"group_by: day": Oscillation in graph

[UltraBlackLinux]

Hey there,
I have this weird oscillation in one of three of my graphs.
See here:

Why is only one of the graphs doing this? They all share pretty much the same config file:

type: custom:mini-graph-card
name: Energy 3d
icon: mdi:clock
group_by: day
hours_to_show: 72
entities:
  - entity: sensor.plug_barge_delta

Any idea what this is about and how to resolve it? I appear to be on 0.13.0.
Thanks!

[ildar170975]

I doubt anyone can give you an explanation since this are YOUR data and it is only YOU who can verify a validity of source data.
So far this is not a BUG (until it is proven otherwise) - means this is a question.
Please ask questions in Community: https://community.home-assistant.io/t/lovelace-mini-graph-card/71385.

[UltraBlackLinux]

what confuses me here is that the graph oscillates even though the points that are next to each other have the exact same value. The other graph cards take the exact same data from a different device and process it in the exact same way.

for context, here's my config (I excluded the third one because it's more of the same):

sensor:
  - platform: statistics
    name: "Plug Auto Energy Hourly"
    entity_id: "sensor.plug_auto_energy"
    state_characteristic: change
    max_age:
        hours: 24
    precision: 3

  - platform: statistics
    name: "Plug Barge Energy Hourly"
    entity_id: "sensor.plug_barge_energy"
    state_characteristic: change
    max_age:
        hours: 24
    precision: 3

template:
  - triggers:
    - trigger: time_pattern
      minutes: 0
    sensor:
      - name: "Plug Auto Reference"
        state: "{{ states('sensor.plug_auto_energy') }}"

      - name: "Plug Barge Reference"
        state: "{{ states('sensor.plug_barge_energy') }}"

  - sensor:
    - name: "Plug Auto Delta"
      unit_of_measurement: "kWh"
      state: "{{ states('sensor.plug_auto_energy')|float - states('sensor.plug_auto_reference')|float }}"

    - name: "Plug Barge Delta"
      unit_of_measurement: "kWh"
      state: "{{ states('sensor.plug_barge_energy')|float - states('sensor.plug_barge_reference')|float }}"

I cannot think of a single reason why one of the graphs would be behaving this way.

If this still doesn't sound like a bug to you I guess I'll ask the community.

[ildar170975]

Look at the case from another side: there are 3 cards with same code, only different entities; and 1 case you find as "strange" - perhaps a particular data set from a particular entity causes a "strange" output. I am not sure that even other 2 cases are correct (you got a 3 days span and some curvy lines). Try playing with other options excluding an interpolation like smoothing, points_per_hour. And of course purge a browser cache.
You need to analyze source data manually - build stock history-graphs for source data & manually create "daily grouping" in Excel, then compare with mini-graph-card.

[UltraBlackLinux]

I frankly don't think that this is an issue with my data at all. I did some checking and it does seem correct. What I find weird is that the graph adds oscillations in there for really no apparent reason. The points are correct, the waves just shouldn't be there - it should be connecting the points with a single more or less straight line.

I did also play around with the data points but that doesn't solve my issue from what I can tell and it's more or less incompatible with the way I'm gathering the data.

[ildar170975]

The points are correct, the waves just shouldn't be there - it should be connecting the points with a single more or less straight line.

Imho any daily graphs for 3 days should have curves containing 3 "planes"...
Means - all your 3 cards look strange.
Let's keep this issue, we'll check it...

[UltraBlackLinux]

I just want to note that I just discovered the meter helper, which does exactly what I've been doing manually with my configuration above. I still however get the exact same oscillations:

The data points still contain identical values.

[worgarside]

FWIW shouldn't it be group_by: date instead of group_by: day? Not sure it'll solve the root issue but it might help group the datapoints together and reduce the oscillation

[UltraBlackLinux]

I'll give that a try, thank you!

[ildar170975]

@UltraBlackLinux
Sorry for a delay with a feedback.
Can you see it still?
Do you know a way to export data for these sensors from your DB to my DB ? Asking because will try to simulate it…

[UltraBlackLinux]

I had to wait a bit to spot some oscillations. Currently I'm just seeing some small ones because the device where it was most obvious started fluctuating. I'll try to get more definitive ones soon: history(1).csv

^^ That's friday the 17th, 11:04 AM to 05:04 PM. Three data points, all three on 0.07 kWh

[UltraBlackLinux]

history(2).csv noticed some rather big and clear oscillations again.

this is the 28th of october, 5:30pm to the 29th, 3:30pm