours we want to use, which would also be taken into account for the optimization process;
Correct. I didn't include trading hours because that would require more 'instances' of EAs, and this version of the testing engine only supports two (breakout, and retrace/counter).

Quoting squalou
- on MONDAYS we have an issue: if i set the BarStart to be less than 2 days, say 300 bars on M1 chart (optimizing will take place only over the last 5 hours), then we don't get any trades on monday because the bars examined will be saturday and sunday bars, ending at 00:00 on monday morning.
At least this is what i observed in (visual) backtest.
If we set BarStart to be more than 2 days, then we still go over useless saturday and sunday bars.
I think we should exclude saturday and sunday from the examined bars.
(and from the bars used...
The EA calculates its starting times based on shift. The bar shift is not converted to time.

Quoting squalou

- "midnight" is estimated in local time.
Meaning that we will not have the same EA behaviour all over the planet, due to the hard-coded midnight trigger.


I hope i'm not completely mistaken here...


Sq.
Midnight should be when the broker's clock strikes 00:00.



These are all excellent questions. Keep them coming!
  • #263
  • Jun 16, 2010 4:27am Jun 16, 2010 4:27am
Ron,

i found a bug in the way the high/low values are determined (took me quite a while to figure out why these @!?@$! iHighest/iLowest() functions did not give me the correct values!!!!)

Here it is:

this code:
Inserted Code
            HighShift = iHighest(NULL, 0,[b] [color=Red]PRICE_HIGH[/color][/b], DayStartShift - RangeEndShift, RangeEndShift);
            HighValue = iHigh(NULL, 0, HighShift);
            LowShift = iLowest(NULL, 0, [b][color=Red]PRICE_LOW[/color][/b], DayStartShift - RangeEndShift, RangeEndShift);

should be:
Inserted Code
            HighShift = iHighest(NULL, 0, [color=Red][b]MODE_HIGH[/b][/color], DayStartShift - RangeEndShift, RangeEndShift);
            HighValue = iHigh(NULL, 0, HighShift);
            LowShift = iLowest(NULL, 0, [b][color=Red]MODE_LOW[/color][/b], DayStartShift - RangeEndShift, RangeEndShift);

sq.
  • #264
  • Jun 18, 2010 4:01am Jun 18, 2010 4:01am
I have been testing this EA for a while and really like it. Thank you for the contribution Ron.

I agree that making it more dynamic and sensitive to fluctuating market conditions would increase profitability. The only times I really see a problem is when new markets open and close. The back test show a counter trend as the better option then things breakout, sometimes missing the big moves.

I wonder if it would be possible add market times to the test variables so it could better predict daily patterns and when when to switch strategy. Would be cool to add news too.

Is this an EA you plan on trying to auto update with you tick data when you figure that out?

Thanks
  • #265
  • Jun 18, 2010 10:03am Jun 18, 2010 10:03am
I also like this EA very much. I was testing it when RR first introduced it, but I was having problems because of it using a lot of resources.

I decided when I got a new computer, I would start testing again. I got a new computer last week and I set it up yesterday. Very nice.

Thanks to you, RR.

Carol
  • #266
  • Jun 18, 2010 10:41am Jun 18, 2010 10:41am
Cheers rfking2 and Eureka.

rfking2, I actually have already done that for a self-optimizing EA. There's a hitch however, I do not want to risk having the trolls and promoters who watch this space steal and sell the EA. I'm currently acquiring international copyright certificates, and moving the bulk of the code to various DLLs.

If you have any more suggestions for DRM-like methods I'd be more than happy to hear them. Until I'm satisfied that there is no chance anyone can successfully sell the EA, I will not release it.
  • #267
  • Jun 18, 2010 3:21pm Jun 18, 2010 3:21pm
I was wondering how many currencies do you trade with using this EA?

I am super impressed with this EA. It's a real winner!!

Thanks, RR for presenting it to the board.

Carol
  • #268
  • Jun 18, 2010 3:51pm Jun 18, 2010 3:51pm
Quoting Eureka
I was wondering how many currencies do you trade with using this EA?

I am super impressed with this EA. It's a real winner!!

Thanks, RR for presenting it to the board.

