Problems with backtesting of FX in .7

[drolles]

NT Support,

I’ve just upgraded .7 and I’ve got problems back testing my strategies. I’m getting errors with Forex position size values when using a fixed factional position sizing.

See screen shot attached. It looks to be around JPY crosses (see Strategy Analyser screen shot attached).

Has an old error been reintroduced? There was a problem around JPY crosses previously. We had to be connected to a price feed – something about cross rate calculation.


Could you please give us a view of the scripts you test to confirm a build is right to ship? What Fx strategies do you run through your builds to confirm they work?


Thanks and regards,

drolles

[NinjaTrader_Austin]

Hi Drolles, there was some change regarding $EURJPY.. I'm not sure if that would apply here but the change (as described in the release notes) is this -
"Fixed
4507
Performance Report
In currency mode, the report from a historical backtest on $EURJPY could have differed from the report of the same strategy ran on a live chart"

I will have someone get back to you on Monday with more clarification.

[drolles]

Austin,

Thanks for the reply.

As per the screen shot the errors not only appearing in EURJPY. My example here is AUDJPY. And I’m seeing other errors on JPY crosses.

Thanks and regards,

drolles

[NinjaTrader_Bertrand]

Hi drolles hope you're well, thanks for the report and clarification - we're looking into it.

[drolles]

Thanks for that.

Given this is a critical issue could we please have an update? Without a messy rollback process I can’t backtest now I’ve upgraded. I’ve resorted to my backup machine.

Thanks,

drolles

[NinjaTrader_Bertrand]

drolles, what version is your backup machine then running on? Could you please clarify?

[drolles]

Bertrand,

My backup install is running 7.0.1000.5.

As per the previous request can I please have an update here? What process is being used for issue? How many resources are deployed to fixing this? What has been found from your testing to date?

Thanks and regards,

drolles

[drolles]

NT Support,

Again, can we please have an update here? When are we going to see some answers to the questions raised?

Thanks,

drolles

[NinjaTrader_Bertrand]

Thanks for the clarifications Daniel, I've reproduced the issue going from the R6 to the R7, so this what you saw updating from your R5 to R7 - I've now asked development to take a look and will reply back as soon as I have more information, for now please stay on the R5 (if you would like to downgrade to it, please send me a PM or email to support).

To clarify another ithem : you would also see no conversion in backtesting, here the execution rate seen would be 1 and that would be expected (which is in contrast to live execution where the needed rate can be streamed in provided the MA / Chart is pulling the data in).

[drolles]

Bertrand,

Thanks for your effort on working through this. I see we have taken a big step that this is a reproducible error.

Could we have an update please?

Thanks and regards,

drolles

[NinjaTrader_Matthew]

drolles,

Bertrand should be follow up with you on Monday.

Thanks for your patience and enjoy your weekend.

[NinjaTrader_Bertrand]

Hi Daniel, development is out for the holiday weekend - once I have an update this week I'll get back to you promptly here.

Thanks for your patience and report,

[NinjaTrader_Bertrand]

Daniel, thanks for the patience here - we closely looked into matters and actually R7 is working correctly while R6 and prior versions unfortunately had a bug in this respect - in backtesting there's simply no conversion done to any PnL generated (as opposed to trading on live data), NT7 would use the lotsize, point value and ticksize set to arrive at the PnL here, which according to this definition is correct for the results we see. Apologies this took so long to get to the bottom of things.

[drolles]

Bertrand,

I’m sorry I’m lost.

I’ve been doing some testing on this tonight, are you saying that in backtest now uses lot and not an outright purchase as it appeared to be using in .6?

If so that doesn’t make sense. I’ve been doing some testing with this tonight. Why would a position size of 10000 behave like a mini-lot? Is that correct?

I’ve shown two screen shots of a backtest using .7 which gives some funny results if it is using lot size.

This just doesn’t make any sense as it looks as if the lot sizing hasn’t changed for USD crosses vis JPY crosses.

I’m sorry but I think this is getting silly now. This issue has been open 2 weeks now and NT still can’t give me an answer that appears to make sense.

Thanks and regards,

Dan

[drolles]

Bertrand,

This issue has now been open for two weeks.

I think it is time to escalate this. I effectively can’t backtest on my development machine while this issue is outstanding, which is just unacceptable.

If we can’t get this fix I would like someone from sales to explain to me while I shouldn’t receive money back for my license purchase if I’m not able to use the software that I’ve purchased for the purpose it is intended.

Thanks and regards,

Dan