Menu
Sister Sites

Options

Bookmark Thread

First Page First Unread Last Page Last Post

Print Thread

Similar Threads

meanVel's Trade Critique Corner 1 reply

critique my EA 18 replies

Critique my system please 12 replies

Trendline only system- Please critique 14 replies

Beginner requesting plan critique 5 replies

Attachments: Critique my backtesting results of this EA
Exit Attachments

Critique my backtesting results of this EA

  • Post #1
  • Quote
  • First Post: May 5, 2016 5:19pm May 5, 2016 5:19pm
Would anyone care to share their thoughts on this EA?


I've been using it live for 3 weeks and has returned 12% so far. Also been ironing out the bugs that caused it to lose one week (Human error to blame in one case)
Attached Image(s) (click to enlarge)
Click to Enlarge Name: Screen Shot 2016-05-05 at 22.12.22.png Size: 122 KB
Click to Enlarge Name: Screen Shot 2016-05-05 at 22.12.13.png Size: 86 KB
So many in the Commercial section want to make me rich! :o
Could you please tell us what the system does in order to give you a realistic opinion.
Old ways dont open new doors
Quoting FX-Flux
Would anyone care to share their thoughts on this EA? I've been using it live for 3 weeks and has returned 12% so far. Also been ironing out the bugs that caused it to lose one week (Human error to blame in one case) {image} {image}
test it with fix lot size and post the result.then we can talk.if ı am not wrong your live acount can margin call soon with the actual risk/trade
until then code seems good
Quoting FX-Flux
Would anyone care to share their thoughts on this EA? I've been using it live for 3 weeks and has returned 12% so far. Also been ironing out the bugs that caused it to lose one week (Human error to blame in one case) {image} {image}
Spread ? and yes, test it with fixed lots.
I will do this as soon as i get back it looks much smoother using fixed lot sizes, i think i had it risking 10% each trade for this test...

I just use current spread, as the TP and SL for this are 20-70pips the spread doesn't seems to effect it much.
So many in the Commercial section want to make me rich! :o
Here is fixed lot size of 0.1
Attached Image(s) (click to enlarge)
Click to Enlarge Name: Screen Shot 2016-05-07 at 19.52.27.png Size: 168 KB
Click to Enlarge Name: Screen Shot 2016-05-07 at 19.51.43.png Size: 84 KB
So many in the Commercial section want to make me rich! :o
Quoting FX-Flux
Here is fixed lot size of 0.1 {image} {image}
first test done with 4 symbol / total trades 2972
this test done with 5 symbol / total trades 3332

??
Sorry, i have recently added another symbol that the strategy works on. Additionally, this test includes the last couple weeks of data. here are the results with just 4 symbols
Attached Image(s) (click to enlarge)
Click to Enlarge Name: Screen Shot 2016-05-08 at 08.04.32.png Size: 124 KB
Click to Enlarge Name: Screen Shot 2016-05-08 at 08.06.56.png Size: 107 KB
So many in the Commercial section want to make me rich! :o
add algorithm specifications.timeframe too.

​​try run foward walks
on fixed Training and trading out of sample periods
cheers
15M time frame.

As i said above, i have been testing this for 3 weeks "forward test"

Quoting januszo
​​fixed Training and trading out of sample periods cheers
im not sure what this means?
So many in the Commercial section want to make me rich! :o
Nothing i have not seen thousands of times before. But lets see how you go with live test. 3 weeks is nothing, so will be good to see how this goes long term. See you in three months or so and good luck!
Quoting mrhigurashi
Nothing i have not seen thousands of times before.
What do you think makes a good EA?

Yeah i look forward to three months time, worth a gamble .
So many in the Commercial section want to make me rich! :o
  • Post #13
  • Quote
  • Edited 10:59am May 8, 2016 4:24am | Edited 10:59am
Quoting FX-Flux
15M time frame. As i said above, i have been testing this for 3 weeks "forward test" {quote} im not sure what this means?
try do few backtests in following manner :
[backtest is that is what you presented us - run on data that were used in optimisation]

assume training period length TRL =4,that is four months (one month or one year for example, you will decide looking on results)
assume tradining out of sample period length TDL=1, one month

Use foward walk software or do it by yourself - repeat 10 (or more) times sequence of steps defined below :
1.optimize your algo on 20151202 20160401 (that is TRL months optimisation period)
2.pick some best parameters,and then use it for simple test run : from 20160402 20160501 (that is TDL months starting one day after optimisation period and we call it foward walk period)
3.check how your algo performed on TRL and TDL -
save result (positive or negative) from TRL period to TRResult
from TDL period to TDResult

4.shift your TRL by TDL months back
Repeat - go to point 1 untill done.
After 10 or more steps sum up results from point 3 -
STR = sum(TRResult)/TRL ,
STD=sum(TDResult)/TDL,
divide STD/STR
if above 0.5 then go live with it.

