Assume some sensor with specified “state_class: measurement” is updated every 30 sec.
According to this: statistics of hourly min, max and average sensor readings is updated every 5 minutes
Means:
Every 5 minutes HA analyses gathered data and calculates minimal, maximal & average values for the last hour.
Then these 3 calculated values are stored in DB and are kept forever.
Then these calculated (“statistical”) data may be displayed by statistics-graph card with a required precision (5minutes, hour etc).
But - here it is specified differently: Each hour it will take a snapshot of supported entities and track different things about the entity state. ... Because they are summarized every hour, they only create 24 entries per day.
So, what is correct:
statistics is calculated every 5 minutes;
statistics is calculated every hour
?
Assume that there are only 24 entries per day.
Then how are “5minutes” data available for the statistics graph card?
I guess that these words are wrong: Each hour it will take a snapshot of supported entities and track different things about the entity state. ... Because they are summarized every hour, they only create 24 entries per day.
For one of my statistics graphs that has months of data I can only see back 7 days of 5 minute data. Hourly data goes back the full record length. So does daily and monthly.
That means that:
– at least these data are collected every 5 minutes;
– these data are “purged” to see only the last 7 days - similarly to “conventional non-statistical” data which are kept for the last xxx days (accordingly to the Recorder’s settings, 10 days by default).
And this does not match to:
Assume there is a statistics for some sensor which is calculated every 5 minutes (i.e. “version 1” from my 1st post).
The “statistics-graph” card when used “period” option = 5minutes → shows all kept statistical data for this sensor w/o any processing - from the beginning (i.e. “all statistics”), not “last 7 days”.
When used with “period = hour” → then from the gathered statistics only “hourly” data are used - and here I do not know whether this is “just let’s take every 12th (60/5) readings” or this is “let’s recalculate min/max/avg for every hour based on 5-minutes-readings” - and again from the beginning.
I.e. the “statistics graph” card may either show all gathered stats (period=5minutes) or recalculated (period=hour,…). So, statistics is always gathered every 5 minutes based on the actual readings within last hour.
These “5 minutes” results are stored in DB (let’s call it “short-term statistics”).
Every hour these values (i.e. calculated at this moment) are stored in a “long-term” part of DB.
The “short-term statistics” persists within “purge interval” - i.e. only last 10 days (by default) may be displayed in a “statistics-graph” card.
The “long-term statistics” persists forever - i.e. whole “hourly” data may be displayed in a “statistics-graph” card.
If bigger periods (day, week, …) are selected for the “statistics-graph” card - then consolidated (i.e. recalculated) data based on that “hourly” data are displayed.
These docs (one, two) are still unclear in this part. That why there is a confusion “why 5minutes statistics is only displayed for last 10 days”.
I had the same question. I looked at my database and there is in fact a table called statistics_short_term which holds 5-minute-interval data going back to the time I have my recorder purge_keep_days set to (in my case 14 days), and there is a separate table called statistics which holds the hourly data going back to the date and hour I installed my first measurement sensor in home assistant. I did not see any other relevant statistics tables, so I suspect you are also correct about HA re-calculating the hourly data to come up with daily, monthly, yearly info.