Carol
Every one.
  • #269
  • Jun 18, 2010 5:33pm Jun 18, 2010 5:33pm
Maybe you could store the sensitive code on your server and make a dll that calls to it when the mt4 is compiled. That way you would not have to worry about pirates.

Is that possible?

Another idea for DRM would come from the Mdriod bot. They have the EA call to their server to ensure a license exists.
  • #270
  • Jun 21, 2010 1:35pm Jun 21, 2010 1:35pm
I'm using this EA with SL=TP=30 on E/U and G/U and it have a loss after 3 weeks !!
It trade counter trend when the market is flat and also when is strong trend ;(
What should I change? Can anyone post the correct settings?
LeW
  • #271
  • Jun 22, 2010 12:40pm Jun 22, 2010 12:40pm
I keep getting an error in the Experts tab. Anyone else getting this?

2010.06.22 13:36:06 HiLo Trader 2 [TD] EURUSDm,M1: EURUSD Error:
<br />
<b>Warning</b>: mysql_connect() [<a href='function.mysql-connect'>function.mysql-connect</a>]: User jedcaban_Feeder already has more than 'max_user_connections' active connections in <b>/home8/jedcaban/public_html/ronaldraygunforex/TickDB/Load.php</b> on line <b>11</b><br />
Could not connect: User jedcaban_Feeder already has more than 'max_user_connections' active connections
  • #272
  • Jun 22, 2010 12:57pm Jun 22, 2010 12:57pm
Quoting Eureka
I keep getting an error in the Experts tab. Anyone else getting this?

2010.06.22 13:36:06 HiLo Trader 2 [TD] EURUSDm,M1: EURUSD Error:
<br />
<b>Warning</b>: mysql_connect() [<a href='function.mysql-connect'>function.mysql-connect</a>]: User jedcaban_Feeder already has more than 'max_user_connections' active connections in <b>/home8/jedcaban/public_html/ronaldraygunforex/TickDB/Load.php</b> on line <b>11</b><br />
Could not connect: User jedcaban_Feeder already has more than 'max_user_connections' active connections
A lot of people are starting to use this and other [TD]-based EAs I'm reaching the capacity in the tick database server. Currently trying to expand available capacity.
  • #273
  • Jun 22, 2010 1:21pm Jun 22, 2010 1:21pm
Quoting Ronald Raygun
A lot of people are starting to use this and other [TD]-based EAs I'm reaching the capacity in the tick database server. Currently trying to expand available capacity.
Thanks for explaining.
  • #274
  • Jun 22, 2010 1:25pm Jun 22, 2010 1:25pm
I just stopped a test on a demo trading every currency on every time frame. I doubt I single handedly over loaded the server, but maybe it will help
  • #275
  • Jun 22, 2010 1:31pm Jun 22, 2010 1:31pm
How was your testing going?
  • #276
  • Jun 22, 2010 1:40pm Jun 22, 2010 1:40pm
Currently roughly 2,000 instances of HiLo trader alone hitting the server. I don't think you're single-handedly causing it.
  • #277
  • Jun 22, 2010 1:42pm Jun 22, 2010 1:42pm
Quoting Eureka
How was your testing going?
Going good. I was thinking that having it on multiple time frames would be like doubling down on a good signal but the shorter time frames usually take the same trades at the same time. So there is no point in having many open. As I have said before in this thread, it is profitable, there are some improvements that can be made. But Ron is already on top of that.
  • #278
  • Jun 22, 2010 1:49pm Jun 22, 2010 1:49pm
Thanks for the info, rfking2. My testing is also profitable. I would have to say since there are that many instances hitting the server, others must be finding it profitable as well.
  • #279
  • Jun 22, 2010 1:49pm Jun 22, 2010 1:49pm
Hey RR, after the EA has been running for a while you don't need the tick database anymore right? My impression was the tick database was for priming the pump when the EA first starts up.
You are in a maze of twisty little passages, all alike.
  • #280
  • Jun 22, 2010 1:53pm Jun 22, 2010 1:53pm
Quoting stevegee58
Hey RR, after the EA has been running for a while you don't need the tick database anymore right? My impression was the tick database was for priming the pump when the EA first starts up.

Good question. I was just about to ask that. :-)
1 trader viewing now