import math

import numpy as np
import pandas as pd

from ktrade.backtest.metrics import block_bootstrap_ci, deflated_sharpe, drawdown_series
from ktrade.backtest.validation import param_grid, walk_forward
from ktrade.strategy.meanrev import ou_half_life
from ktrade.strategy.momentum import TimeSeriesMomentum
from tests.conftest import synthetic_candles


def test_drawdown():
    dd = drawdown_series(pd.Series([100, 110, 99, 120, 60]))
    assert math.isclose(dd.min(), 0.5 - 1)


def test_ou_half_life_detects_mean_reversion_and_random_walk():
    rng = np.random.default_rng(0)
    x = np.zeros(2000)
    for i in range(1, 2000):
        x[i] = 0.9 * x[i - 1] + rng.normal()
    hl = ou_half_life(pd.Series(x))
    assert 4 < hl < 10  # theory: -ln2/ln(0.9) ≈ 6.6
    rw = pd.Series(np.cumsum(rng.normal(size=2000)))
    assert math.isnan(ou_half_life(rw)) or ou_half_life(rw) > 100


def test_bootstrap_ci_contains_point_estimate():
    rng = np.random.default_rng(1)
    r = rng.normal(0.0005, 0.01, 2000)
    bpy = 365 * 24
    ci = block_bootstrap_ci(r, bpy, n_boot=200)
    s = r.mean() / r.std(ddof=1) * math.sqrt(bpy)
    assert ci["sharpe_ci"][0] <= s <= ci["sharpe_ci"][1]


def test_deflated_sharpe_decreases_with_trials():
    a = deflated_sharpe(1.5, 2000, 365 * 24, 1, 0.0, 0.0)["dsr"]
    b = deflated_sharpe(1.5, 2000, 365 * 24, 100, 0.0, 0.0)["dsr"]
    assert a > b


def test_walk_forward_produces_oos_folds():
    df = synthetic_candles(n=1500, seed=11, drift=0.0005)
    wf = walk_forward(TimeSeriesMomentum, df, interval_min=60, grid={"lookback": [24, 48], "ema_fast": [10], "ema_slow": [40]}, n_folds=3)
    assert len(wf["folds"]) == 3 and wf["n_trials"] == 2
    for f in wf["folds"]:
        assert f["train"][1] < f["test"][0]  # OOS strictly after IS
    assert len(param_grid(TimeSeriesMomentum, {"lookback": [1, 24]})) == 1  # 1 < min → dropped