quant_engine/tests/test_strategy_optimizer.py

"""Bounded optimizer runs actual core strategies and rejects ambiguous ranking."""

import numpy as np
import pandas as pd
import pytest

from quant_engine import strategy_optimizer as optimizer
from quant_engine.strategy_contracts import STRATEGIES, strategy_parameters
from quant_engine.strategy_research import BenchmarkInput, run_strategy_research
from quant_engine import strategy_contracts


def feed(values=None):
    values = np.asarray(
        values if values is not None else 10 + 2 * np.sin(np.arange(80) / 2), dtype=float
    )
    return pd.DataFrame(
        dict.fromkeys(("open", "high", "low", "close"), values),
        index=pd.date_range("2026-01-01", periods=len(values), freq="B"),
    )


@pytest.mark.parametrize("name", STRATEGIES)
def test_each_default_strategy_uses_the_real_ledger(name):
    defaults = strategy_parameters(name)
    key = next(iter(defaults))
    result = optimizer.optimize_strategy_research(
        name,
        feed(),
        asset="SYNTHETIC",
        param_grid={key: [defaults[key]]},
        objective="total_return",
        initial_cash=1000,
        commission=0.01,
        stamp_duty=0.002,
    )
    direct = run_strategy_research(
        name, feed(), asset="SYNTHETIC", initial_cash=1000, commission=0.01, stamp_duty=0.002
    )
    assert len(result.trials) == 1
    assert result.trials[0].result.ledger.positions == direct.ledger.positions
    assert result.trials[0].result.ledger.daily_executions == direct.ledger.daily_executions
    assert result.trials[0].score == direct.metrics["total_return"]
    assert result.decision_eligible is False


def test_real_negative_zero_scores_and_explicit_costs_sort_without_defaults():
    result = optimizer.optimize_strategy_research(
        "BuyAndHold",
        feed([10, 10, 9, 8]),
        asset="SYNTHETIC",
        param_grid={"buy_pct": [0.5, 0, 0.25]},
        objective="total_return",
        initial_cash=1000,
        commission=0.01,
        stamp_duty=0.002,
    )
    assert [trial.parameters["buy_pct"] for trial in result.trials] == [0, 0.25, 0.5]
    assert [trial.score for trial in result.trials] == pytest.approx([0, -0.0525, -0.105])
    assert [trial.result.ledger.total_costs for trial in result.trials] == pytest.approx(
        [0, 2.5, 5]
    )


def test_stable_ties_retain_canonical_axis_and_candidate_order():
    result = optimizer.optimize_strategy_research(
        "MACross",
        feed([10] * 40),
        asset="SYNTHETIC",
        param_grid={"atr_period": [2, 0], "fast": [5, 4]},
        objective="total_return",
        commission=0,
        stamp_duty=0,
    )
    assert [
        (trial.parameters["fast"], trial.parameters["atr_period"]) for trial in result.trials
    ] == [(5, 2), (5, 0), (4, 2), (4, 0)]


def test_hundred_combinations_are_unique_actual_results():
    result = optimizer.optimize_strategy_research(
        "MACross",
        feed(),
        asset="SYNTHETIC",
        param_grid={"fast": list(range(1, 11)), "slow": list(range(11, 21))},
        objective="total_return",
        commission=0,
        stamp_duty=0,
    )
    assert len(result.trials) == 100
    assert len({tuple(trial.parameters.items()) for trial in result.trials}) == 100
    assert all(len(trial.result.ledger.positions) == 80 for trial in result.trials)


@pytest.mark.parametrize(
    "grid,objective",
    [
        ({"fast": []}, "total_return"),
        ({"fast": [5, 40]}, "total_return"),
        ({"fast": [True]}, "total_return"),
        ({"fast": [5]}, "unknown"),
        ({"fast": [5, 5]}, "total_return"),
        (
            {"fast": list(range(1, 11)), "slow": list(range(11, 21)), "atr_period": [0, 1]},
            "total_return",
        ),
    ],
)
def test_invalid_or_partly_invalid_grid_never_starts_a_strategy(monkeypatch, grid, objective):
    calls = []
    monkeypatch.setattr(optimizer, "run_strategy_research", lambda *a, **kw: calls.append(1))
    with pytest.raises(
        ValueError, match=r"grid|Grid|candidate|number|less than|objective|Duplicate"
    ):
        optimizer.optimize_strategy_research(
            "MACross", feed(), asset="SYNTHETIC", param_grid=grid, objective=objective
        )
    assert calls == []


def test_history_failure_for_one_candidate_prevents_all_runs(monkeypatch):
    calls = []
    monkeypatch.setattr(optimizer, "run_strategy_research", lambda *a, **kw: calls.append(1))
    with pytest.raises(ValueError, match=r"history"):
        optimizer.optimize_strategy_research(
            "MACross",
            feed([10] * 40),
            asset="SYNTHETIC",
            param_grid={"slow": [30, 100]},
            objective="total_return",
        )
    assert calls == []


def test_requested_benchmark_error_prevents_every_run(monkeypatch):
    calls = []
    monkeypatch.setattr(optimizer, "run_strategy_research", lambda *a, **kw: calls.append(1))
    with pytest.raises(ValueError, match=r"benchmark source"):
        optimizer.optimize_strategy_research(
            "BuyAndHold",
            feed(),
            asset="SYNTHETIC",
            param_grid={"buy_pct": [0.5, 1]},
            objective="total_return",
            benchmark=BenchmarkInput("source_error"),
        )
    assert calls == []


def test_undefined_sharpe_fails_the_ranking_instead_of_winning_as_zero():
    with pytest.raises(ValueError, match=r"objective.*unavailable"):
        optimizer.optimize_strategy_research(
            "BuyAndHold",
            feed([10] * 4),
            asset="SYNTHETIC",
            param_grid={"buy_pct": [0, 0.5]},
            objective="sharpe_ratio",
            commission=0,
            stamp_duty=0,
        )


def test_large_grid_fails_before_creating_cartesian_product(monkeypatch):
    calls = []
    monkeypatch.setattr(strategy_contracts, "product", lambda *a, **kw: calls.append("product"))
    monkeypatch.setattr(optimizer, "run_strategy_research", lambda *a, **kw: calls.append("run"))
    with pytest.raises(ValueError, match=r"limit"):
        optimizer.optimize_strategy_research(
            "MACross",
            feed(),
            asset="SYNTHETIC",
            param_grid={
                "fast": list(range(1, 11)),
                "slow": list(range(11, 21)),
                "atr_period": list(range(10)),
                "atr_mult": list(range(1, 11)),
            },
        )
    assert calls == []