Buckets:

download
raw
3.86 kB
"""Tests for the ELO rating system — calculation, K-factor, edge cases."""
from __future__ import annotations
import pytest
from tradearena.core.elo import (
ESTABLISHED_THRESHOLD,
K_FACTOR_ESTABLISHED,
K_FACTOR_NEW,
EloResult,
calculate_elo_change,
expected_score,
k_factor,
)
class TestKFactor:
def test_new_bot_gets_high_k(self):
assert k_factor(0) == K_FACTOR_NEW
assert k_factor(29) == K_FACTOR_NEW
def test_established_bot_gets_low_k(self):
assert k_factor(30) == K_FACTOR_ESTABLISHED
assert k_factor(100) == K_FACTOR_ESTABLISHED
def test_threshold_boundary(self):
assert k_factor(ESTABLISHED_THRESHOLD - 1) == K_FACTOR_NEW
assert k_factor(ESTABLISHED_THRESHOLD) == K_FACTOR_ESTABLISHED
class TestExpectedScore:
def test_equal_ratings(self):
assert expected_score(1200, 1200) == pytest.approx(0.5)
def test_higher_rated_favored(self):
es = expected_score(1400, 1200)
assert es > 0.5
assert es < 1.0
def test_lower_rated_underdog(self):
es = expected_score(1200, 1400)
assert es < 0.5
assert es > 0.0
def test_symmetry(self):
ea = expected_score(1200, 1400)
eb = expected_score(1400, 1200)
assert ea + eb == pytest.approx(1.0)
def test_400_point_gap(self):
# 400 points difference = ~91% expected for the higher rated
es = expected_score(1600, 1200)
assert es == pytest.approx(0.9091, abs=0.01)
class TestCalculateEloChange:
def test_equal_ratings_winner_gains(self):
r1, r2 = calculate_elo_change(1200, 1200, 1.0, 0, 0)
assert r1.new_rating > 1200
assert r2.new_rating < 1200
assert r1.change > 0
assert r2.change < 0
def test_equal_ratings_draw_no_change(self):
r1, r2 = calculate_elo_change(1200, 1200, 0.5, 0, 0)
assert r1.new_rating == pytest.approx(1200, abs=0.01)
assert r2.new_rating == pytest.approx(1200, abs=0.01)
def test_upset_gives_bigger_gain(self):
# Low-rated player (1000) beats high-rated player (1400)
r1_upset, _ = calculate_elo_change(1000, 1400, 1.0, 0, 0)
# High-rated player (1400) beats low-rated player (1000) — expected
r1_expected, _ = calculate_elo_change(1400, 1000, 1.0, 0, 0)
assert r1_upset.change > r1_expected.change
def test_established_players_change_less(self):
# New players (K=32)
r1_new, _ = calculate_elo_change(1200, 1200, 1.0, 0, 0)
# Established players (K=16)
r1_est, _ = calculate_elo_change(1200, 1200, 1.0, 50, 50)
assert abs(r1_new.change) > abs(r1_est.change)
def test_new_vs_established_asymmetric_k(self):
# New bot (K=32) vs established (K=16) — both at 1200, new bot wins
r1, r2 = calculate_elo_change(1200, 1200, 1.0, 5, 50)
# New bot gains more than established bot loses (different K-factors)
assert r1.change == pytest.approx(16.0, abs=0.01) # K=32 * 0.5
assert r2.change == pytest.approx(-8.0, abs=0.01) # K=16 * -0.5
def test_returns_elo_result_dataclass(self):
r1, r2 = calculate_elo_change(1200, 1200, 1.0, 0, 0)
assert isinstance(r1, EloResult)
assert isinstance(r2, EloResult)
assert r1.old_rating == 1200
assert r2.old_rating == 1200
def test_large_elo_gap_minimal_change_for_expected_win(self):
# 1600 beats 800 — expected, so minimal ELO change
r1, r2 = calculate_elo_change(1600, 800, 1.0, 0, 0)
assert r1.change < 3 # almost no gain for expected win
assert r2.change > -3
def test_loss_decreases_rating(self):
r1, r2 = calculate_elo_change(1200, 1200, 0.0, 0, 0)
assert r1.new_rating < 1200
assert r2.new_rating > 1200

Xet Storage Details

Size:
3.86 kB
·
Xet hash:
f5a6eca4a6640c1e476687827201cb5b94654faf64d7bfaa252f592bfec16f7a

Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.