Writing a question that works¶

The difference between a question that tracks the world and one that sits near 0.5 is usually not the model. It is what you told it.

Tell it how to read the numbers¶

Jev makes a judgment. It does not compare numbers.

Hand it a reading of 1.2 and a threshold of 5 buried in prose and it will not work out which side the reading falls on. Measured on the laundry question at 1.2 W and at 1450 W, five runs per cell:

What the state carried	idle	running	separation
the readings alone	0.47	0.26	+0.21
the rule in `background:`	0.70	0.10	+0.60
the comparison done in the template	0.74	0.06	+0.69
both	0.72	0.04	+0.68

Either fix roughly triples the separation, and they do not stack, so do one.

Put the rule in the questionDo the comparison yourself

No Jinja needed. This is the field labelled Standing facts in the UI.

background: >-
  This machine draws under 5 W when idle and over 300 W while a programme runs.

Exact and free, and it hands over a conclusion in words.

state: >-
  The washing machine is
  {% if states('sensor.washing_machine_power') | float(0) < 5 %}
  drawing almost no power, which means it is idle
  {% else %}
  drawing {{ states('sensor.washing_machine_power') }} W, so a programme is running
  {% endif %}.

Placement is most of the effect¶

The same sentence put in the state rather than in the question measured +0.33, about half of what it is worth on the question.

So background: beats writing the same rule into your template.

Ask one thing¶

A question that measures two things at once answers near 0.5 with low confidence, because both halves are true some of the time.

Two questions wearing one coat

Is the laundry finished and has nobody dealt with it?

One question

Is the laundry finished but still sitting in the machine?

If you need both, ask both. They go in the same request and cost almost no extra time.

A note beside the readings¶

A target on its own sends readings. Adding a note puts your text alongside them.

Asking whether the laundry was finished, about a power sensor and a door sensor, returned 0.31 with the readings alone and 0.80 after one sentence saying the programme had finished 14 minutes ago.

The model cannot know what your sensors do not say.

Use the preview¶

Everything above is guesswork until you look. The preview shows the exact state and the answer it gets, before the question exists. Change a word, look again.