Market Replay depends on Days To Load

[ATI user]

code, data, output sent attn: PatrickH

note that steps for reproducing an issue is in the code..with remarks on results

this does not address the main problem which is replaying code on diff days and getting diff results from same code and same data...this is the biggest problem as I can not reproduce original results with original code as replay changes

however...maybe this bug will reveal the problem with the other.

[NinjaTrader_Bertrand]

Thanks ATI user, we will check into and get back to you with our results.

[NinjaTrader_Bertrand]

Thanks for the replay files sent in, we're testing on our end.

[NinjaTrader_Bertrand]

ATI_user, I understand Patrick was working via email further with you - please expect an update via this route. Thanks.

[ATI user]

I sent Patrick an email on another issue with replay..i.e. it fills on the opening price of a bar...which is impossible live...real fills are at least one tick off market price for market order...simulated fills can not be at opening tick price

[NinjaTrader_Matthew]

Thanks ATI - we have received and are reviewing. Please let us know if you have anything else you wish to add to these reports.

[ATI user]

replay issue is definitely related to how ninja syncs/correlates historical data with replay data ...i.e. ninja needs enough historical data to run replay data

....For example, if you have replay-formatted data from 20120813 through 20120928 (34 trading days plus weekends is over 50 days), then you would connect to live feeds for Zenfire primary and secondary, if any, and set the chart instrument to the current contract month. Then right-click, select DataSeries and set the DaysToLoad to ’99′. This includes the 33 days recommended for replay DaysToLoad plus 66 additional historical days. The result is that ninja downloads enough historical data from both feeds to extend more than 33 days beyond your earliest replay-formatted data from 20120813. Then disconnect both feeds, connect to Market Replay and change the DaysToLoad setting back to 33. Now you can replay any date in your range. Again, it is best to work backwards chronologically so Ninja does not lose its correlation between the historical and replay data.

[ATI user]

a further thought on my last post and the biggest issue (to me) of different results on different days replaying same date/data with same code...

since replay and historical seem (to me) to be closely related to the problem, and since with each new live day the amount of historical data changes (i.e. ninja loads another day of historical than it did yesterday), the relationship between the stored replay-formatted data (which has not changed) and the new historical data is changed... and this change could be why I can not get the same results I did yesterday

the only way to test this would be to have replay ignore historical and just run the replay-formatted data which is constant.... in the case of my strategy, it does not use historical data...so I should get exactly the same results every time...on any day... which is essential for developing code

there is no doubt in my mind that the historical data is the variable that is causing the replay problems... this because changing the days to load has a big effect on replay results...or even whether replay works or not....this also since by going backwards chronologically replay works.. possibly because the historical data is not being changed? i.e. works during the test day/session but not between test days/sessions

[NinjaTrader_PatrickH]

Hello ATI user,

Thank you for this report.

We will continue to review this matter with other cases we receive on the correlation between historical and Market Replay data.

If you come across a specific scenario where we can re-create the same scenario please advise us of that information.

Please let me know if I may be of further assistance.

[ATI user]

it has been a while...any resolution?...is there a fix coming?

[NinjaTrader_PatrickH]

Hello ATI user,

Thank you for your response on this matter

Have you come across a specific scenario that we would be able to re-create on our end? If you come across a specific scenario where we can re-create the same scenario please advise us of that information.

[ATI user]

Hello ATI user,

Thank you for your response on this matter

Have you come across a specific scenario that we would be able to re-create on our end? If you come across a specific scenario where we can re-create the same scenario please advise us of that information.

thought you sent me an email weeks ago saying you were able to reproduce the problem 'exactly' and were working with others to resolve....is that no longer the case? or are you referring to a different issue?

[NinjaTrader_PatrickH]

Hello ATI User,

Thank you for your update on this matter.

thought you sent me an email weeks ago saying you were able to reproduce the problem 'exactly' and were working with others to resolve....is that no longer the case? or are you referring to a different issue?

Yes I was able to reproduce the scenario we discussed before where if you use Close[0] and start Replay on exactly the second of the start of the session you can get incorrect results due to the fact that NinjaTrader builds a minute bar from hh:mm:00 to hh:mm:59.
Starting at hh:mm:00 will actually be one second into the bar, remember to start the Replay a few seconds before the session start time.

Please let me know if I may be of further assistance.

[ATI user]

I have noticed no difference starting a second after the session start

[NinjaTrader_PatrickH]

Hello ATI user,

I have noticed no difference starting a second after the session start

Please start the Replay a few seconds before the session start time in order to pull the correct Close[0] on session start.

Please let me know if I may be of further assistance.