example configs.

helping_ikbensuper.md

The hacky, not-reccomended, may break in future solution would be this

platform: bayesian
prior: 0.6
name: occupancy_livingroom_0
probability_threshold: 0.87
observations:
  - entity_id: "input_boolean.occupancy_override_livingroom_0"
    prob_given_true: 0.9
    prob_given_false: 0.1
    platform: "state"
    to_state: "on"
  - entity_id: "input_boolean.occupancy_override_livingroom_0"
    prob_given_true: 0.5
    prob_given_false: 0.5
    platform: "state"
    to_state: "off"
  - entity_id: "binary_sensor.pir_sensor_livingroom_0"
    prob_given_true: 0.45
    prob_given_false: 0.2
    platform: "state"
    to_state: "on"
  - entity_id: "binary_sensor.pir_sensor_livingroom_0"
    prob_given_true: 0.5
    prob_given_false: 0.5
    platform: "state"
    to_state: "off"
  - entity_id: "binary_sensor.dop_sensor_livingroom_0"
    prob_given_true: 0.45
    prob_given_false: 0.2
    platform: "state"
    to_state: "on"
  - entity_id: "binary_sensor.dop_sensor_livingroom_0"
    prob_given_true: 0.5
    prob_given_false: 0.5
    platform: "state"
    to_state: "off"

platform: oldbayesian
prior: 0.6
name: home_precense_ben
probability_threshold: 0.85
observations:
  - entity_id: "device_tracker.ben"
    prob_given_true: 0.9
    prob_given_false: 0.2
    platform: "state"
    to_state: "home"
  - entity_id: "device_tracker.ben"
    prob_given_true: 0.5
    prob_given_false: 0.5
    platform: "state"
    to_state: "not_home"
  - entity_id: "input_boolean.home_precense_homekit_ben"
    prob_given_true: 0.9
    prob_given_false: 0.2
    platform: "state"
    to_state: "on"
  - entity_id: "input_boolean.home_precense_homekit_ben"
    prob_given_true: 0.5
    prob_given_false: 0.5
    platform: "state"
    to_state: "off"

helpingAltycoder.md

The hacky, not-reccomended, may break in future solution would be this

  - name: 'family'
    platform: bayesian
    prior: 0.83
    probability_threshold: 0.86
    observations:
      - entity_id: 'device_tracker.nmap_person1'
        prob_given_true: 0.95
        prob_given_false: 0.125
        platform: 'state'
        to_state: 'home'
      - entity_id: 'device_tracker.nmap_person1' # this duplicate entry for not_home is uninformative so bayesian effectively ingores it.
        prob_given_true: 0.5
        prob_given_false: 0.5
        platform: 'state'
        to_state: 'not_home'
      - entity_id: 'device_tracker.person1phoneapp'
        prob_given_true: 0.8
        prob_given_false: 0.5
        platform: 'state'
        to_state: 'home'
      - entity_id: 'device_tracker.person1phoneapp'
        prob_given_true: 0.5
        prob_given_false: 0.5
        platform: 'state'
        to_state: 'not_home'
      - entity_id: 'device_tracker.life360_person1'
        prob_given_true: 0.6
        prob_given_false: 0.5
        platform: 'state'
        to_state: 'home'
      - entity_id: 'device_tracker.life360_person1'
        prob_given_true: 0.5
        prob_given_false: 0.5
        platform: 'state'
        to_state: 'not_home'

And so on

@Altycoder, Note how there are 2 entries for each entity? Configuring an uninformative entry for the off to_state for every entity is the difference. This is essentially what the old logic was (incorrectly) doing in the background before. It makes zero sense mathematically but if it works on your config....

- entity_id: "input_boolean.occupancy_override_livingroom_0"
    prob_given_true: 0.5
    prob_given_false: 0.5
    platform: "state"
    to_state: "off"