How reliable is backtests?
forex_trader_139412
Member Since Jul 16, 2013
385 posts
Sep 22, 2014 at 19:31
How reliable is the MT4's strategy tester with 90% modelling quality?
Member Since Nov 21, 2011
1718 posts
Sep 22, 2014 at 22:46
It depends on what type of strategy you are backtesting.
- Scalping strategy is 100% useless
- Scalping strategy is 100% useless
forex_trader_139412
Member Since Jul 16, 2013
385 posts
Sep 23, 2014 at 06:33
Not a scalper..
forex_trader_202879
Member Since Aug 07, 2014
406 posts
Sep 23, 2014 at 07:02
CrazyTrader posted:
It depends on what type of strategy you are backtesting.
- Scalping strategy is 100% useless
Hello. Why would you say scalping strategies and backtesting is 100% useless? Such open ended statements reveals your bias towards BIAS TRADING which is investing. Scalpers make money on both sides of the coin, as investors only look to FOLLOW the trend. We see how well that worked from 1.39xx to 1.30xx on EU huh..
Sep 23, 2014 at 07:08
It's very unreliable. MT makes up the data as it goes along. You'll get a different result from running a test while online and while offline. You'll get a different result on a different processor.
It's got no use other than to see if your code is working.
It's got no use other than to see if your code is working.
forex_trader_139412
Member Since Jul 16, 2013
385 posts
Sep 23, 2014 at 07:32
Is there any way of doing a reliable backtest then to see how your EA will perform before going live?
Sep 23, 2014 at 10:03
Just test it. You can run as many instances of MT as you want to. I usually have about 10 going. If you have an idea code it, run it. See what happens.
Your alternative is to ditch MT and write your own systems in JAVA or C++ and go test it on Oanda's data...
Can also try something ike this: https://pepperstone.com/trading-platforms/ctrader-calgo.php
I don't know how accurate they are, but quite frankly anything will be more accurate than MT.
Your alternative is to ditch MT and write your own systems in JAVA or C++ and go test it on Oanda's data...
Can also try something ike this: https://pepperstone.com/trading-platforms/ctrader-calgo.php
I don't know how accurate they are, but quite frankly anything will be more accurate than MT.
Member Since Jun 03, 2010
696 posts
Sep 23, 2014 at 11:17
RSTrading posted:
How reliable is the MT4's strategy tester with 90% modelling quality?
Not reliable even at 1% ;)
PAMM MANAGER // Professional Fund Manager
forex_trader_139412
Member Since Jul 16, 2013
385 posts
Sep 23, 2014 at 11:49
I suppose you meant 'not even at 100%'?
So does that extra 9% make a huge difference. Should one rather go that route?
I know backtesting doesn't factor in stuff like spread and commission, but lets say you set the spread to a ridiculous figure and know what your commission is per trade, will that do?
What do you guys then do to get accurate backtesting?
So does that extra 9% make a huge difference. Should one rather go that route?
I know backtesting doesn't factor in stuff like spread and commission, but lets say you set the spread to a ridiculous figure and know what your commission is per trade, will that do?
What do you guys then do to get accurate backtesting?
Member Since Nov 21, 2011
1718 posts
Sep 23, 2014 at 12:52
Many people think backtesting is for testing past. No it doesn't. Backtesting is designed to check if EA you coded run as expected. If not, then it means you have to modify logic code until you get what you want. That's is.
forex_trader_139412
Member Since Jul 16, 2013
385 posts
Sep 23, 2014 at 13:06
(edited Sep 23, 2014 at 13:35)
oh ok.. Then I have been under the wrong impression for years.. So backtesting will not give you past performance? You will only be able to see if your coding was done right.
forex_trader_139412
Member Since Jul 16, 2013
385 posts
Sep 23, 2014 at 13:25
Is there a way then to see how your (already established 100% working and doing what you want) EA performs on historical info?
Sep 23, 2014 at 13:42
Sep 23, 2014 at 13:49
Probabaly already made you throw some very good systems in the dustbin and made you try losing ones.
Look do the test, take your EA run it for a week, then back test the same week. Compare the results. If you can run it on a second broker.
You'll end up with three sets of results.
You just can't build a future on that. And you're wasting your time trying.
My advice to you. Test live, you see very quickly if you like the system or not.
Look do the test, take your EA run it for a week, then back test the same week. Compare the results. If you can run it on a second broker.
You'll end up with three sets of results.
You just can't build a future on that. And you're wasting your time trying.
My advice to you. Test live, you see very quickly if you like the system or not.
forex_trader_139412
Member Since Jul 16, 2013
385 posts
Sep 23, 2014 at 18:29
Thanks, will try that, and will take your advice..
Sep 23, 2014 at 20:21
Look even on live tests, I've got one on O running two accounts same code, just different trade sizes.
One has 11 trades open the other 20. The difference ? The spread. I have code watching what's going on spread wise and it will sit on it's hands if it doesn't like the spread. We're talking milliseconds here between quires.
You're not going to get an accurate results from a free piece of software that makes up data is it goes along. Next back test you run look at the bottom right corner. See the data coming through. Then run it again, it will pull data again. So what it pull through on the previous test that it doesn't have now ?
This is before spread problems, server problems and so forth that just doesn't feature in the backtests. The differences between running live and back testing will be anywhere from 20%, 50% up...
One has 11 trades open the other 20. The difference ? The spread. I have code watching what's going on spread wise and it will sit on it's hands if it doesn't like the spread. We're talking milliseconds here between quires.
You're not going to get an accurate results from a free piece of software that makes up data is it goes along. Next back test you run look at the bottom right corner. See the data coming through. Then run it again, it will pull data again. So what it pull through on the previous test that it doesn't have now ?
This is before spread problems, server problems and so forth that just doesn't feature in the backtests. The differences between running live and back testing will be anywhere from 20%, 50% up...
