Historical data
-
I have received conflicting advice on the importance of the modelling quality number in my back tests. Some claim that you need 99% and tickstory can get this number for you if set it up correctly, however others claim that it is not important and all that matters is that you analyse your results to make sure the EA is doing what you have instructed. I would also like to query if you are best to backtest using ticks or candle close. Any opinions would be appreciated.
-
What's the best historical data to use for backtesting EAs?
-
Tickstory?
-
if for you are not enough only basic prices "open, close, high, low" and you need "all" ticks for backtest, this will be always as result, even with renko:
https://ctrlv.cz/shots/2016/11/05/x6fJ.png
-
look, it depends how is your EA created .... if you are working with candle[0] (it is current candle, not closed) - for this case also data from tickstory are not enough for you, you need do forward test on demo (few months) to see, if it is doing right (profit) ..... but for basic functionality (only if it is opening/closing on conditions and other criteria) you dont need 99%, or are you?
if you are working with candle[1], 2, 3, x - which are closed candles - here you also dont need 99%, why? .... if you are working with renko candles and you are generating them from tickstory data, they will be not complete, I am 111111% sure, that they will not be complete on 99%, because if you know how is renko generator working, you will be know that you need left this generator oagain on forward to generate data from real ticks, or again forward test on demo ....Between trading and gambling is a very small gap, be careful.
-
OK @miro1360, that makes sense. I mostly use candle 1 close. My analysis seems to accurately reflect what I am trying to achieve and I don't get 99% accuracy so I think I am good to go... I have found a variance between results on candle close and tick data, I will continue to forward test to find out which has greater consistence.
