36 lines
1.1 KiB
Python
36 lines
1.1 KiB
Python
from typing import Iterable
|
|
|
|
K_FACTOR = 10
|
|
BETA = 200
|
|
|
|
|
|
def rerank(ratings: Iterable[int], winning_player_idx: int) -> Iterable[int]:
|
|
expectations = _expectations(ratings)
|
|
return [
|
|
rating
|
|
+ (K_FACTOR * ((1 if winning_player_idx == idx else 0) - expectations[idx]))
|
|
for idx, rating in enumerate(ratings)
|
|
]
|
|
|
|
|
|
def _expectations(ratings: Iterable[int]) -> Iterable[int]:
|
|
return [
|
|
_calculate_expectation(rating, ratings[:idx] + ratings[idx + 1 :])
|
|
for idx, rating in enumerate(ratings)
|
|
]
|
|
|
|
|
|
def _calculate_expectation(rating: int, other_ratings: Iterable[int]) -> int:
|
|
return sum(
|
|
[_pairwise_expectation(rating, other_rating) for other_rating in other_ratings]
|
|
) / (float(len(other_ratings) + 1) * len(other_ratings) / 2)
|
|
|
|
|
|
def _pairwise_expectation(rating: int, other_rating: int) -> Iterable[int]:
|
|
"""
|
|
Gives the expected score of `rating` against `other_rating`
|
|
"""
|
|
diff = float(other_rating) - float(rating)
|
|
f_factor = 2 * BETA # rating disparity
|
|
return 1.0 / (1 + 10 ** (diff / f_factor))
|