Data for backtesting - page 3

New comment
 
Fernando Carreiro #:
A suitable value might be the average, or the average plus/minus one standard deviation to compare the two to evaluate robustness.

hmm yes maybe a mean +1 or 2 sigma. That's a good idea.


It's a shame that's not configurable in tickstory or the broker. Short of writing my own code to run through the tick data and calculate it for 1 min bar to bar I could get broader stats over the whole period which would give an indication. Then I could run it with a fixed spread.

 
EdFuk #:

As can be seen there is a large variation and the tickstory data is significantly different to the Darwinex "real" spread (Grey vs red lines)

This is probably not the best choice for researching your trading strategies.

 
fxsaber #:

This is probably not the best choice for researching your trading strategies.

OK thanks, but it’d be conservative, right? If it worked with Darwinex it’d probably work with others?

 
EdFuk #: OK thanks, but it’d be conservative, right? If it worked with Darwinex it’d probably work with others?

Yes, maybe, no!

Forex is not a centralised exchange. Brokers have different liquidity providers and data feeds.

So, what works for one broker may not work for another, independent of the spread.

It depends on how your strategy works and how resilient and adaptable it can be to those differences.

 
Fernando Carreiro #:

Yes, maybe, no!

Forex is not a centralised exchange. Brokers have different liquidity providers and data feeds.

So, what works for one broker may not work for another, independent of the spread.

It depends on how your strategy works and how resilient and adaptable it can be to those differences.

Sure, I get this. But my strategies will be running on M15 + and based on closing prices. They are not HFT strategies based on every tick. Agreed there will be differences from broker to broker but if I’m within 10/15% on the initial coarse backtests that’s great and then I will then test with tick data. If a strategy works with one broker but not another it’s not robust enough. I will be keeping things simple but obviously want to be accurate as necessary. 

 
EdFuk #: Sure, I get this. But my strategies will be running on M15 + and based on closing prices. They are not HFT strategies based on every tick. Agreed there will be differences from broker to broker but if I’m within 10/15% on the initial coarse backtests that’s great and then I will then test with tick data. If a strategy works with one broker but not another it’s not robust enough. I will be keeping things simple but obviously want to be accurate as necessary. 

Even if it isn't HFT, there can still be differences even on hourly or even daily charts. You will have to test it on the data of different brokers to find out. Just relying on dukascopy data is not ideal.

 
Fernando Carreiro #:
broker's real tick data
Fernando Carreiro #:

If you use a reliable broker, they often provide quality historical data quite far back.

Using the broker's real tick data, also allows you test with real and floating spreads, instead of fixed spreads.

Don't be stuck in the MT4 mind-set. MT5 is different and you don't need 3rd party tools for testing with real and quality historical data.


hi. which broker you supposed to provide real tick data on mt5 ?

 
ma abbasi #:


hi. which broker you supposed to provide real tick data on mt5 ?

All major and reputable brokers do.
 
Dominik Christian Egert #:
All major and reputable brokers do.

any suggest ? (i contact some broker, they say, they dont have old data , lower time frame ,for example : M1 , from 2000-2023 )

i try alpari,liteforex,fxdd.!

thank's for help

 
ma abbasi #:

any suggest ? (i contact some broker, they say, they dont have old data , lower time frame ,for example : M1 , from 2000-2023 )

i try alpari,liteforex,fxdd.!

thank's for help

You were talking about tick data, now about timeframe data, it's different things. These brokers certainly have M1 data from 2000, but certainly not tick data.
123
New comment