Playback test results vary depending on how timeframes are organized.

[alrafer]

Hi NT team,

I have a strategy that generate some good results, depending on the time period of course, and I'd like to test it as a whole and by chunks (every 15 days for example).
However, when I pick a 90 day period and test it as a whole, or split into 6 15day periods the results are relatively different (very different if I use the Playback option). The chart is ES 03-22, I am using historical data afaik, 1000 ticks data series.

One thing I noticed is that in playback, some trades are somehow different depending on the Timeframe chosen (for example same trades in 12 days vs 45 days), even if I didn't change anything else - only the start/stop times. I know there will be discrepancies between Real Data and Historical, but not sure why there are when I chose diff start/end dates, for the same trades.

What could be the cause of these (sometimes big) discrepancies I see?

Thanks.

[NinjaTrader_Kate]

Hello alrafer,

Thank you for your post.

Are you using Break at EOD for the data series so the bars start at the same place each day? Does the strategy use Exit on Session Close or does it hold positions between sessions?

I would expect that if your strategy calculates historically that different results could be seen when using different start/end dates as you could have trades that would have started in a previous session that wouldn't be seen when changing the start date, so you could see entirely different entries and exits.

Thanks in advance; I look forward to assisting you further.

[alrafer]

Thanks for the insights Kate, it can be related to these configs you mention... I am using Break at EOD, yep, since it comes by default with the chart. And the strategy uses Exit on Session Close too.

Now that I look at these... That should be the best config to prevent these discrepancies in from happening, right?
Anything else I should check to avoid them, as much as possible?

[NinjaTrader_Kate]

Hello alrafer,

Thank you for your reply.

Keep in mind that when you're using Playback and have historical data loaded on the chart that the strategy will process over that data first, before the playback starts. If you're changing the date range of the playback the strategy will be starting from X number of days of historical data back which may influence trades that are taken when the strategy gets to the actual playback. Exit on Session close and Break at EOD should help, but I would not expect to see exactly identical trades when the strategy is started from different days.

Please let us know if we may be of further assistance to you.

[alrafer]

Thanks Kate, I'll run a test again (this time with Market Replay data) and I'll let you know if I have further questions.