EA testing

[Ken Long]

I installed the Demark this past weekend with no problem. I simply clicked the link, chose save, and saved it in my experts folder. No problem. Installing the indicator was a little more difficult as I couldnt locate the folder right off. I ended up clicking run, which brought up the Meta editor. Then I created a new indicator file, named it and cut and pasted the source code in. Everything is working perfectly.

These are installed on my personal trading computer. I dont like to run it 24 hrs a day. I turn it on in the morning and use them on a discresionary basis. In just the past two days I've found the Demark to be a fine directional robot. I choose the direction and time frame, and modify the s/l and t/p settings to fit, and set it to work.
When it hits I adjust the s/l and t/p settings to fit the chart, and trail my own stops. The trailing stop feature on MT doesnt seem to trail a stop loss, only a profit stop, and for some reason the t/s feature on the Demark doesnt seem to work right. The system automaticaly trails the s/l regardless of how you set the t/s feature.

This seems to be a great robot for directional tend trading.

[Batchboy]

Since this thread started out as "EA testing", I choose it to ask my question!
I can get lots of demo back-1Min-data but hardly anything at all for live back-1Min-data!!!!
This means my development of Invisible_FX_Landscape is screwed!!! The live data does seem somewhat different than demo, so switching MT4 from demo to live account only gets a couple days of back-1Min-data. Help! I've successfully optmized and back-tested this EA of mine for 3 months data, but now for live???? No data! What to do!?
jerry

[PipCop]

Originally Posted by Batchboy

The live data does seem somewhat different than demo, so switching MT4 from demo to live account only gets a couple days of back-1Min-data. Help! I've successfully optmized and back-tested this EA of mine for 3 months data, but now for live???? No data! What to do!?
jerry

This is why I am so skeptical one back tests. One minute data, at least what I have found, means it has 4 pieces of information:

Open
High
Low
Close

What it doesn't have is every single tick . Therefore, when you back test a robot, you are testing against OHLC data, not tick -by-tick data. The result is that most robots are designed against bad data, and it's no surprise they don't work live.

The only solution is to test live or find tick -by-tick data and backtest against that. However, as I understand it, MetaTrader doesn't handle tick -by-tick data very well for back testing, so you'd need something like Forex Tester.

[Batchboy]

yes I know, that is why I had designed new EA to commit to logic decisions from static C(-1) bar after it had just formed! You'll see that in my new EA thread and also about general EA designing thread. So of course, I'm well aware, but this is another step to the problem, live 1Min being different than demo side, and live isn't storing but couple of days, it's a horrible situation, so I may have another large step of finding an EA that stores the live data (not talking about tick by ticjk problem, that is already solved by taking care to use on end-of-bar close static close values for decsion and also to do the SL in dynamic manner) or of having one created. 1ka found something from mql4 site that is a conversion chart EA from 2005, so I'll have to see how this new arm of project will go.

Even after I found or have a data storer EA, then I have to just keep it running to store the data for a few months before I can do the optimization again! Oh well, at least I found that if you go by the strict standard of EA design to avoid the tick -by-tick problem you can bypass demo testing and go right into a $100 live test, why? Because I found the live demo running of an EA under such conformaties will do as back-tests do, thus the jump past demo-ing to a micro-live$$ is possible, so minus one step, add back-in another soon as I have a storer of waiting for data to store for 3 months!

jb

[Batchboy]

So I wrote both Yaroslav of forex-experts.com and Witold of mqlsoft.com to get a quoate for an EA that can attach to a 1Min chart to a bar-data-farm.

I Notice posted progress on this pipcop forum has been a fence post for me of where the dirt road hidden below all the over-grown shrubery is, it's been quite an adventure. The end of the dirt road out in the mountains may go over a cliff someday or may yield a mountain top frankly, because a wonderful EA (apart from false-god-like EA's like yoww) are really only and ultimately optimum dice throwing game, in example of this EA I finished with 1ka, it's dice throw of 40%-wins/60%-lose, and (on 1 mini) gainers-$16, losers -$6, which "on average" means $475 in 3 months and 9% drawdowns......over last 3 months during midnite to 1am (PacTime), but no doubt the optimization run on live data storage will come out different! It already has different parameters and zoomed-in hours to operate on a std-demo. I'm running both std-demo & mini-demo live now in real time now.

