Strategies work awesome in SQ, but not in MT4
30 replies
FXHelper
10 years ago #115392
I am having a hell of a time getting these beautiful strategies to translate into anything meaningful in MT4.
I am typically lucky to get 50% of the performance, and like I said thats on a good day, under the same conditions and in SQ I add a pip or more to slippage to make extra room for error.
I am not scalping, right now I am working on the D1 timeframe.
Any suggestions will be greatly appreciated
0
FXHelper
10 years ago #138509
Is a spread of 4 realistic for gbpjpy? I think its 7-15 if you calculate with commission. For the backtests I tend to use a high spread and slippage.
I use GlobalPrime, and the GJ spread is 3 pips… 7-15 thats a little nuts…
0
FXHelper
10 years ago #138510
i have almost same results in sq3; problem is in mt4 somewhere
if the problem is in mt4 backtest, will the same problem happen when your demo/live trading?
I mean this could render the whole exercise useless
0
clonex / Ivan Hudec
10 years ago #138511
data in mt4 are from ducas?
0
FXHelper
10 years ago #138513
data in mt4 are from ducas?
no its the brokers own data… I have not figured out yet how to load dukascopy data into my MT4
0
pknoell
10 years ago #138514
I use GlobalPrime, and the GJ spread is 3 pips… 7-15 thats a little nuts…
The spread is 3 pips + commission. I would test it with a spread of 10-15. If it is profitable with that spread it should work in real life. I test hard.
0
Threshold
10 years ago #138515
That is pretty sickening news,
ok, So i have to delete all data, and redownload every symbol, making sure that its weekends data not included… to start, ( i guess i have to delete all the strategies, as they are all worthless…
sucks, but better now than with real money.
Thanks
Thats my guess atleast. I dont’ think you’ll have to redownload them unless you deleted it all. You can just open and see your previous settings if you had weekends included or removed…
Finding good data and the process of preparing and handling it is extremely tedious and time consuming and above all else the most important.
0
mikeyc
10 years ago #138516
I use GlobalPrime, and the GJ spread is 3 pips… 7-15 thats a little nuts…
Global Prime feed is based on New York trading hours (Eastern Standard Time + 7 hrs to bring 5pm close in line with midnight on MT4), with US daylight saving time applied.
Dukascopy is a GMT feed with no daylight saving time applied.
There’s your answer.
If you are serious about developing strategies you need to use something like Birt’s Tick Data Suite to manage your data in MT4. You also need a broker that matches Dukascopy or you need to shift and apply DST to the dukascopy data to match Global Prime.
This all takes a lot of time and effort to get right, but without doing this any real trading will be worthless.
0
mabi
10 years ago #138518
@FXHelper
Since you are working with D1 time frame the problem is most liklely that you have trades that opens and closes on the same bar / day. In order to be able to check this you would need to use tickdata to be able to look inside the bars ( weather the trade reached the profit or loss first ) This goes for all strategies that have trades that open and closes within the same bar in what ever timeframe you are working on. I have found many “holyshit” holy grail strategies that turned out to have this problem.
0
FXHelper
10 years ago #138521
@FXHelper
Since you are working with D1 time frame the problem is most liklely that you have trades that opens and closes on the same bar / day. In order to be able to check this you would need to use tickdata to be able to look inside the bars ( weather the trade reached the profit or loss first ) This goes for all strategies that have trades that open and closes within the same bar in what ever timeframe you are working on. I have found many “holyshit” holy grail strategies that turned out to have this problem.
I never use open bar… they are always generated on 1 min bars…
the issue I was/am having, and I really wish someone would actually read the issue is, when doing the MT4 backtest, some trades would open exactly like they would in SQ, and others ( every other one) would open and close right away…
I dont know how thats a data issue…
i really think its MT4, which creates the question of, will the MT4 action on backtest is a foreshadowing of how it will act on demo tests.. ( since the trades dont trade every few minutes, its hard to sit and wait 3 months to get the answer)
0
Threshold
10 years ago #138524
I downloaded your strategy and tested it and am getting the same problem.
I think the strategy’s exit condition is being met on the same day the orders are opening in many instances so they are getting closed immediately for the cost of spread (a loss) whereas SQ seems to wait to close them the next day.
0
FXHelper
10 years ago #138525
I downloaded your strategy and tested it and am getting the same problem.
I think the strategy’s exit condition is being met on the same day the orders are opening so they are getting closed immediately for the cost of spread (a loss) whereas SQ seems to wait to close them the next day.
That could be a HUGE flaw in SQ then…
I noticed that its closing within 10 minutes of opening on MT4…
0
Threshold
10 years ago #138526
Always backtest and double check them on both platforms, there are definitely bugs. Maybe Mark or Tomas can provide you a code snippet to edit it if that is the case.
0
AC1962
10 years ago #138528
Hi guys
I initially had very similar problems with differences between SQ and MT4 backtest, but was able to resolve this after obtaining and ‘strictly’ following the MT4 “99% Quality Backtesting Guide”, freely available from AutotradingAcademy.com, if you attend one of their Webinars (only 45-60min long and quite interesting if you don’t mind the sales pitch). Though you have to use Birt’s Tick Data Suite for MT4 (costing $97 + $10/mth) to obtain the 99% backtest modelling accuracy. Nowadays, I find that using free Dukascopy tick data as both SQ ‘Real Tick’ data and MT4-TDS ‘Every Tick’ data, the results are virtually identical: except ‘Draw Down’ results which MT4-TDS generally yields more conservative results than SQ. Very rarely, but sometimes, SQ and MT4-TDS backtest results may still differ significantly, but this is probably a sign of an overly curve-fit strategy in SQ, which probably needs to be dumped anyway.
Important: make sure that downloaded Dukascopy tick data is set to your broker’s server time, + DST if applicable.
I hope the above is of help.
AC1962
0
FXHelper
10 years ago #138529
Hi guysI initially had very similar problems with differences between SQ and MT4 backtest, but was able to resolve this after obtaining and ‘strictly’ following the MT4 “99% Quality Backtesting Guide”, freely available from AutotradingAcademy.com, if you attend one of their Webinars (only 45-60min long and quite interesting if you don’t mind the sales pitch). Though you have to use Birt’s Tick Data Suite for MT4 (costing $97 + $10/mth) to obtain the 99% backtest modelling accuracy. Nowadays I find that using free Dukascopy tick data as both SQ ‘Real Tick’ data and Tick Data Suite for MT4 ‘Every Tick’ data, the results are virtually identical: except ‘Draw Down’ results which Tick Data Suite for MT4 generally yields more conservative results than SQ. Very rarely, but sometimes, SQ and Tick Data Suite for MT4 backtest results may still differ significantly, but this is probably a sign of an overly curve-fit strategy in SQ, which probably needs to be dumped anyway.Most important: make sure that downloaded Dukascopy tick data is set to your broker’s server time, + DST if applicable.
I hope the above is of help.
AC1962
I agree, just bought birts tick data… which makes sense as you should be testing it on the same data..
none of my strategies are time sensitive, what I did find is that the MT4, keeps opening and closing everyother trade instantly..
I personally am starting to think that its a MT4 backtesting issue, but who knows..
0
Rimantas Petrauskas
10 years ago #139642
Hi Quants,
glad to see so much interest in 99% quality backtesting.
We have created The Ultimate MT4 Backtesting Guide for those who want to learn how to backtest on MT4 and reach 99% modelling quality with a variable historical spread.
It’s absolutely free and explains how to use the newly released Tick Data Suite 2, which now is even more powerful than any other MT4 backtesting engine.
0