Difference between Generation and retesting
11 replies
Diego Michielan
4 years ago #257532
Hi I have a problem, I’m not able to understand why the EURUSD D1 strategies in generation phase and in the retest give extremely different resutls. I attach the screenshots and the strategy, could someone help?
Thanks a lto
Diego
tomas262
4 years ago #257543
Hello Diego,
before the strategy is retested do you load the strategy config into the Retester? Check the screen attached
You need to retest strategies using the very same settings they were built with
Diego Michielan
4 years ago #257548
Hi, thank you, I did and it seems now working. However testing in an higher TF I got the situation attached for all the strategies. Any clue why this could happen? I’ve also made a backtest in MT4 and the strategies produce trades.
Thanks again for your support.
Diego
hankeys
4 years ago #257553
you must be doing something wrong but hard to say what…
if you have the same settings in builder and retester the backtest is the same, there is no problem
did you have the same view turned on and see really the same info?
You want to be a profitable algotrader? We started using StrateQuant software in early 2014. For now we have a very big knowhow for building EAs for every possible types of markets. We share this knowhow, apps, tools and also all final strategies with real traders. If you want to join us, fill in the FORM.
Diego Michielan
4 years ago #257567
Hi I reset everything and now seems to be fine, the trades are generated, probably there were some mistakes in configuration but don’t know what.
Now the only small discrepancy I still notice is this one: during retesting, changing only the data range of histricial data I got small differences in results of the first year. This happens for all strategies, not just in some cases.
For example see screenshots below: same strategy, all same parameters into retester except for the data start day: in one case 01.05.2003 in the other 01.01.2011. I got different results in 2011. I’ve checked the list of trades and seems to be in the shorter period (starting from 2011) few trades at the beginning of the year are missed. I thought this could be due to some shifting periods in indicators in entry signals but this seems to be not the case either. the singals for the strategy in fact are
LongEntrySignal = (Close(Main chart)[1] crosses below HeikenAshiClose(Main chart)[1]);
ShortEntrySignal = (Close(Main chart)[1] crosses above HeikenAshiClose(Main chart)[1]);
Any previous experience in such behaviour?
THanks a lot.
Diego
mabi
4 years ago #257604
Diego Michielan
4 years ago #257606
Thanks but this is not the case. No pending orders and no indicator’s periods, below the pseudocode. This behaviour happens for all the strategies and different TF (I tested H1, H4, D1)
hankeys
4 years ago #257613
you are doing something strange…you didnt describe what are you doing…i am not getting any differences
You want to be a profitable algotrader? We started using StrateQuant software in early 2014. For now we have a very big knowhow for building EAs for every possible types of markets. We share this knowhow, apps, tools and also all final strategies with real traders. If you want to join us, fill in the FORM.
Diego Michielan
4 years ago #257615
I simply retest strategies with two backtest data setting. first retest: from 2003.05.05 til 2020.04.10, second retest from 2011.01.01 til 2020.04.10. No other changes and I got different result for 2011.
hankeys
4 years ago #257621
it could be simple as mabi said….you have open trade from 2010 to 2011 or the logic for entry is delayed…
how long did the difference last in 2011? is it question of some first trades?
because long period of bars valid could mean, that the logic of strategy could be different
You want to be a profitable algotrader? We started using StrateQuant software in early 2014. For now we have a very big knowhow for building EAs for every possible types of markets. We share this knowhow, apps, tools and also all final strategies with real traders. If you want to join us, fill in the FORM.
Diego Michielan
4 years ago #257660
From the ones I analysed 3/4 months (since March/April). As I said the strategy enters at market (no pending orders) and is based on Ichimoku, no indicators or periods which can apparently cause the delay. I start to believe this is general issue since it always show up in all the backtests. Does this not happen to you also? Tnx
hankeys
4 years ago #257677
i think i dont have this issue
You want to be a profitable algotrader? We started using StrateQuant software in early 2014. For now we have a very big knowhow for building EAs for every possible types of markets. We share this knowhow, apps, tools and also all final strategies with real traders. If you want to join us, fill in the FORM.
Viewing 11 replies - 1 through 11 (of 11 total)