Trend system EA needs back testing | Page 2 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Trend System - Now Testing 15 replies

Doji Strategy Needs an EA for thorough Testing 9 replies

Strategy Testing - Testing 2EAs simultaneously 5 replies

Back testing my system. @ connect FXCM 5 replies

With backtesting, who needs forward testing? 15 replies

Trend system EA needs back testing

  • #21
  • Apr 19, 2013 7:37pm Apr 19, 2013 7:37pm
Quoting cja
I agree with FerruFx backtesting is only any good to test that the EA functions as it should, I am currently forward testing an EA on a Demo account that makes a good solid profit yet if I run it on the BT over exactly the same time period it comes up with a totally different result, part of it I suspect is because the EA has some MTF code and the BT does not handle MTF very well. When I forward test the EA on a Live account once I am satisfied with the basic set up it will probably produce very different results yet again. A forward test on a Demo...
Exactly
MT4/MT5 EAs/Indicators/Alerts coder
  • #22
  • Apr 19, 2013 8:40pm Apr 19, 2013 8:40pm
Quoting FerruFx
This was not the point. The point was that a BT can't tell you what real time trading will give you when you will go LIVE. Past results are not a guarantee for live results ...
When you do your BT with 99% tick data you will be able to use/set fixed spread.

No one is saying FT result will be as good as the BT, what I am saying is that “if the BT result is good then there is 99% chance that FT result will be good”. Ok?

You can't have a bad BT result and get good result in FT.

Regards
No Risk No Reward
  • #23
  • Apr 19, 2013 8:45pm Apr 19, 2013 8:45pm
Quoting fmfx1
what I am saying is that “if the BT result is good then there is 99% chance that FT result will be good”. Ok?
False.

Quoting fmfx1
You can't have a bad BT result and get good result in FT.
False.
MT4/MT5 EAs/Indicators/Alerts coder
  • #24
  • Apr 20, 2013 2:54am Apr 20, 2013 2:54am
Quoting FerruFx
Let me give you the reason why a backtest on week open and on week-end will give you different results: Backtester uses the current platform spread. That means if the spread was 20 pips at friday close, all your backtests on week-end will be done with a 20 pips spread. And on market open, if you run a backtest around news release and the spread is 50 pips ... Do you get the picture ?
I get the picture, but if you disconnect your platform from your Broker the Spread doesn't change. Do you understand ? You can also use tools to set the spread for your testing.

So all my Strategy Test runs can be performed under the same conditions regardless of the day or time.

Do you know how to disconnect your Terminal from your Broker ? there are several ways, I do it using an incorrect logon ID, just make one up with a few random numbers, the logon fails and your Terminal is disconnected.
20 pips a day isn't too much to ask . . .
  • #25
  • Apr 20, 2013 2:56am Apr 20, 2013 2:56am
Quoting FerruFx
This was not the point. The point was that a BT can't tell you what real time trading will give you when you will go LIVE. Past results are not a guarantee for live results ...
So a forward test is no help either, when it is complete all the data use is from the past. The thing that it does give is a real simulation of delays when opening, modifying and closing orders, slippage, news, etc.
20 pips a day isn't too much to ask . . .
  • #26
  • Apr 20, 2013 3:48am Apr 20, 2013 3:48am
Quoting RaptorUK
I get the picture, but if you disconnect your platform from your Broker the Spread doesn't change. Do you understand ? You can also use tools to set the spread for your testing. So all my Strategy Test runs can be performed under the same conditions regardless of the day or time. Do you know how to disconnect your Terminal from your Broker ? there are several ways, I do it using an incorrect logon ID, just make one up with a few random numbers, the logon fails and your Terminal is disconnected.
Yes of course, but the tester will use the last known spread. If your last connection was ended with a spread of 20, the tester will use it for all your backtests.
MT4/MT5 EAs/Indicators/Alerts coder
  • #27
  • Apr 20, 2013 3:51am Apr 20, 2013 3:51am
Quoting RaptorUK
The thing that it does give is a real simulation of delays when opening, modifying and closing orders, slippage, news, etc.
This is exactly the point. Real trading and backtest can't be compared.
MT4/MT5 EAs/Indicators/Alerts coder
  • #28
  • Apr 20, 2013 3:56am Apr 20, 2013 3:56am
Quoting FerruFx
False. False.
Why false? You can do better than that.

Regards
No Risk No Reward
  • #29
  • Apr 20, 2013 4:12am Apr 20, 2013 4:12am
Quoting fmfx1
Why false? You can do better than that. Regards
I already answered in an earlier post.

Good backtest isn't a guarantee of good live trading.

Bad backtest isn't a guarantee of bad live trading.

This was discussed many times in many threads ...
MT4/MT5 EAs/Indicators/Alerts coder
  • #30
  • Apr 20, 2013 6:23am Apr 20, 2013 6:23am
Quoting FerruFx
What I meant is that a backtesting result can't tell you what you can expect in the future. Forward trading is what you really get.
What I interpret the point that RaptorUK is making is that the a "forward test" "can't tell you either what you can expect in the future.
Say you create a strategy... and when you backtest the "first time", (say you run a backtest of 2 months) that is exactly as running the strategy in a forward demo for 2 months.. you get what you get.

Now, if you curve fit that strategy for profits with those 2 past months .. and you run that optimized set for the next 2 month... that is also as running
2 months of forward test. You get what you get.

So, "nothing" that you run in back test guaranteed same results in the future... and nothing you run in a forward test guarantee results in the future.
The only thing both gives is an "idea" of what you can get out.

Obviously I am talking about "well executed" back test... with tick data and really knowing what one is doing.
The problem is most traders do not know nor understand all the 1000 details of performing a good back test.

J.
  • #31
  • Apr 20, 2013 6:44am Apr 20, 2013 6:44am
Quoting jeuro
What I interpret the point that RaptorUK is making is that the a "forward test" "can't tell you either what you can expect in the future.
I was not saying that a forward test can or can't tell what to expect in the future.

I was telling that a backtest can't tell what to expect in real live trading.
MT4/MT5 EAs/Indicators/Alerts coder
  • #32
  • Apr 20, 2013 4:27pm Apr 20, 2013 4:27pm
Quoting FerruFx
I was not saying that a forward test can or can't tell what to expect in the future. I was telling that a backtest can't tell what to expect in real live trading.
Well .. yea.. I can agree to a point..but in general we know certain "expectations" that occur in real money account, and
we have control to factor them IN to the backtest.

Is all about having the "proper expectation" according to the strategy we are testing.
Let say I am testing an scalper strategy ( I don't use them but lets assume that) . We know in a real money account
we will get slippage or re-quote some 50% of the time.. say 3 pips average and maybe 5 in the Asian session.. then I
factor that in and perform the back test.

In other world I took out the "predictable" factor that happens in real money account and I isolated the strategy itself
testing the logic for triggering and so forth... Now, I have a very decent idea of how the strategy would behave
in real money account and could make a decision to trade it. ...

In "my opinion" , I do not call "forward testing" when done in a real money account. I call it "trading it". The only forward
I do in demo is to test the tech things of the code... if executing property etc. Not the logic.. the logic I already know
from the back test.

J.
  • #33
  • Last Post: Apr 20, 2013 7:06pm Apr 20, 2013 7:06pm
Quoting jeuro
In "my opinion" , I do not call "forward testing" when done in a real money account. I call it "trading it". The only forward I do in demo is to test the tech things of the code... if executing property etc. Not the logic.. the logic I already know from the back test. J.
I fully agree with this.
MT4/MT5 EAs/Indicators/Alerts coder
0 traders viewing now