![]() |
|
||||
|
Belzebut,
I think the results thrown away are basically optimisations that failed to remain in profit. The lack of doco and/or logs and/or error msgs etc is a little disturbing. Obviously any backtester isn't gonna be a patch on forward testing. Interesting to do some forward testing and compare results on the backtester for the same period though. Metaquotes still have a ways to go with MT4, least of all the Client API. Cheers |
|
||||
|
I think the diffrence in backtesting results between MT3 and MT4 an live trading is not so hard to explain.
Live trading is the real thing. Every tick is really comming from the market. The strategy tester is using a summary of ticks. Know as a bar. If you run a backtest on a 1 day chart only OHLC are availble to you. While the real market generates 8640 ticks a day (1 tick/10 secs) We will never be able to know what the real market movement was within that bar/day. MT4 uses a principle known as fractals to interpolate the data. Read generate 8636 ticks. There is a theory that the market behaves on the smaller timeframes just like on the higher timeframes. Example 2 decade of data. There are 10 yearly bars up. Then 4 down. Then 6 up. Now zoom in on the first leg of 10 yearly bars up. In that leg you will find 60 monthly bars up. 24 monthly bars down and 36 monthly bars up. You can zoom in way down to tick level. 10pips up 4 pips down 6 pips up. If you start a fractal backtest on daily bars you sometimes see short message that it's loading data from a smaller timeframes. With minute bars you can make the extrapolation much more accurate because now you know the range from minute to minute. Unfortunately there is limited minute data availble. If you backtest 15 years of data you will notice that the further you go back in history the less accurate the backtest becomes. Just because 15 years back there are no lower timeframes availble. 3 month back you still have 4 hour timefrimes but not 1 minute so the interpolation is less accurate. If the fractal principle is coded well a reasoneble backtest can made. But no way near real tickdata. Backtesting is the hot issue on *all* trading platforms. testing a strategy by forward testing is (almost) impossible. IMO you need at least 100 trades to calculate statistical probabilities of your system. Many systems only generate 1 trade a day. Thats 100 days of testing. During testing easily 5 diffrent sets of parameters are used (often 50 diffrent sets) So 500 days of forward testing. And now consider that some systems only generate 1 trade a week. Such longterm systems can only be backtested manually. But that is close to slavelabor. So people tend to base the quality of their system on 5 trades. The ultimate backtester is very easy to program for the vendor of trading programs. Just store all ticks and on backtesting replay them. 100 years of tickdata takes only ~4GB storage per pair. Nothing by todays standards. Now we just need someone who writes such a tool :-) Just my 2 pips
__________________
If your system fails just swap long & short :whistling |
|
||||
|
Somewhere a MT employee said that besides of the 3 backtesting strategies in MT the user could also use for example a file that contains 1 minute data back to 1970.
Unfortunately his English is even worse than mine and he used both 1 minute bar and tickdata in one sentence. So it's not clear to me tickdata is also supported. But importing from file gives new possiblities. For example play a bit with tickdata and make it look like 1 minute bars. The big Q is now does anyone has any experience with importing from file? I can't find info on it.
__________________
If your system fails just swap long & short :whistling |
|
|||
|
Could someone tell me whether MetaTrade is able to accept historical intraday data for market replay and trading simulator with real-time P/L tracking function ? The data is in standard TradeStation format like this.
Date, time, open, high, low, close 10082008, 9:45, 10322, 10333, 9525, 9955 Can anyone please tell me ? many thanks in advance.... |
![]() |
«
Previous Thread
|
Next Thread
»
| Thread Tools | |
| Display Modes | |
|
|
Similar Threads
|
||||
| Thread | Thread Starter | Forum | Replies | Last Post |
| Important: ZeroCode Final Release | scorpion | MetaTrader and ZeroCode | 44 | 09-13-2006 12:59 AM |
| Getting spread in MetaTrader 4 | karan | MetaTrader and ZeroCode | 3 | 05-23-2006 05:31 AM |
| Neurex goes to metatrader 4 !!! | Belzebut | MetaTrader and ZeroCode | 1 | 07-07-2005 11:14 AM |
| Security Issues in MetaTrader 4 | scorpion | MetaTrader and ZeroCode | 3 | 01-27-2005 03:51 PM |
| Upcoming MetaTrader Version 4 | scorpion | MetaTrader and ZeroCode | 12 | 11-26-2004 01:30 AM |
All times are GMT. The time now is 12:42 PM.
Powered by vBulletin Version 3.6.2
Copyright ©2000 - 2008, Jelsoft Enterprises Ltd.
SEO by vBSEO 3.1.0
Copyright ©2000 - 2008, Jelsoft Enterprises Ltd.
SEO by vBSEO 3.1.0
Registered members have access to special online forex currency trading tools, software, mt4 expert advisors and indicators. Register now

















Linear Mode


