Backtesting - Real Tick Data / Tross - page 6
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Hi, Tross
I testing with tick data Gain Capital.
But I don't know, if is able to believe record. Use only *fxt datathat the am generated, without election RECALCULATE.
Examples for EURUSD
1. downloads *.csv data
2. copy FXTHeader.mqh to \experts\include
3. copy simple_csv2fxt.mq4 to \experts\scripts
4. copy GainTicks2fxt.mq4 to \experts\scripts
5. restart Terminal
6. copy csv files to \expert\files and rename EURUSD_ticks.csv
7. run terminal without internet or without password...not loggeed!
7. open graph EURUSD , switch timeframe M1 and drag and drop script GainTicks2fxt.mq4 to graph.
8. after convert... look journal, *.fxt in dir files.
9. repeat any timeframe
10. copy *.fxt to tester\history and dont enable switch RECALCULATE
This is EURUSD my repaired data 2005..2006 GAIN CAPITAL
http://rapidshare.com/files/9633327/EUR_USD.zip.html
http://rapidshare.com/files/9635504/GBP_USD.zip.html
http://rapidshare.com/files/9643753/USD_CHF.zip.html
http://rapidshare.com/files/9645952/USD_JPY.zip.html"6. copy csv files to \expert\files and rename EURUSD_ticks.csv"
Did you really mean to copy the "*.csv" file to "\expert\files"?
There is no way one could open a chart from that file... if you do step 6 and try to odo step 7 you will have not data at all to open any chart, unless you have connected and downloaded data, but that would crap everything up.
In order to use the csv data I could import to my history, but this doesn't work too.
And else: "10. copy *.fxt to tester\history and dont enable switch RECALCULATE" Where would that be originally generated in to make us have to copy to another place?
Am I missing something?
i dont think this idea can ever work. MT can only ever use 1 min increments and throwing smaller times scales at it will just aggregate those ticks to 1 min bars. How do you run an expert based on its internal settings without pressing the recalc button? Even if it was possible i would suspect GAIN data is highly dubious at best being they are a broker that often trades way off market and stop hunts all the time so they have loads of spikes which are not real to market. So in effect your trying to find that last 10% accuracy on data which does not reflect an honest broker.
In my experince i ran a 4h system live all last year and compared the live trades to backtest. In about 1 in 20 trades a stop was hit by 1 pips when backtest missed but also fills were made when backtest missed yet the gain for the year and PF value was correct. Im seeing the same now on scalper system, PF is about the same live but not every trade is faithful to back test 1 in 20 will show stop hit when missed on backtest and profitable trades appear not seen on backtest.
In essence a well designed EA thats has been backtested to 90% on 1 min ticks is a very very close example to real life and if it shows a profit it will work live.
Bolt,
I really think MT developers and the brokers should seriously think about providing quality data. Even 1 min data that can be downloaded from MT is not true data that you can trade off. I have EAs that trade based on closed bars (ie, is not dependent of intrabar ticks) and have shown great profits with 90% modeling quality. However, they fall on their face when forward tested.
I wish the brokers sell quality data at a reasonable price. It is better to spend some money to get quality data rather than base your trading decisions and risk losing more money on unrealistic data (typically like that you can download from MT).
I agree, I have been forward testing over this week, the EA made about 1.5%. (over the same period) 90% backtest shows huge loss, 99% backtest shows huge win. A lot can happen in 1 minute. This is a scalping system, to me this indicates the downloaded 90% data is less volatile then the live feed and the tick data is more volatile. To me it seems the only surefire way to backtest systems like this is to use broker specific data, there seems to be only one broker who does this and they don't support metatrader.
The whole backtesting thing has far more holes in it than it should, they leave a huge doors open for you to delude yourself into thinking that an EA is profitable and the documentation is sadly lacking. As pointed out, the MT4 file structure does not really support tick data or ask/bid separation, what does this tell you? I say it's been done deliberately.
Bottom line, if they really wanted to make it good, they could. The limitatons seem thought out in advance to be subtle.
my experience is all on forward testing, but i think these few sentence might help, so i make a post. i have demo accounts with IBFX, North Finance, and Crown Forex. ofcourse everybody is talking about North Finance's is not clean ; where Crown's volume is sucks. That all i know, but if you're running a forward test with scalping EA which close trade and open trade oftenly, you can actually find out Crown forex's result did better then any other broker. Where North Finance at 2nd place, and IBFX is totally out of the picture. therefor, backtest for me is only a quick reference. in order to refine settings on EA, that will take a lot of time to run forward test, even years to prove the EA is profitable or not. 10point3 is currently running live with my own interbankfx, so far so good. withdraw twice a month, wipe out once a month. reset settings every withdrawal or refund account after wipe out. 10point3 is averagely 50% capital gain / month after deduction of withdrawal or refunding account. current stage is trying to work out money management that can act like myself to protect the capital from being traded and get the margin call. and because of itchy hand, has reconstructed another new MACD version of cost averaging system, and ready to participate on the comming world championship.
Regards,
David
Goblin_Options with my options...
Hi Tross, congratulations for this very useful thread!
Can you please backtest "Goblin_ options" with the exact settings as on the attached statement.
Basically I only changed as follows:
Lots= 0.01
TP= 10
Pips=10
MaxTrades=10
Thanks in advance
Hello,
If posible please backtest this EA. eurusd pair on 1m chart.
Hi Tross, congratulations for this very useful thread!
Can you please backtest "Goblin_ options" with the exact settings as on the attached statement.
Basically I only changed as follows:
Lots= 0.01
TP= 10
Pips=10
MaxTrades=10
Thanks in advanceYou bet, interesting approach, here it is
Hello, If posible please backtest this EA. eurusd pair on 1m chart.
Hey, can't seem to get this to open any trades.
Don't have a lot of time to debug it, did it backtest for you?
I couldn't get it to backtest on MIG or Velocity.
lemme know
Hey, can't seem to get this to open any trades.
Don't have a lot of time to debug it, did it backtest for you?
I couldn't get it to backtest on MIG or Velocity.
lemme knowHmm... i seem to get the same problem. Nevermind, please backtest this EA... its the same strategy but for gbpusd.
Thanks