SQX 130
Hello,
Tomas Matejka asked me to open a ticket here.
I’ve just encountered a strange difference between 1m test precision and real tick-real spread.
Despite the fact that i’ve used quite a strict spread setting for 1m test it is still even not close to real tick real spread result.
Please have a look,
I’ve attached everything here.
Probably it isn't a bug, but Tomas agreed that it looks odd.
Thank you in advance
Attachment to Tomasnzdchfspread.zip added
thank your for your investigation.
I found spread settings here. https://www.dukascopy.com/swiss/english/marketwatch/average-spreads/
It is stated that nzd chf pair has spread 1.9-2.1 more than 96% of time. So I've used 1.9 as a reliable one.Also I've made a quick backtest for you and there is a still a problem.
I am attaching those strategies here.
Attachment strategies.zip added
Hi Valery,
I have investigated this a little deeper and performed Monte Carlo tests with randomized spread.From monte carlo equity curves it is clear that first four strategies are sensitive to spread - the number of trades varies a lot depending on the spread used.
These four strategies also use a very similar entry price levels.
The last one - 4.12.429 is different, the number of trades is almost constant accross all spread settings and higher spreads only cause lower overall profit.
I attach the backtested strategies, you can take a look.
Best regards,
Tomas
If I compare backtests using minute data and real ticks with custom spread, the results are much more similar.
Looking at the tick data of NZDCHF I think the applied spread 1.9 pips is quite low, I can see 4 or 5 pips in the data itself.