Description

Not sure if this user error or an actual bug, but I am seeing this behaviour with all models (not just PlackettLuce):

When trying to simulate draws in 1:1 matches, either both players gain rating or 1 looses and 1 gains rating. I would have expected the ratings to be the same.

The documentation implies that ties are supported by the models.

Version

6.1.3

Minimal Reproduction

from openskill.models import PlackettLuce

model = PlackettLuce()

player_1 = model.rating()
player_2 = model.rating()

print(player_1)
print(player_2)

# with [0, 0] player_1 always gains rating, player_2 always loses rating
# with [x, x] | x != 0 both players gain rating
scores = [1, 1]

new_ratings = model.rate([[player_1], [player_2]], scores=scores)

print(new_ratings)

Outputs

[[PlackettLuceRating(mu=27.635389493140497, sigma=8.06590141354368)], [PlackettLuceRating(mu=27.635389493140497, sigma=8.06590141354368)]]

Version

v6