resample_to_interval date shifted

[ykoretskyi]

resample_to_interval date shifted

[xmatthias]

It's correct as is (but it's often misunderstood - and most likely a huge cause of bugs, too).

since all date columns are candle open columns - the proper merge for a :40 candle (goes from :40-:45) is the :44 candle (going from :44-:45).
To simplify thinking - make a list of "when will each (short and long) candle complete" - then it may make more sense.

Your "how it should be" uses a 5m candle (from :40-:45) at :41 (when then :40-:41 candle is finished) before the candle is complete. The only value you can reliably use from this candle at this point in time is the candle's open - which is usually not what you need (and which freqtrade won't give you in live mode, either).

For a more detailed explanation, please refer to this answer i gave a few weeks ago #268 (comment)

---

As a side note: your data doesn't seem to make much sense overall (ignoring timestamps for a moment).
your volume at 5m is lower than at 1m - which is surely not correct (and also ohlc values don't really align with the 1m data).
As such - you should consider rechecking the code you're using to generate the 5m data - maybe compare it to the usage sample.