PS
before going live - at least try 50 or more steps in the above algo
  •  Soros
  • Joined Sep 2012 | Status: Edge,Phsycology And Money Managemen | 949 Posts
looks like the backtests i use to get with million dollar pips long ago.

well to be honest, we cant give you an accurate backtest without having a basic idea of the logic behind the strategy, each logic has its Pro's and Cons, so when you backtest you need to have such in mind and try to establish what the worst that could happen with the system. basicaly backtesting is for this:

1. To get a Basic Idea of the system.
2. To Establish the systems breaking point if any.
3. To try and get the right settings for the system and see when in what market conditions system performs best.
4. To have a basic idea of maximum drawdown so you can correctly measure the risk acording to your risk tollerance.
5. To know if a system is potentialy profitable or not.

Well there are more points i gues but the points i put above are my personal points that i look for, just the basic. i usualy work on a system for quite some time before trusting it, i know that a system can easily turn on me, so i must know what im getting myself into and prepare myself mentaly.
I am what Many Dream to be but only a few can achieve, im a part of the 1%
  •  Soros
  • Joined Sep 2012 | Status: Edge,Phsycology And Money Managemen | 949 Posts
by just looking at your system it seems like it uses some breakout trailing stop algorithm, if so then i would be very cautious with such systems.
I am what Many Dream to be but only a few can achieve, im a part of the 1%
Quoting FX-Flux
{quote} What do you think makes a good EA? Yeah i look forward to three months time, worth a gamble .
Well, im yet to see any EA that works long term on a decent regulated broker. So, not sure i can answer that question. Like aliens or ghosts, not saying i don't believe they exists, just haven't seen one myself.
Soros - some questions arised - you say :


2. To Establish the systems breaking point if any.
- what is breaking ponit? - stop EA for sending orders when ie.: ADX is below 20?

3. To try and get the right settings for the system and see when in what market conditions system performs best.
- u assume marked conditions are detected within EA or simply put EA online or offline when you think so?

4. To have a basic idea of maximum drawdown so you can correctly measure the risk acording to your risk tollerance.
- i have system that shows 7% DD on backtest,
then i run it live and suffering 10%DD
turn it off or not ? )
Do you have any simple rules that answer above or similar questions?
Quoting Soros
by just looking at your system it seems like it uses some breakout trailing stop algorithm, if so then i would be very cautious with such systems.
You are correct, it is a very simple breakout system - no TS though.

As for backtesting and optimisation - it didn't really need much optimisation, as i had already whittled the strategy down manually while i was still practicing, the only parameters that needed optimising were the ones to try spot fakeOuts. When i optimised i used around 60% of the data available to optimise and the remaining data to prove.

I think the drawback here is that some of the pairs it works on are correlated so you find that it sometimes makes trades around the same time of day.

I do think the algo needs more work, but i was quite happy with these results so far.
So many in the Commercial section want to make me rich! :o
focus on backtests as from 2013 to present

do not include pre-2013 data on your BT as there is no tick data for those time periods
if you do 2000-2013 BT - almost all EAs win because no tick data is available
practising coitus interruptus
  • Post #20
  • Quote
  • Edited 6:13am May 8, 2016 6:02am | Edited 6:13am
Quoting avibe
focus on backtests as from 2013 to present do not include pre-2013 data on your BT as there is no tick data for those time periods if you do 2000-2013 BT - almost all EAs win because no tick data is available
There is a way around this in your coding and if you restrict your trading timeframe to say D1. For example you can do long range tests on D1 provided your entry and exit are based on the open of the bar immediately following the signal (provided that the signal is still in force on the open). That way you can confidently test the system without having to worry about tick data accuracy. There are 4 accurate control points to every bar...being the open, the close, the high and the low. The rest is algorithmically derived in MT4 when no tick data is available.

PS Dukascopy tick data is available on many forex pairs back to 2003 or so....so you can get free tick data for an extended test if you want to choose lower timeframes.

Attached Image (click to enlarge)
Click to Enlarge Name: Capture.JPG Size: 107 KB
0 traders viewing now