Could some friend help me?
-
Good evening, friends, at least here in Brazil is 22 hours.
I need to mount an EA that opens for example for each purchase order that it makes, open together a pending sales order with the same value as the purchase order.
Let's say, win with the high, but protect the capital by opening the sales order with say 40 pipis of stop gain and if after these 40 pipis of gain continue falling, after closing the sales order with gain open a new sales order after 10 pipis.
I tried to swear, but I honestly could not! Any good souls to help? -
this strategy is not profitable, was tested many times intensively in the tester by many experienced guys (even by me), and yeah, is not profitable
-
But my God, do we all think the same way? Guys, does anyone have something new to share? Let's help ... Where are the guys from NASA?
Look at this design I made, do the back test in eurusd 15minutes for 1 month ... It's too beautiful
-
What is this history from the market telling you about the future?
People are looking for many different paterns as profitable holy grail, but that never worked - acutally it worked only in tester for some period of time, but not for the future.
You can trade probability and create tools for that. Following this member can help you: https://www.forexfactory.com/alphaomegaBetween trading and gambling is a very small gap, be careful.
-
In my view:
rule number #1: avoid overfitting
rule number #2: backtesting results from "in sample" data means absolutely nothing. Out-of-sample is the only thing that matters.
rule number #3: never underestimate rules #1 and #2. -
What do you do to avoid over-fitting?
-
Validate a trading system across multiple independent segments of data that were not available during optimization.
Optimize 2010-2011 (past, in sample)
Test 2011-2012 ("future", out os sample OOS #1)Optimize again 2011-2012 (that became past, in sample)
Test 2012-2013 (out of sample #2).... repeat until all data is used, then calculate the combined performance of OOS segments. This is called walk forward analysis and simulates how well a strategy deals with changing markets. If not done carefully, you contaminate OOS segments by using them for optimisation before walk forward analysis. I see many people don't understand the concept and end up curve fitting the walk forward analysis as well.... so, it's important to do the homework (google Robert Pardo)
-
"Thanks GOD of trading to let @miro1360 be with us"
WFA = Walk-Forward Analysis
-
It troubles me when someone suggests a certain strategy doesn't or worse can't work. With all the tools and techniques available with fxDreema and Metatrader pretty much any strategy can be made to work.
If it doesn't work right now you haven't found its flaws and adjusted for them yet.
There may be cleaner or easier strategies or methods. But it sure feels to me like you could make a pig fly with these tools with a little effort.
No magic. Just patterns, analysis, some funny colored ovals on the screen and optimization.
My warped view of the world ...
-
Good morning, my dear friend, imagine that the FXdreema is a great boat, and we are all trying to row to the same side.
In this case, I think you could get me a doubt.
What would rule 1 overfitting? Could you explain more?
And what would rule 2 "in sample"?
I am very lay and I would like to learn more with you ... -
Thank you dear friend, I'm going to study and learn more.
My aggravation is my English a little bit limited ...
Big hug -
Overfitting is the process of adjusting a model (e.g. a trading strategy) to work so perfectly in test data (e.g. the past) that it only "fits" that piece of data and absolutely nothing else. This is, in my view, the most common mistake in trading system development. So what happens in most cases is the trader has an excellent backtest, but the performance collapses immediately after starting on live trading.
Common question is: "how can we know the future then?".... obviously you can't, but what you can do is to break the past in multiple segments, develop and optimize a system in older segments, then pick the systems variations you judged "best" in that old segment and test it in a new, forward segment.... in that manner you can better estimate (never guarantee) whether your model has predictive power into new data, since this is the only thing a trader is looking for (a pattern in the past that is repeated in the future).
So, back to your other question: In sample is the "past", and out of sample is the "future", in relative terms. ..... 2012 is the future of 2011... 2013 is the future of 2012,.. 2014 is the future of 2013.... you got the idea.
So you have 2010 as in sample, optimise your strategy there and run the best option in 2011 (out of sample). Register the 2011 results alone... then 2011 becomes the past. So now it's in sample and you can optimise with 2011 data. So you do that and run the best option alone in 2012.... then repeat until you reach yesterday's trading session. If the combined performance of OOS segments is good, you have a system that has a good chance of adapting to next year's market, and continue to do so, until it doesn't (so you turn it off)... Again, nothing is guaranteed in trading, we're just talking about statistics and probability, on a context that is always changing. So you have to understand the risks and be aware that the chance of success is tiny tiny.
-
I understood, so I figured an EA would open 2 orders regardless of the bullish or bullish trend.
For each purchase, sale pending.
Or blocks with trend reading where you take a reading of 1 month for trend where defining this trend, would trigger the respective block to the scenario until this change takes effect.
Because the market comes down to this, regardless of the market, it is made of tops and bottoms, supports and resistances ... and finally cycles of trends ...Meanwhile I still dream of being Bobby Axelrod in (Bilions) in the Netflix series