Depending on the market, the results in 3 month will no doubt vary, but if looked at as an average projection.....you get the idea.....paint blue dot on 4 sides of 6 of the dice and place an equal bet on each dice roll, would you play the game? No matter how scientific, the bottom-line scientifically is the game..... roll-the-dice-please-for-the-next-toss! You'll have times of spuriouslly having a series of losers or series of winners ........and times where it goes back 'n forth. Up to actually doing these tests on a micro$$ account, it shall be all scientific if you want those performance figures to be "hammered into the rock face" before live$$ run forward. Does someone else want to convince me any method or EA would do so cleverly otherwise, cleverly & scientifically too?

Albeit DTS-1 is clever idea that accumulates wins & loses and waits for a teeter-toter of either camp side to win and closes all the small trades, clever in opposition of martingales and grid technoiques where you never know when or if much teades will close!
jb

[Batchboy]

Originally Posted by PipCop

Thanks for the kind words. I am glad the site offers some hope, as I hate to see so many folks lose so much money!

My many thank-yous too sir pipcop!
Have a suggestion to consider:
Just below this line on front-page:
PipCop - Real Forex robot reviews on real accounts!


Put a score card!
E.G.: Since date XX/XX/XX:
15 demos running
15 failed demo run EAs
3 Live$$ accts running
7 live$-accts failed
0 live$-accts continue past 3 months thus far re: ANY gains greater than loses

You see, us public can microscopically get into Robot reviews w/ no macroscopic view from get-go as to what is going on and what we're up against! This in bold Black Italic letters would be the score that would appropriately cut into any early hope/greed factors we all have to some degree.
Going totally scientific,
Jerry

[PipCop]

I like the scorecard idea. I'll have to see how I can go about doing that.

[Batchboy]

Originally Posted by PipCop

I like the scorecard idea. I'll have to see how I can go about doing that.

Yet another idea chief to put just above the scorecard:
Some kind of motto pertaining to the ongoing nature of the results thus far, like "If you have an EA that works 6 months or more live$$ operated in our servers here (!), then you win $10,000**"

**(paid over a period of time from members here)

So since not sure that can be done, would you believe:

"Nobody's EA is really making it long-term, will yours?"

I know you have some kind of motto below "The Pipcop", but that was thought up when you originated site, it's more like a purpose statement, I believe, not a motto (?), and definitely not after seeing where this was really going with any scorecard(s) at hand, continuing:

"Where we dare you to expose your EA LIVE!! Here! Now!"

Have others submit their motto ideas in this thread and maybe do a poll to arrive at best one, ...........more:

"So far EA's are Toys! All failing within a few months on live accounts to date. Is your EA man enough?"

OK, had my creative cup-of-tea for the dayl, enjoy you all!

jb

[Ken Long]

Originally Posted by Batchboy

"
"So far EA's are Toys! All failing within a few months on live accounts to date.



jb

Toys certainly, but semi inteligent toys. I have always had my doubts about trusting any purely mechanical trading system. However many of these EA's are enormously capable, just not completely failproof.

Aside from inteligent design on part of the creator, I think what is missing most often is inteligent use on part of the operator. I have yet to meet any simple mechanical system that can adapt to a changing environment and manage a position through those changes, that job is ours.
Perhaps there are more complicated algorithms with adaptive technologies, but that idea is most likely beyond our pockets and this site. However, even some inexpensive EA's can be used well by a discretionary trader who knows how to manage a position.

Mechanical trading systems are an aid, not a grail, or a trading end all. Most are due to fail at one time or anouther.

[Batchboy]

Originally Posted by Ken Long

Mechanical trading systems are an aid, not a grail, or a trading end all. Most are due to fail at one time or anouther.

That's why I wanted to come up with my own EA (see thread about, "Invisible_fx_Landscape") that covers all the bases, and in case of trader's manual interjection I put trade-time windows, optimize, 1 thru 23 and 1 thru 4 hours in time-length......search for. A kind of "Land" in between total-auto and interjected-manually. And it may work, no guarantees of course, even then I'll be optimizing every weekend (interjected manually again!) to "be on top of it" (i.e., market changes). Just have to see
jb